/src/gdal/ogr/ogrsf_frmts/generic/ogrlayerarrow.cpp
Line | Count | Source (jump to first uncovered line) |
1 | | /****************************************************************************** |
2 | | * |
3 | | * Project: OpenGIS Simple Features Reference Implementation |
4 | | * Purpose: Parts of OGRLayer dealing with Arrow C interface |
5 | | * Author: Even Rouault, <even dot rouault at spatialys.com> |
6 | | * |
7 | | ****************************************************************************** |
8 | | * Copyright (c) 2022-2023, Even Rouault <even dot rouault at spatialys.com> |
9 | | * |
10 | | * SPDX-License-Identifier: MIT |
11 | | ****************************************************************************/ |
12 | | |
13 | | #include "ogrsf_frmts.h" |
14 | | #include "ogr_api.h" |
15 | | #include "ogr_recordbatch.h" |
16 | | #include "ograrrowarrayhelper.h" |
17 | | #include "ogrlayerarrow.h" |
18 | | #include "ogr_p.h" |
19 | | #include "ogr_swq.h" |
20 | | #include "ogr_wkb.h" |
21 | | #include "ogr_p.h" |
22 | | #include "ogrlayer_private.h" |
23 | | |
24 | | #include "cpl_float.h" |
25 | | #include "cpl_json.h" |
26 | | #include "cpl_time.h" |
27 | | |
28 | | #include <algorithm> |
29 | | #include <cassert> |
30 | | #include <cinttypes> |
31 | | #include <limits> |
32 | | #include <utility> |
33 | | #include <set> |
34 | | |
35 | | constexpr const char *MD_GDAL_OGR_TYPE = "GDAL:OGR:type"; |
36 | | constexpr const char *MD_GDAL_OGR_ALTERNATIVE_NAME = |
37 | | "GDAL:OGR:alternative_name"; |
38 | | constexpr const char *MD_GDAL_OGR_COMMENT = "GDAL:OGR:comment"; |
39 | | constexpr const char *MD_GDAL_OGR_DEFAULT = "GDAL:OGR:default"; |
40 | | constexpr const char *MD_GDAL_OGR_SUBTYPE = "GDAL:OGR:subtype"; |
41 | | constexpr const char *MD_GDAL_OGR_WIDTH = "GDAL:OGR:width"; |
42 | | constexpr const char *MD_GDAL_OGR_UNIQUE = "GDAL:OGR:unique"; |
43 | | constexpr const char *MD_GDAL_OGR_DOMAIN_NAME = "GDAL:OGR:domain_name"; |
44 | | |
45 | | constexpr char ARROW_LETTER_BOOLEAN = 'b'; |
46 | | constexpr char ARROW_LETTER_INT8 = 'c'; |
47 | | constexpr char ARROW_LETTER_UINT8 = 'C'; |
48 | | constexpr char ARROW_LETTER_INT16 = 's'; |
49 | | constexpr char ARROW_LETTER_UINT16 = 'S'; |
50 | | constexpr char ARROW_LETTER_INT32 = 'i'; |
51 | | constexpr char ARROW_LETTER_UINT32 = 'I'; |
52 | | constexpr char ARROW_LETTER_INT64 = 'l'; |
53 | | constexpr char ARROW_LETTER_UINT64 = 'L'; |
54 | | constexpr char ARROW_LETTER_FLOAT16 = 'e'; |
55 | | constexpr char ARROW_LETTER_FLOAT32 = 'f'; |
56 | | constexpr char ARROW_LETTER_FLOAT64 = 'g'; |
57 | | constexpr char ARROW_LETTER_STRING = 'u'; |
58 | | constexpr char ARROW_LETTER_LARGE_STRING = 'U'; |
59 | | constexpr char ARROW_LETTER_BINARY = 'z'; |
60 | | constexpr char ARROW_LETTER_LARGE_BINARY = 'Z'; |
61 | | constexpr char ARROW_LETTER_DECIMAL = 'd'; |
62 | | constexpr char ARROW_2ND_LETTER_LIST = 'l'; |
63 | | constexpr char ARROW_2ND_LETTER_LARGE_LIST = 'L'; |
64 | | |
65 | | static inline bool IsStructure(const char *format) |
66 | 0 | { |
67 | 0 | return format[0] == '+' && format[1] == 's' && format[2] == 0; |
68 | 0 | } |
69 | | |
70 | | static inline bool IsMap(const char *format) |
71 | 0 | { |
72 | 0 | return format[0] == '+' && format[1] == 'm' && format[2] == 0; |
73 | 0 | } |
74 | | |
75 | | static inline bool IsFixedWidthBinary(const char *format) |
76 | 0 | { |
77 | 0 | return format[0] == 'w' && format[1] == ':'; |
78 | 0 | } |
79 | | |
80 | | static inline int GetFixedWithBinary(const char *format) |
81 | 0 | { |
82 | 0 | return atoi(format + strlen("w:")); |
83 | 0 | } |
84 | | |
85 | | static inline bool IsList(const char *format) |
86 | 0 | { |
87 | 0 | return format[0] == '+' && format[1] == ARROW_2ND_LETTER_LIST && |
88 | 0 | format[2] == 0; |
89 | 0 | } |
90 | | |
91 | | static inline bool IsLargeList(const char *format) |
92 | 0 | { |
93 | 0 | return format[0] == '+' && format[1] == ARROW_2ND_LETTER_LARGE_LIST && |
94 | 0 | format[2] == 0; |
95 | 0 | } |
96 | | |
97 | | static inline bool IsFixedSizeList(const char *format) |
98 | 0 | { |
99 | 0 | return format[0] == '+' && format[1] == 'w' && format[2] == ':'; |
100 | 0 | } |
101 | | |
102 | | static inline int GetFixedSizeList(const char *format) |
103 | 0 | { |
104 | 0 | return atoi(format + strlen("+w:")); |
105 | 0 | } |
106 | | |
107 | | static inline bool IsDecimal(const char *format) |
108 | 0 | { |
109 | 0 | return format[0] == ARROW_LETTER_DECIMAL && format[1] == ':'; |
110 | 0 | } |
111 | | |
112 | | static inline bool IsBoolean(const char *format) |
113 | 0 | { |
114 | 0 | return format[0] == ARROW_LETTER_BOOLEAN && format[1] == 0; |
115 | 0 | } |
116 | | |
117 | | static inline bool IsInt8(const char *format) |
118 | 0 | { |
119 | 0 | return format[0] == ARROW_LETTER_INT8 && format[1] == 0; |
120 | 0 | } |
121 | | |
122 | | static inline bool IsUInt8(const char *format) |
123 | 0 | { |
124 | 0 | return format[0] == ARROW_LETTER_UINT8 && format[1] == 0; |
125 | 0 | } |
126 | | |
127 | | static inline bool IsInt16(const char *format) |
128 | 0 | { |
129 | 0 | return format[0] == ARROW_LETTER_INT16 && format[1] == 0; |
130 | 0 | } |
131 | | |
132 | | static inline bool IsUInt16(const char *format) |
133 | 0 | { |
134 | 0 | return format[0] == ARROW_LETTER_UINT16 && format[1] == 0; |
135 | 0 | } |
136 | | |
137 | | static inline bool IsInt32(const char *format) |
138 | 0 | { |
139 | 0 | return format[0] == ARROW_LETTER_INT32 && format[1] == 0; |
140 | 0 | } |
141 | | |
142 | | static inline bool IsUInt32(const char *format) |
143 | 0 | { |
144 | 0 | return format[0] == ARROW_LETTER_UINT32 && format[1] == 0; |
145 | 0 | } |
146 | | |
147 | | static inline bool IsInt64(const char *format) |
148 | 0 | { |
149 | 0 | return format[0] == ARROW_LETTER_INT64 && format[1] == 0; |
150 | 0 | } |
151 | | |
152 | | static inline bool IsUInt64(const char *format) |
153 | 0 | { |
154 | 0 | return format[0] == ARROW_LETTER_UINT64 && format[1] == 0; |
155 | 0 | } |
156 | | |
157 | | static inline bool IsFloat16(const char *format) |
158 | 0 | { |
159 | 0 | return format[0] == ARROW_LETTER_FLOAT16 && format[1] == 0; |
160 | 0 | } |
161 | | |
162 | | static inline bool IsFloat32(const char *format) |
163 | 0 | { |
164 | 0 | return format[0] == ARROW_LETTER_FLOAT32 && format[1] == 0; |
165 | 0 | } |
166 | | |
167 | | static inline bool IsFloat64(const char *format) |
168 | 0 | { |
169 | 0 | return format[0] == ARROW_LETTER_FLOAT64 && format[1] == 0; |
170 | 0 | } |
171 | | |
172 | | static inline bool IsString(const char *format) |
173 | 0 | { |
174 | 0 | return format[0] == ARROW_LETTER_STRING && format[1] == 0; |
175 | 0 | } |
176 | | |
177 | | static inline bool IsLargeString(const char *format) |
178 | 0 | { |
179 | 0 | return format[0] == ARROW_LETTER_LARGE_STRING && format[1] == 0; |
180 | 0 | } |
181 | | |
182 | | static inline bool IsBinary(const char *format) |
183 | 0 | { |
184 | 0 | return format[0] == ARROW_LETTER_BINARY && format[1] == 0; |
185 | 0 | } |
186 | | |
187 | | static inline bool IsLargeBinary(const char *format) |
188 | 0 | { |
189 | 0 | return format[0] == ARROW_LETTER_LARGE_BINARY && format[1] == 0; |
190 | 0 | } |
191 | | |
192 | | static inline bool IsTimestampInternal(const char *format, char chType) |
193 | 0 | { |
194 | 0 | return format[0] == 't' && format[1] == 's' && format[2] == chType && |
195 | 0 | format[3] == ':'; |
196 | 0 | } |
197 | | |
198 | | static inline bool IsTimestampSeconds(const char *format) |
199 | 0 | { |
200 | 0 | return IsTimestampInternal(format, 's'); |
201 | 0 | } |
202 | | |
203 | | static inline bool IsTimestampMilliseconds(const char *format) |
204 | 0 | { |
205 | 0 | return IsTimestampInternal(format, 'm'); |
206 | 0 | } |
207 | | |
208 | | static inline bool IsTimestampMicroseconds(const char *format) |
209 | 0 | { |
210 | 0 | return IsTimestampInternal(format, 'u'); |
211 | 0 | } |
212 | | |
213 | | static inline bool IsTimestampNanoseconds(const char *format) |
214 | 0 | { |
215 | 0 | return IsTimestampInternal(format, 'n'); |
216 | 0 | } |
217 | | |
218 | | static inline bool IsTimestamp(const char *format) |
219 | 0 | { |
220 | 0 | return IsTimestampSeconds(format) || IsTimestampMilliseconds(format) || |
221 | 0 | IsTimestampMicroseconds(format) || IsTimestampNanoseconds(format); |
222 | 0 | } |
223 | | |
224 | | static inline const char *GetTimestampTimezone(const char *format) |
225 | 0 | { |
226 | 0 | return IsTimestamp(format) ? format + strlen("tm?:") : ""; |
227 | 0 | } |
228 | | |
229 | | /************************************************************************/ |
230 | | /* TestBit() */ |
231 | | /************************************************************************/ |
232 | | |
233 | | inline bool TestBit(const uint8_t *pabyData, size_t nIdx) |
234 | 0 | { |
235 | 0 | return (pabyData[nIdx / 8] & (1 << (nIdx % 8))) != 0; |
236 | 0 | } |
237 | | |
238 | | /************************************************************************/ |
239 | | /* SetBit() */ |
240 | | /************************************************************************/ |
241 | | |
242 | | inline void SetBit(uint8_t *pabyData, size_t nIdx) |
243 | 0 | { |
244 | 0 | pabyData[nIdx / 8] |= (1 << (nIdx % 8)); |
245 | 0 | } |
246 | | |
247 | | /************************************************************************/ |
248 | | /* UnsetBit() */ |
249 | | /************************************************************************/ |
250 | | |
251 | | inline void UnsetBit(uint8_t *pabyData, size_t nIdx) |
252 | 0 | { |
253 | 0 | pabyData[nIdx / 8] &= uint8_t(~(1 << (nIdx % 8))); |
254 | 0 | } |
255 | | |
256 | | /************************************************************************/ |
257 | | /* DefaultReleaseSchema() */ |
258 | | /************************************************************************/ |
259 | | |
260 | | static void OGRLayerReleaseSchema(struct ArrowSchema *schema, |
261 | | bool bFullFreeFormat) |
262 | 0 | { |
263 | 0 | CPLAssert(schema->release != nullptr); |
264 | 0 | if (bFullFreeFormat || STARTS_WITH(schema->format, "w:") || |
265 | 0 | STARTS_WITH(schema->format, "tsm:")) |
266 | 0 | { |
267 | 0 | CPLFree(const_cast<char *>(schema->format)); |
268 | 0 | } |
269 | 0 | CPLFree(const_cast<char *>(schema->name)); |
270 | 0 | CPLFree(const_cast<char *>(schema->metadata)); |
271 | 0 | if (schema->children) |
272 | 0 | { |
273 | 0 | for (int i = 0; i < static_cast<int>(schema->n_children); ++i) |
274 | 0 | { |
275 | 0 | if (schema->children[i] && schema->children[i]->release) |
276 | 0 | { |
277 | 0 | schema->children[i]->release(schema->children[i]); |
278 | 0 | CPLFree(schema->children[i]); |
279 | 0 | } |
280 | 0 | } |
281 | 0 | CPLFree(schema->children); |
282 | 0 | } |
283 | 0 | if (schema->dictionary) |
284 | 0 | { |
285 | 0 | if (schema->dictionary->release) |
286 | 0 | { |
287 | 0 | schema->dictionary->release(schema->dictionary); |
288 | 0 | CPLFree(schema->dictionary); |
289 | 0 | } |
290 | 0 | } |
291 | 0 | schema->release = nullptr; |
292 | 0 | } |
293 | | |
294 | | static void OGRLayerPartialReleaseSchema(struct ArrowSchema *schema) |
295 | 0 | { |
296 | 0 | OGRLayerReleaseSchema(schema, /* bFullFreeFormat = */ false); |
297 | 0 | } |
298 | | |
299 | | static void OGRLayerFullReleaseSchema(struct ArrowSchema *schema) |
300 | 0 | { |
301 | 0 | OGRLayerReleaseSchema(schema, /* bFullFreeFormat = */ true); |
302 | 0 | } |
303 | | |
304 | | /** Release a ArrowSchema. |
305 | | * |
306 | | * To be used by driver implementations that have a custom GetArrowStream() |
307 | | * implementation. |
308 | | * |
309 | | * @param schema Schema to release. |
310 | | * @since GDAL 3.6 |
311 | | */ |
312 | | |
313 | | void OGRLayer::ReleaseSchema(struct ArrowSchema *schema) |
314 | 0 | { |
315 | 0 | OGRLayerPartialReleaseSchema(schema); |
316 | 0 | } |
317 | | |
318 | | /************************************************************************/ |
319 | | /* AddDictToSchema() */ |
320 | | /************************************************************************/ |
321 | | |
322 | | static void AddDictToSchema(struct ArrowSchema *psChild, |
323 | | const OGRCodedFieldDomain *poCodedDomain) |
324 | 0 | { |
325 | 0 | const OGRCodedValue *psIter = poCodedDomain->GetEnumeration(); |
326 | 0 | int nLastCode = -1; |
327 | 0 | int nCountNull = 0; |
328 | 0 | uint32_t nCountChars = 0; |
329 | 0 | for (; psIter->pszCode; ++psIter) |
330 | 0 | { |
331 | 0 | if (CPLGetValueType(psIter->pszCode) != CPL_VALUE_INTEGER) |
332 | 0 | { |
333 | 0 | return; |
334 | 0 | } |
335 | 0 | int nCode = atoi(psIter->pszCode); |
336 | 0 | if (nCode <= nLastCode || nCode - nLastCode > 100) |
337 | 0 | { |
338 | 0 | return; |
339 | 0 | } |
340 | 0 | for (int i = nLastCode + 1; i < nCode; ++i) |
341 | 0 | { |
342 | 0 | nCountNull++; |
343 | 0 | } |
344 | 0 | if (psIter->pszValue != nullptr) |
345 | 0 | { |
346 | 0 | const size_t nLen = strlen(psIter->pszValue); |
347 | 0 | if (nLen > std::numeric_limits<uint32_t>::max() - nCountChars) |
348 | 0 | return; |
349 | 0 | nCountChars += static_cast<uint32_t>(nLen); |
350 | 0 | } |
351 | 0 | else |
352 | 0 | nCountNull++; |
353 | 0 | nLastCode = nCode; |
354 | 0 | } |
355 | | |
356 | 0 | auto psChildDict = static_cast<struct ArrowSchema *>( |
357 | 0 | CPLCalloc(1, sizeof(struct ArrowSchema))); |
358 | 0 | psChild->dictionary = psChildDict; |
359 | 0 | psChildDict->release = OGRLayerPartialReleaseSchema; |
360 | 0 | psChildDict->name = CPLStrdup(poCodedDomain->GetName().c_str()); |
361 | 0 | psChildDict->format = "u"; |
362 | 0 | if (nCountNull) |
363 | 0 | psChildDict->flags = ARROW_FLAG_NULLABLE; |
364 | 0 | } |
365 | | |
366 | | /************************************************************************/ |
367 | | /* DefaultGetArrowSchema() */ |
368 | | /************************************************************************/ |
369 | | |
370 | | /** Default implementation of the ArrowArrayStream::get_schema() callback. |
371 | | * |
372 | | * To be used by driver implementations that have a custom GetArrowStream() |
373 | | * implementation. |
374 | | * |
375 | | * @since GDAL 3.6 |
376 | | */ |
377 | | int OGRLayer::GetArrowSchema(struct ArrowArrayStream *, |
378 | | struct ArrowSchema *out_schema) |
379 | 0 | { |
380 | 0 | const bool bIncludeFID = CPLTestBool( |
381 | 0 | m_aosArrowArrayStreamOptions.FetchNameValueDef("INCLUDE_FID", "YES")); |
382 | 0 | const bool bDateTimeAsString = m_aosArrowArrayStreamOptions.FetchBool( |
383 | 0 | GAS_OPT_DATETIME_AS_STRING, false); |
384 | 0 | memset(out_schema, 0, sizeof(*out_schema)); |
385 | 0 | out_schema->format = "+s"; |
386 | 0 | out_schema->name = CPLStrdup(""); |
387 | 0 | out_schema->metadata = nullptr; |
388 | 0 | auto poLayerDefn = GetLayerDefn(); |
389 | 0 | const int nFieldCount = poLayerDefn->GetFieldCount(); |
390 | 0 | const int nGeomFieldCount = poLayerDefn->GetGeomFieldCount(); |
391 | 0 | const int nChildren = 1 + nFieldCount + nGeomFieldCount; |
392 | |
|
393 | 0 | out_schema->children = static_cast<struct ArrowSchema **>( |
394 | 0 | CPLCalloc(nChildren, sizeof(struct ArrowSchema *))); |
395 | 0 | int iSchemaChild = 0; |
396 | 0 | if (bIncludeFID) |
397 | 0 | { |
398 | 0 | out_schema->children[iSchemaChild] = static_cast<struct ArrowSchema *>( |
399 | 0 | CPLCalloc(1, sizeof(struct ArrowSchema))); |
400 | 0 | auto psChild = out_schema->children[iSchemaChild]; |
401 | 0 | ++iSchemaChild; |
402 | 0 | psChild->release = OGRLayer::ReleaseSchema; |
403 | 0 | const char *pszFIDName = GetFIDColumn(); |
404 | 0 | psChild->name = |
405 | 0 | CPLStrdup((pszFIDName && pszFIDName[0]) ? pszFIDName |
406 | 0 | : DEFAULT_ARROW_FID_NAME); |
407 | 0 | psChild->format = "l"; |
408 | 0 | } |
409 | 0 | for (int i = 0; i < nFieldCount; ++i) |
410 | 0 | { |
411 | 0 | const auto poFieldDefn = poLayerDefn->GetFieldDefn(i); |
412 | 0 | if (poFieldDefn->IsIgnored()) |
413 | 0 | { |
414 | 0 | continue; |
415 | 0 | } |
416 | | |
417 | 0 | out_schema->children[iSchemaChild] = static_cast<struct ArrowSchema *>( |
418 | 0 | CPLCalloc(1, sizeof(struct ArrowSchema))); |
419 | 0 | auto psChild = out_schema->children[iSchemaChild]; |
420 | 0 | ++iSchemaChild; |
421 | 0 | psChild->release = OGRLayer::ReleaseSchema; |
422 | 0 | psChild->name = CPLStrdup(poFieldDefn->GetNameRef()); |
423 | 0 | if (poFieldDefn->IsNullable()) |
424 | 0 | psChild->flags = ARROW_FLAG_NULLABLE; |
425 | 0 | const auto eType = poFieldDefn->GetType(); |
426 | 0 | const auto eSubType = poFieldDefn->GetSubType(); |
427 | 0 | const char *item_format = nullptr; |
428 | 0 | switch (eType) |
429 | 0 | { |
430 | 0 | case OFTInteger: |
431 | 0 | { |
432 | 0 | if (eSubType == OFSTBoolean) |
433 | 0 | psChild->format = "b"; |
434 | 0 | else if (eSubType == OFSTInt16) |
435 | 0 | psChild->format = "s"; |
436 | 0 | else |
437 | 0 | psChild->format = "i"; |
438 | |
|
439 | 0 | const auto &osDomainName = poFieldDefn->GetDomainName(); |
440 | 0 | if (!osDomainName.empty()) |
441 | 0 | { |
442 | 0 | auto poDS = GetDataset(); |
443 | 0 | if (poDS) |
444 | 0 | { |
445 | 0 | const auto poFieldDomain = |
446 | 0 | poDS->GetFieldDomain(osDomainName); |
447 | 0 | if (poFieldDomain && |
448 | 0 | poFieldDomain->GetDomainType() == OFDT_CODED) |
449 | 0 | { |
450 | 0 | const OGRCodedFieldDomain *poCodedDomain = |
451 | 0 | static_cast<const OGRCodedFieldDomain *>( |
452 | 0 | poFieldDomain); |
453 | 0 | AddDictToSchema(psChild, poCodedDomain); |
454 | 0 | } |
455 | 0 | } |
456 | 0 | } |
457 | |
|
458 | 0 | break; |
459 | 0 | } |
460 | | |
461 | 0 | case OFTInteger64: |
462 | 0 | psChild->format = "l"; |
463 | 0 | break; |
464 | | |
465 | 0 | case OFTReal: |
466 | 0 | { |
467 | 0 | if (eSubType == OFSTFloat32) |
468 | 0 | psChild->format = "f"; |
469 | 0 | else |
470 | 0 | psChild->format = "g"; |
471 | 0 | break; |
472 | 0 | } |
473 | | |
474 | 0 | case OFTString: |
475 | 0 | case OFTWideString: |
476 | 0 | psChild->format = "u"; |
477 | 0 | break; |
478 | | |
479 | 0 | case OFTBinary: |
480 | 0 | { |
481 | 0 | if (poFieldDefn->GetWidth() > 0) |
482 | 0 | psChild->format = |
483 | 0 | CPLStrdup(CPLSPrintf("w:%d", poFieldDefn->GetWidth())); |
484 | 0 | else |
485 | 0 | psChild->format = "z"; |
486 | 0 | break; |
487 | 0 | } |
488 | | |
489 | 0 | case OFTIntegerList: |
490 | 0 | { |
491 | 0 | if (eSubType == OFSTBoolean) |
492 | 0 | item_format = "b"; |
493 | 0 | else if (eSubType == OFSTInt16) |
494 | 0 | item_format = "s"; |
495 | 0 | else |
496 | 0 | item_format = "i"; |
497 | 0 | break; |
498 | 0 | } |
499 | | |
500 | 0 | case OFTInteger64List: |
501 | 0 | item_format = "l"; |
502 | 0 | break; |
503 | | |
504 | 0 | case OFTRealList: |
505 | 0 | { |
506 | 0 | if (eSubType == OFSTFloat32) |
507 | 0 | item_format = "f"; |
508 | 0 | else |
509 | 0 | item_format = "g"; |
510 | 0 | break; |
511 | 0 | } |
512 | | |
513 | 0 | case OFTStringList: |
514 | 0 | case OFTWideStringList: |
515 | 0 | item_format = "u"; |
516 | 0 | break; |
517 | | |
518 | 0 | case OFTDate: |
519 | 0 | psChild->format = "tdD"; |
520 | 0 | break; |
521 | | |
522 | 0 | case OFTTime: |
523 | 0 | psChild->format = "ttm"; |
524 | 0 | break; |
525 | | |
526 | 0 | case OFTDateTime: |
527 | 0 | { |
528 | 0 | const char *pszPrefix = "tsm:"; |
529 | 0 | const char *pszTZOverride = |
530 | 0 | m_aosArrowArrayStreamOptions.FetchNameValue("TIMEZONE"); |
531 | 0 | if (bDateTimeAsString) |
532 | 0 | { |
533 | 0 | psChild->format = "u"; |
534 | 0 | } |
535 | 0 | else if (pszTZOverride && EQUAL(pszTZOverride, "unknown")) |
536 | 0 | { |
537 | 0 | psChild->format = CPLStrdup(pszPrefix); |
538 | 0 | } |
539 | 0 | else if (pszTZOverride) |
540 | 0 | { |
541 | 0 | psChild->format = CPLStrdup( |
542 | 0 | (std::string(pszPrefix) + pszTZOverride).c_str()); |
543 | 0 | } |
544 | 0 | else |
545 | 0 | { |
546 | 0 | const int nTZFlag = poFieldDefn->GetTZFlag(); |
547 | 0 | if (nTZFlag == OGR_TZFLAG_MIXED_TZ || |
548 | 0 | nTZFlag == OGR_TZFLAG_UTC) |
549 | 0 | { |
550 | 0 | psChild->format = |
551 | 0 | CPLStrdup(CPLSPrintf("%sUTC", pszPrefix)); |
552 | 0 | } |
553 | 0 | else if (nTZFlag == OGR_TZFLAG_UNKNOWN || |
554 | 0 | nTZFlag == OGR_TZFLAG_LOCALTIME) |
555 | 0 | { |
556 | 0 | psChild->format = CPLStrdup(pszPrefix); |
557 | 0 | } |
558 | 0 | else |
559 | 0 | { |
560 | 0 | psChild->format = CPLStrdup( |
561 | 0 | (pszPrefix + OGRTZFlagToTimezone(nTZFlag, "UTC")) |
562 | 0 | .c_str()); |
563 | 0 | } |
564 | 0 | } |
565 | 0 | break; |
566 | 0 | } |
567 | 0 | } |
568 | | |
569 | 0 | if (item_format) |
570 | 0 | { |
571 | 0 | psChild->format = "+l"; |
572 | 0 | psChild->n_children = 1; |
573 | 0 | psChild->children = static_cast<struct ArrowSchema **>( |
574 | 0 | CPLCalloc(1, sizeof(struct ArrowSchema *))); |
575 | 0 | psChild->children[0] = static_cast<struct ArrowSchema *>( |
576 | 0 | CPLCalloc(1, sizeof(struct ArrowSchema))); |
577 | 0 | psChild->children[0]->release = OGRLayer::ReleaseSchema; |
578 | 0 | psChild->children[0]->name = CPLStrdup("item"); |
579 | 0 | psChild->children[0]->format = item_format; |
580 | 0 | } |
581 | |
|
582 | 0 | std::vector<std::pair<std::string, std::string>> oMetadata; |
583 | |
|
584 | 0 | if (eType == OFTDateTime && bDateTimeAsString) |
585 | 0 | { |
586 | 0 | oMetadata.emplace_back( |
587 | 0 | std::pair(MD_GDAL_OGR_TYPE, OGR_GetFieldTypeName(eType))); |
588 | 0 | } |
589 | |
|
590 | 0 | const char *pszAlternativeName = poFieldDefn->GetAlternativeNameRef(); |
591 | 0 | if (pszAlternativeName && pszAlternativeName[0]) |
592 | 0 | oMetadata.emplace_back( |
593 | 0 | std::pair(MD_GDAL_OGR_ALTERNATIVE_NAME, pszAlternativeName)); |
594 | |
|
595 | 0 | const char *pszDefault = poFieldDefn->GetDefault(); |
596 | 0 | if (pszDefault && pszDefault[0]) |
597 | 0 | oMetadata.emplace_back(std::pair(MD_GDAL_OGR_DEFAULT, pszDefault)); |
598 | |
|
599 | 0 | const std::string &osComment = poFieldDefn->GetComment(); |
600 | 0 | if (!osComment.empty()) |
601 | 0 | oMetadata.emplace_back(std::pair(MD_GDAL_OGR_COMMENT, osComment)); |
602 | |
|
603 | 0 | if (eType == OFTString && eSubType == OFSTJSON) |
604 | 0 | { |
605 | 0 | oMetadata.emplace_back( |
606 | 0 | std::pair(ARROW_EXTENSION_NAME_KEY, EXTENSION_NAME_ARROW_JSON)); |
607 | 0 | } |
608 | 0 | else if (eSubType != OFSTNone && eSubType != OFSTBoolean && |
609 | 0 | eSubType != OFSTFloat32) |
610 | 0 | { |
611 | 0 | oMetadata.emplace_back(std::pair( |
612 | 0 | MD_GDAL_OGR_SUBTYPE, OGR_GetFieldSubTypeName(eSubType))); |
613 | 0 | } |
614 | 0 | if (eType == OFTString && poFieldDefn->GetWidth() > 0) |
615 | 0 | { |
616 | 0 | oMetadata.emplace_back(std::pair( |
617 | 0 | MD_GDAL_OGR_WIDTH, CPLSPrintf("%d", poFieldDefn->GetWidth()))); |
618 | 0 | } |
619 | 0 | if (poFieldDefn->IsUnique()) |
620 | 0 | { |
621 | 0 | oMetadata.emplace_back(std::pair(MD_GDAL_OGR_UNIQUE, "true")); |
622 | 0 | } |
623 | 0 | if (!poFieldDefn->GetDomainName().empty()) |
624 | 0 | { |
625 | 0 | oMetadata.emplace_back(std::pair(MD_GDAL_OGR_DOMAIN_NAME, |
626 | 0 | poFieldDefn->GetDomainName())); |
627 | 0 | } |
628 | |
|
629 | 0 | if (!oMetadata.empty()) |
630 | 0 | { |
631 | 0 | uint64_t nLen64 = sizeof(int32_t); |
632 | 0 | for (const auto &oPair : oMetadata) |
633 | 0 | { |
634 | 0 | nLen64 += sizeof(int32_t); |
635 | 0 | nLen64 += oPair.first.size(); |
636 | 0 | nLen64 += sizeof(int32_t); |
637 | 0 | nLen64 += oPair.second.size(); |
638 | 0 | } |
639 | 0 | if (nLen64 < |
640 | 0 | static_cast<uint64_t>(std::numeric_limits<int32_t>::max())) |
641 | 0 | { |
642 | 0 | const size_t nLen = static_cast<size_t>(nLen64); |
643 | 0 | char *pszMetadata = static_cast<char *>(CPLMalloc(nLen)); |
644 | 0 | psChild->metadata = pszMetadata; |
645 | 0 | size_t offsetMD = 0; |
646 | 0 | int32_t nSize = static_cast<int>(oMetadata.size()); |
647 | 0 | memcpy(pszMetadata + offsetMD, &nSize, sizeof(nSize)); |
648 | 0 | offsetMD += sizeof(int32_t); |
649 | 0 | for (const auto &oPair : oMetadata) |
650 | 0 | { |
651 | 0 | nSize = static_cast<int32_t>(oPair.first.size()); |
652 | 0 | memcpy(pszMetadata + offsetMD, &nSize, sizeof(nSize)); |
653 | 0 | offsetMD += sizeof(int32_t); |
654 | 0 | memcpy(pszMetadata + offsetMD, oPair.first.data(), |
655 | 0 | oPair.first.size()); |
656 | 0 | offsetMD += oPair.first.size(); |
657 | |
|
658 | 0 | nSize = static_cast<int32_t>(oPair.second.size()); |
659 | 0 | memcpy(pszMetadata + offsetMD, &nSize, sizeof(nSize)); |
660 | 0 | offsetMD += sizeof(int32_t); |
661 | 0 | memcpy(pszMetadata + offsetMD, oPair.second.data(), |
662 | 0 | oPair.second.size()); |
663 | 0 | offsetMD += oPair.second.size(); |
664 | 0 | } |
665 | |
|
666 | 0 | CPLAssert(offsetMD == nLen); |
667 | 0 | CPL_IGNORE_RET_VAL(offsetMD); |
668 | 0 | } |
669 | 0 | else |
670 | 0 | { |
671 | | // Extremely unlikely ! |
672 | 0 | CPLError(CE_Warning, CPLE_AppDefined, |
673 | 0 | "Cannot write ArrowSchema::metadata due to " |
674 | 0 | "too large content"); |
675 | 0 | } |
676 | 0 | } |
677 | 0 | } |
678 | | |
679 | 0 | const char *const pszGeometryMetadataEncoding = |
680 | 0 | m_aosArrowArrayStreamOptions.FetchNameValue( |
681 | 0 | "GEOMETRY_METADATA_ENCODING"); |
682 | 0 | const char *pszExtensionName = EXTENSION_NAME_OGC_WKB; |
683 | 0 | if (pszGeometryMetadataEncoding) |
684 | 0 | { |
685 | 0 | if (EQUAL(pszGeometryMetadataEncoding, "OGC")) |
686 | 0 | pszExtensionName = EXTENSION_NAME_OGC_WKB; |
687 | 0 | else if (EQUAL(pszGeometryMetadataEncoding, "GEOARROW")) |
688 | 0 | pszExtensionName = EXTENSION_NAME_GEOARROW_WKB; |
689 | 0 | else |
690 | 0 | CPLError(CE_Warning, CPLE_NotSupported, |
691 | 0 | "Unsupported GEOMETRY_METADATA_ENCODING value: %s", |
692 | 0 | pszGeometryMetadataEncoding); |
693 | 0 | } |
694 | 0 | for (int i = 0; i < nGeomFieldCount; ++i) |
695 | 0 | { |
696 | 0 | const auto poFieldDefn = poLayerDefn->GetGeomFieldDefn(i); |
697 | 0 | if (poFieldDefn->IsIgnored()) |
698 | 0 | { |
699 | 0 | continue; |
700 | 0 | } |
701 | | |
702 | 0 | out_schema->children[iSchemaChild] = CreateSchemaForWKBGeometryColumn( |
703 | 0 | poFieldDefn, "z", pszExtensionName); |
704 | |
|
705 | 0 | ++iSchemaChild; |
706 | 0 | } |
707 | |
|
708 | 0 | out_schema->n_children = iSchemaChild; |
709 | 0 | out_schema->release = OGRLayer::ReleaseSchema; |
710 | 0 | return 0; |
711 | 0 | } |
712 | | |
713 | | /************************************************************************/ |
714 | | /* CreateSchemaForWKBGeometryColumn() */ |
715 | | /************************************************************************/ |
716 | | |
717 | | /** Return a ArrowSchema* corresponding to the WKB encoding of a geometry |
718 | | * column. |
719 | | */ |
720 | | |
721 | | /* static */ |
722 | | struct ArrowSchema * |
723 | | OGRLayer::CreateSchemaForWKBGeometryColumn(const OGRGeomFieldDefn *poFieldDefn, |
724 | | const char *pszArrowFormat, |
725 | | const char *pszExtensionName) |
726 | 0 | { |
727 | 0 | CPLAssert(strcmp(pszArrowFormat, "z") == 0 || |
728 | 0 | strcmp(pszArrowFormat, "Z") == 0); |
729 | 0 | if (!EQUAL(pszExtensionName, EXTENSION_NAME_OGC_WKB) && |
730 | 0 | !EQUAL(pszExtensionName, EXTENSION_NAME_GEOARROW_WKB)) |
731 | 0 | { |
732 | 0 | CPLError(CE_Failure, CPLE_NotSupported, |
733 | 0 | "Unsupported extension name '%s'. Defaulting to '%s'", |
734 | 0 | pszExtensionName, EXTENSION_NAME_OGC_WKB); |
735 | 0 | pszExtensionName = EXTENSION_NAME_OGC_WKB; |
736 | 0 | } |
737 | 0 | auto psSchema = static_cast<struct ArrowSchema *>( |
738 | 0 | CPLCalloc(1, sizeof(struct ArrowSchema))); |
739 | 0 | psSchema->release = OGRLayer::ReleaseSchema; |
740 | 0 | const char *pszGeomFieldName = poFieldDefn->GetNameRef(); |
741 | 0 | if (pszGeomFieldName[0] == '\0') |
742 | 0 | pszGeomFieldName = DEFAULT_ARROW_GEOMETRY_NAME; |
743 | 0 | psSchema->name = CPLStrdup(pszGeomFieldName); |
744 | 0 | if (poFieldDefn->IsNullable()) |
745 | 0 | psSchema->flags = ARROW_FLAG_NULLABLE; |
746 | 0 | psSchema->format = strcmp(pszArrowFormat, "z") == 0 ? "z" : "Z"; |
747 | 0 | std::string osExtensionMetadata; |
748 | 0 | if (EQUAL(pszExtensionName, EXTENSION_NAME_GEOARROW_WKB)) |
749 | 0 | { |
750 | 0 | const auto poSRS = poFieldDefn->GetSpatialRef(); |
751 | 0 | if (poSRS) |
752 | 0 | { |
753 | 0 | char *pszPROJJSON = nullptr; |
754 | 0 | poSRS->exportToPROJJSON(&pszPROJJSON, nullptr); |
755 | 0 | if (pszPROJJSON) |
756 | 0 | { |
757 | 0 | osExtensionMetadata = "{\"crs\":"; |
758 | 0 | osExtensionMetadata += pszPROJJSON; |
759 | 0 | osExtensionMetadata += '}'; |
760 | 0 | CPLFree(pszPROJJSON); |
761 | 0 | } |
762 | 0 | else |
763 | 0 | { |
764 | 0 | CPLError(CE_Warning, CPLE_AppDefined, |
765 | 0 | "Cannot export CRS of geometry field %s to PROJJSON", |
766 | 0 | poFieldDefn->GetNameRef()); |
767 | 0 | } |
768 | 0 | } |
769 | 0 | } |
770 | 0 | size_t nLen = sizeof(int32_t) + sizeof(int32_t) + |
771 | 0 | strlen(ARROW_EXTENSION_NAME_KEY) + sizeof(int32_t) + |
772 | 0 | strlen(pszExtensionName); |
773 | 0 | if (!osExtensionMetadata.empty()) |
774 | 0 | { |
775 | 0 | nLen += sizeof(int32_t) + strlen(ARROW_EXTENSION_METADATA_KEY) + |
776 | 0 | sizeof(int32_t) + osExtensionMetadata.size(); |
777 | 0 | } |
778 | 0 | char *pszMetadata = static_cast<char *>(CPLMalloc(nLen)); |
779 | 0 | psSchema->metadata = pszMetadata; |
780 | 0 | size_t offsetMD = 0; |
781 | 0 | *reinterpret_cast<int32_t *>(pszMetadata + offsetMD) = |
782 | 0 | osExtensionMetadata.empty() ? 1 : 2; |
783 | 0 | offsetMD += sizeof(int32_t); |
784 | 0 | *reinterpret_cast<int32_t *>(pszMetadata + offsetMD) = |
785 | 0 | static_cast<int32_t>(strlen(ARROW_EXTENSION_NAME_KEY)); |
786 | 0 | offsetMD += sizeof(int32_t); |
787 | 0 | memcpy(pszMetadata + offsetMD, ARROW_EXTENSION_NAME_KEY, |
788 | 0 | strlen(ARROW_EXTENSION_NAME_KEY)); |
789 | 0 | offsetMD += static_cast<int>(strlen(ARROW_EXTENSION_NAME_KEY)); |
790 | 0 | *reinterpret_cast<int32_t *>(pszMetadata + offsetMD) = |
791 | 0 | static_cast<int32_t>(strlen(pszExtensionName)); |
792 | 0 | offsetMD += sizeof(int32_t); |
793 | 0 | memcpy(pszMetadata + offsetMD, pszExtensionName, strlen(pszExtensionName)); |
794 | 0 | offsetMD += strlen(pszExtensionName); |
795 | 0 | if (!osExtensionMetadata.empty()) |
796 | 0 | { |
797 | 0 | *reinterpret_cast<int32_t *>(pszMetadata + offsetMD) = |
798 | 0 | static_cast<int32_t>(strlen(ARROW_EXTENSION_METADATA_KEY)); |
799 | 0 | offsetMD += sizeof(int32_t); |
800 | 0 | memcpy(pszMetadata + offsetMD, ARROW_EXTENSION_METADATA_KEY, |
801 | 0 | strlen(ARROW_EXTENSION_METADATA_KEY)); |
802 | 0 | offsetMD += static_cast<int>(strlen(ARROW_EXTENSION_METADATA_KEY)); |
803 | 0 | *reinterpret_cast<int32_t *>(pszMetadata + offsetMD) = |
804 | 0 | static_cast<int32_t>(osExtensionMetadata.size()); |
805 | 0 | offsetMD += sizeof(int32_t); |
806 | 0 | memcpy(pszMetadata + offsetMD, osExtensionMetadata.c_str(), |
807 | 0 | osExtensionMetadata.size()); |
808 | 0 | offsetMD += osExtensionMetadata.size(); |
809 | 0 | } |
810 | 0 | CPLAssert(offsetMD == nLen); |
811 | 0 | CPL_IGNORE_RET_VAL(offsetMD); |
812 | 0 | return psSchema; |
813 | 0 | } |
814 | | |
815 | | /************************************************************************/ |
816 | | /* StaticGetArrowSchema() */ |
817 | | /************************************************************************/ |
818 | | |
819 | | /** Default implementation of the ArrowArrayStream::get_schema() callback. |
820 | | * |
821 | | * To be used by driver implementations that have a custom GetArrowStream() |
822 | | * implementation. |
823 | | * |
824 | | * @since GDAL 3.6 |
825 | | */ |
826 | | int OGRLayer::StaticGetArrowSchema(struct ArrowArrayStream *stream, |
827 | | struct ArrowSchema *out_schema) |
828 | 0 | { |
829 | 0 | auto poLayer = static_cast<ArrowArrayStreamPrivateDataSharedDataWrapper *>( |
830 | 0 | stream->private_data) |
831 | 0 | ->poShared->m_poLayer; |
832 | 0 | if (poLayer == nullptr) |
833 | 0 | { |
834 | 0 | CPLError(CE_Failure, CPLE_NotSupported, |
835 | 0 | "Calling get_schema() on a freed OGRLayer is not supported"); |
836 | 0 | return EINVAL; |
837 | 0 | } |
838 | 0 | return poLayer->GetArrowSchema(stream, out_schema); |
839 | 0 | } |
840 | | |
841 | | /************************************************************************/ |
842 | | /* DefaultReleaseArray() */ |
843 | | /************************************************************************/ |
844 | | |
845 | | static void OGRLayerDefaultReleaseArray(struct ArrowArray *array) |
846 | 0 | { |
847 | 0 | if (array->buffers) |
848 | 0 | { |
849 | 0 | for (int i = 0; i < static_cast<int>(array->n_buffers); ++i) |
850 | 0 | VSIFreeAligned(const_cast<void *>(array->buffers[i])); |
851 | 0 | CPLFree(array->buffers); |
852 | 0 | } |
853 | 0 | if (array->children) |
854 | 0 | { |
855 | 0 | for (int i = 0; i < static_cast<int>(array->n_children); ++i) |
856 | 0 | { |
857 | 0 | if (array->children[i] && array->children[i]->release) |
858 | 0 | { |
859 | 0 | array->children[i]->release(array->children[i]); |
860 | 0 | CPLFree(array->children[i]); |
861 | 0 | } |
862 | 0 | } |
863 | 0 | CPLFree(array->children); |
864 | 0 | } |
865 | 0 | if (array->dictionary) |
866 | 0 | { |
867 | 0 | if (array->dictionary->release) |
868 | 0 | { |
869 | 0 | array->dictionary->release(array->dictionary); |
870 | 0 | CPLFree(array->dictionary); |
871 | 0 | } |
872 | 0 | } |
873 | 0 | array->release = nullptr; |
874 | 0 | } |
875 | | |
876 | | /** Release a ArrowArray. |
877 | | * |
878 | | * To be used by driver implementations that have a custom GetArrowStream() |
879 | | * implementation. |
880 | | * |
881 | | * @param array Arrow array to release. |
882 | | * @since GDAL 3.6 |
883 | | */ |
884 | | void OGRLayer::ReleaseArray(struct ArrowArray *array) |
885 | 0 | { |
886 | 0 | OGRLayerDefaultReleaseArray(array); |
887 | 0 | } |
888 | | |
889 | | /************************************************************************/ |
890 | | /* IsValidField() */ |
891 | | /************************************************************************/ |
892 | | |
893 | | static inline bool IsValidField(const OGRField *psRawField) |
894 | 0 | { |
895 | 0 | return (!(psRawField->Set.nMarker1 == OGRUnsetMarker && |
896 | 0 | psRawField->Set.nMarker2 == OGRUnsetMarker && |
897 | 0 | psRawField->Set.nMarker3 == OGRUnsetMarker) && |
898 | 0 | !(psRawField->Set.nMarker1 == OGRNullMarker && |
899 | 0 | psRawField->Set.nMarker2 == OGRNullMarker && |
900 | 0 | psRawField->Set.nMarker3 == OGRNullMarker)); |
901 | 0 | } |
902 | | |
903 | | /************************************************************************/ |
904 | | /* AllocValidityBitmap() */ |
905 | | /************************************************************************/ |
906 | | |
907 | | static uint8_t *AllocValidityBitmap(size_t nSize) |
908 | 0 | { |
909 | 0 | auto pabyValidity = static_cast<uint8_t *>( |
910 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE((1 + nSize + 7) / 8)); |
911 | 0 | if (pabyValidity) |
912 | 0 | { |
913 | | // All valid initially |
914 | 0 | memset(pabyValidity, 0xFF, (nSize + 7) / 8); |
915 | 0 | } |
916 | 0 | return pabyValidity; |
917 | 0 | } |
918 | | |
919 | | /************************************************************************/ |
920 | | /* FillArray() */ |
921 | | /************************************************************************/ |
922 | | |
923 | | template <class T, typename TMember> |
924 | | static bool FillArray(struct ArrowArray *psChild, |
925 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
926 | | const size_t nFeatureCountLimit, const bool bIsNullable, |
927 | | TMember member, const int i) |
928 | 0 | { |
929 | 0 | psChild->n_buffers = 2; |
930 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
931 | 0 | uint8_t *pabyValidity = nullptr; |
932 | 0 | T *panValues = static_cast<T *>( |
933 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(sizeof(T) * (1 + nFeatureCountLimit))); |
934 | 0 | if (panValues == nullptr) |
935 | 0 | return false; |
936 | 0 | psChild->buffers[1] = panValues; |
937 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat) |
938 | 0 | { |
939 | 0 | auto &poFeature = apoFeatures[iFeat]; |
940 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
941 | 0 | if (IsValidField(psRawField)) |
942 | 0 | { |
943 | 0 | panValues[iFeat] = static_cast<T>((*psRawField).*member); |
944 | 0 | } |
945 | 0 | else if (bIsNullable) |
946 | 0 | { |
947 | 0 | panValues[iFeat] = 0; |
948 | 0 | ++psChild->null_count; |
949 | 0 | if (pabyValidity == nullptr) |
950 | 0 | { |
951 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
952 | 0 | psChild->buffers[0] = pabyValidity; |
953 | 0 | if (pabyValidity == nullptr) |
954 | 0 | return false; |
955 | 0 | } |
956 | 0 | UnsetBit(pabyValidity, iFeat); |
957 | 0 | } |
958 | 0 | else |
959 | 0 | { |
960 | 0 | panValues[iFeat] = 0; |
961 | 0 | } |
962 | 0 | } |
963 | 0 | return true; |
964 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:bool FillArray<short, int OGRField::*>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, int OGRField::*, int) Unexecuted instantiation: ogrlayerarrow.cpp:bool FillArray<int, int OGRField::*>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, int OGRField::*, int) Unexecuted instantiation: ogrlayerarrow.cpp:bool FillArray<long, long long OGRField::*>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, long long OGRField::*, int) Unexecuted instantiation: ogrlayerarrow.cpp:bool FillArray<float, double OGRField::*>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, double OGRField::*, int) Unexecuted instantiation: ogrlayerarrow.cpp:bool FillArray<double, double OGRField::*>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, double OGRField::*, int) |
965 | | |
966 | | /************************************************************************/ |
967 | | /* FillBoolArray() */ |
968 | | /************************************************************************/ |
969 | | |
970 | | template <typename TMember> |
971 | | static bool FillBoolArray(struct ArrowArray *psChild, |
972 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
973 | | const size_t nFeatureCountLimit, |
974 | | const bool bIsNullable, TMember member, const int i) |
975 | 0 | { |
976 | 0 | psChild->n_buffers = 2; |
977 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
978 | 0 | uint8_t *pabyValidity = nullptr; |
979 | 0 | uint8_t *panValues = static_cast<uint8_t *>( |
980 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE((nFeatureCountLimit + 7 + 1) / 8)); |
981 | 0 | if (panValues == nullptr) |
982 | 0 | return false; |
983 | 0 | memset(panValues, 0, (nFeatureCountLimit + 7) / 8); |
984 | 0 | psChild->buffers[1] = panValues; |
985 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat) |
986 | 0 | { |
987 | 0 | auto &poFeature = apoFeatures[iFeat]; |
988 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
989 | 0 | if (IsValidField(psRawField)) |
990 | 0 | { |
991 | 0 | if ((*psRawField).*member) |
992 | 0 | SetBit(panValues, iFeat); |
993 | 0 | } |
994 | 0 | else if (bIsNullable) |
995 | 0 | { |
996 | 0 | ++psChild->null_count; |
997 | 0 | if (pabyValidity == nullptr) |
998 | 0 | { |
999 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1000 | 0 | psChild->buffers[0] = pabyValidity; |
1001 | 0 | if (pabyValidity == nullptr) |
1002 | 0 | return false; |
1003 | 0 | } |
1004 | 0 | UnsetBit(pabyValidity, iFeat); |
1005 | 0 | } |
1006 | 0 | } |
1007 | 0 | return true; |
1008 | 0 | } |
1009 | | |
1010 | | /************************************************************************/ |
1011 | | /* FillListArray() */ |
1012 | | /************************************************************************/ |
1013 | | |
1014 | | struct GetFromIntegerList |
1015 | | { |
1016 | | static inline int getCount(const OGRField *psRawField) |
1017 | 0 | { |
1018 | 0 | return psRawField->IntegerList.nCount; |
1019 | 0 | } |
1020 | | |
1021 | | static inline const int *getValues(const OGRField *psRawField) |
1022 | 0 | { |
1023 | 0 | return psRawField->IntegerList.paList; |
1024 | 0 | } |
1025 | | }; |
1026 | | |
1027 | | struct GetFromInteger64List |
1028 | | { |
1029 | | static inline int getCount(const OGRField *psRawField) |
1030 | 0 | { |
1031 | 0 | return psRawField->Integer64List.nCount; |
1032 | 0 | } |
1033 | | |
1034 | | static inline const GIntBig *getValues(const OGRField *psRawField) |
1035 | 0 | { |
1036 | 0 | return psRawField->Integer64List.paList; |
1037 | 0 | } |
1038 | | }; |
1039 | | |
1040 | | struct GetFromRealList |
1041 | | { |
1042 | | static inline int getCount(const OGRField *psRawField) |
1043 | 0 | { |
1044 | 0 | return psRawField->RealList.nCount; |
1045 | 0 | } |
1046 | | |
1047 | | static inline const double *getValues(const OGRField *psRawField) |
1048 | 0 | { |
1049 | 0 | return psRawField->RealList.paList; |
1050 | 0 | } |
1051 | | }; |
1052 | | |
1053 | | template <class OffsetType, class T, class GetFromList> |
1054 | | static size_t |
1055 | | FillListArray(struct ArrowArray *psChild, |
1056 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1057 | | const size_t nFeatureCountLimit, const bool bIsNullable, |
1058 | | const int i, const size_t nMemLimit) |
1059 | 0 | { |
1060 | 0 | psChild->n_buffers = 2; |
1061 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
1062 | 0 | uint8_t *pabyValidity = nullptr; |
1063 | 0 | OffsetType *panOffsets = |
1064 | 0 | static_cast<OffsetType *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE( |
1065 | 0 | sizeof(OffsetType) * (1 + nFeatureCountLimit))); |
1066 | 0 | if (panOffsets == nullptr) |
1067 | 0 | return 0; |
1068 | 0 | psChild->buffers[1] = panOffsets; |
1069 | |
|
1070 | 0 | OffsetType nOffset = 0; |
1071 | 0 | size_t nFeatCount = 0; |
1072 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat, ++nFeatCount) |
1073 | 0 | { |
1074 | 0 | panOffsets[iFeat] = nOffset; |
1075 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1076 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1077 | 0 | if (IsValidField(psRawField)) |
1078 | 0 | { |
1079 | 0 | const unsigned nCount = GetFromList::getCount(psRawField); |
1080 | 0 | if (nCount > static_cast<size_t>(nMemLimit - nOffset)) |
1081 | 0 | { |
1082 | 0 | if (nFeatCount == 0) |
1083 | 0 | return 0; |
1084 | 0 | break; |
1085 | 0 | } |
1086 | 0 | nOffset += static_cast<OffsetType>(nCount); |
1087 | 0 | } |
1088 | 0 | else if (bIsNullable) |
1089 | 0 | { |
1090 | 0 | ++psChild->null_count; |
1091 | 0 | if (pabyValidity == nullptr) |
1092 | 0 | { |
1093 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1094 | 0 | psChild->buffers[0] = pabyValidity; |
1095 | 0 | if (pabyValidity == nullptr) |
1096 | 0 | return 0; |
1097 | 0 | } |
1098 | 0 | UnsetBit(pabyValidity, iFeat); |
1099 | 0 | } |
1100 | 0 | } |
1101 | 0 | panOffsets[nFeatCount] = nOffset; |
1102 | |
|
1103 | 0 | psChild->n_children = 1; |
1104 | 0 | psChild->children = static_cast<struct ArrowArray **>( |
1105 | 0 | CPLCalloc(1, sizeof(struct ArrowArray *))); |
1106 | 0 | psChild->children[0] = static_cast<struct ArrowArray *>( |
1107 | 0 | CPLCalloc(1, sizeof(struct ArrowArray))); |
1108 | 0 | auto psValueChild = psChild->children[0]; |
1109 | |
|
1110 | 0 | psValueChild->release = OGRLayerDefaultReleaseArray; |
1111 | 0 | psValueChild->n_buffers = 2; |
1112 | 0 | psValueChild->buffers = |
1113 | 0 | static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
1114 | 0 | psValueChild->length = nOffset; |
1115 | 0 | T *panValues = static_cast<T *>( |
1116 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(sizeof(T) * (nOffset + 1))); |
1117 | 0 | if (panValues == nullptr) |
1118 | 0 | return 0; |
1119 | 0 | psValueChild->buffers[1] = panValues; |
1120 | |
|
1121 | 0 | nOffset = 0; |
1122 | 0 | for (size_t iFeat = 0; iFeat < nFeatCount; ++iFeat) |
1123 | 0 | { |
1124 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1125 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1126 | 0 | if (IsValidField(psRawField)) |
1127 | 0 | { |
1128 | 0 | const int nCount = GetFromList::getCount(psRawField); |
1129 | 0 | const auto paList = GetFromList::getValues(psRawField); |
1130 | 0 | if (sizeof(*paList) == sizeof(T)) |
1131 | 0 | memcpy(panValues + nOffset, paList, nCount * sizeof(T)); |
1132 | 0 | else |
1133 | 0 | { |
1134 | 0 | for (int j = 0; j < nCount; ++j) |
1135 | 0 | { |
1136 | 0 | panValues[nOffset + j] = static_cast<T>(paList[j]); |
1137 | 0 | } |
1138 | 0 | } |
1139 | 0 | nOffset += static_cast<OffsetType>(nCount); |
1140 | 0 | } |
1141 | 0 | } |
1142 | |
|
1143 | 0 | return nFeatCount; |
1144 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:unsigned long FillListArray<int, short, GetFromIntegerList>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, int, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:unsigned long FillListArray<int, int, GetFromIntegerList>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, int, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:unsigned long FillListArray<int, long, GetFromInteger64List>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, int, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:unsigned long FillListArray<int, float, GetFromRealList>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, int, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:unsigned long FillListArray<int, double, GetFromRealList>(ArrowArray*, std::__1::deque<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> >, std::__1::allocator<std::__1::unique_ptr<OGRFeature, std::__1::default_delete<OGRFeature> > > >&, unsigned long, bool, int, unsigned long) |
1145 | | |
1146 | | template <class OffsetType, class GetFromList> |
1147 | | static size_t |
1148 | | FillListArrayBool(struct ArrowArray *psChild, |
1149 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1150 | | const size_t nFeatureCountLimit, const bool bIsNullable, |
1151 | | const int i, const size_t nMemLimit) |
1152 | 0 | { |
1153 | 0 | psChild->n_buffers = 2; |
1154 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
1155 | 0 | uint8_t *pabyValidity = nullptr; |
1156 | 0 | OffsetType *panOffsets = |
1157 | 0 | static_cast<OffsetType *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE( |
1158 | 0 | sizeof(OffsetType) * (1 + nFeatureCountLimit))); |
1159 | 0 | if (panOffsets == nullptr) |
1160 | 0 | return 0; |
1161 | 0 | psChild->buffers[1] = panOffsets; |
1162 | |
|
1163 | 0 | OffsetType nOffset = 0; |
1164 | 0 | size_t nFeatCount = 0; |
1165 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat, ++nFeatCount) |
1166 | 0 | { |
1167 | 0 | panOffsets[iFeat] = nOffset; |
1168 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1169 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1170 | 0 | if (IsValidField(psRawField)) |
1171 | 0 | { |
1172 | 0 | const unsigned nCount = GetFromList::getCount(psRawField); |
1173 | 0 | if (nCount > static_cast<size_t>(nMemLimit - nOffset)) |
1174 | 0 | { |
1175 | 0 | if (nFeatCount == 0) |
1176 | 0 | return 0; |
1177 | 0 | break; |
1178 | 0 | } |
1179 | 0 | nOffset += static_cast<OffsetType>(nCount); |
1180 | 0 | } |
1181 | 0 | else if (bIsNullable) |
1182 | 0 | { |
1183 | 0 | ++psChild->null_count; |
1184 | 0 | if (pabyValidity == nullptr) |
1185 | 0 | { |
1186 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1187 | 0 | psChild->buffers[0] = pabyValidity; |
1188 | 0 | if (pabyValidity == nullptr) |
1189 | 0 | return 0; |
1190 | 0 | } |
1191 | 0 | UnsetBit(pabyValidity, iFeat); |
1192 | 0 | } |
1193 | 0 | } |
1194 | 0 | panOffsets[nFeatCount] = nOffset; |
1195 | |
|
1196 | 0 | psChild->n_children = 1; |
1197 | 0 | psChild->children = static_cast<struct ArrowArray **>( |
1198 | 0 | CPLCalloc(1, sizeof(struct ArrowArray *))); |
1199 | 0 | psChild->children[0] = static_cast<struct ArrowArray *>( |
1200 | 0 | CPLCalloc(1, sizeof(struct ArrowArray))); |
1201 | 0 | auto psValueChild = psChild->children[0]; |
1202 | |
|
1203 | 0 | psValueChild->release = OGRLayerDefaultReleaseArray; |
1204 | 0 | psValueChild->n_buffers = 2; |
1205 | 0 | psValueChild->buffers = |
1206 | 0 | static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
1207 | 0 | psValueChild->length = nOffset; |
1208 | 0 | uint8_t *panValues = static_cast<uint8_t *>( |
1209 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE((nOffset + 7 + 1) / 8)); |
1210 | 0 | if (panValues == nullptr) |
1211 | 0 | return 0; |
1212 | 0 | memset(panValues, 0, (nOffset + 7) / 8); |
1213 | 0 | psValueChild->buffers[1] = panValues; |
1214 | |
|
1215 | 0 | nOffset = 0; |
1216 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat) |
1217 | 0 | { |
1218 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1219 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1220 | 0 | if (IsValidField(psRawField)) |
1221 | 0 | { |
1222 | 0 | const int nCount = GetFromList::getCount(psRawField); |
1223 | 0 | const auto paList = GetFromList::getValues(psRawField); |
1224 | |
|
1225 | 0 | for (int j = 0; j < nCount; ++j) |
1226 | 0 | { |
1227 | 0 | if (paList[j]) |
1228 | 0 | SetBit(panValues, nOffset + j); |
1229 | 0 | } |
1230 | 0 | nOffset += static_cast<OffsetType>(nCount); |
1231 | 0 | } |
1232 | 0 | } |
1233 | |
|
1234 | 0 | return nFeatCount; |
1235 | 0 | } |
1236 | | |
1237 | | /************************************************************************/ |
1238 | | /* FillStringArray() */ |
1239 | | /************************************************************************/ |
1240 | | |
1241 | | template <class T> |
1242 | | static size_t |
1243 | | FillStringArray(struct ArrowArray *psChild, |
1244 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1245 | | const size_t nFeatureCountLimit, const bool bIsNullable, |
1246 | | const int i, const size_t nMemLimit) |
1247 | 0 | { |
1248 | 0 | psChild->n_buffers = 3; |
1249 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(3, sizeof(void *))); |
1250 | 0 | uint8_t *pabyValidity = nullptr; |
1251 | 0 | T *panOffsets = static_cast<T *>( |
1252 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(sizeof(T) * (1 + nFeatureCountLimit))); |
1253 | 0 | if (panOffsets == nullptr) |
1254 | 0 | return 0; |
1255 | 0 | psChild->buffers[1] = panOffsets; |
1256 | |
|
1257 | 0 | size_t nOffset = 0; |
1258 | 0 | size_t nFeatCount = 0; |
1259 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat, ++nFeatCount) |
1260 | 0 | { |
1261 | 0 | panOffsets[iFeat] = static_cast<T>(nOffset); |
1262 | 0 | const auto psRawField = apoFeatures[iFeat]->GetRawFieldRef(i); |
1263 | 0 | if (IsValidField(psRawField)) |
1264 | 0 | { |
1265 | 0 | const size_t nLen = strlen(psRawField->String); |
1266 | 0 | if (nLen > nMemLimit - nOffset) |
1267 | 0 | { |
1268 | 0 | if (nFeatCount == 0) |
1269 | 0 | return 0; |
1270 | 0 | break; |
1271 | 0 | } |
1272 | 0 | nOffset += static_cast<T>(nLen); |
1273 | 0 | } |
1274 | 0 | else if (bIsNullable) |
1275 | 0 | { |
1276 | 0 | ++psChild->null_count; |
1277 | 0 | if (pabyValidity == nullptr) |
1278 | 0 | { |
1279 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1280 | 0 | psChild->buffers[0] = pabyValidity; |
1281 | 0 | if (pabyValidity == nullptr) |
1282 | 0 | return 0; |
1283 | 0 | } |
1284 | 0 | UnsetBit(pabyValidity, iFeat); |
1285 | 0 | } |
1286 | 0 | } |
1287 | 0 | panOffsets[nFeatCount] = static_cast<T>(nOffset); |
1288 | |
|
1289 | 0 | char *pachValues = |
1290 | 0 | static_cast<char *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE(nOffset + 1)); |
1291 | 0 | if (pachValues == nullptr) |
1292 | 0 | return 0; |
1293 | 0 | psChild->buffers[2] = pachValues; |
1294 | |
|
1295 | 0 | nOffset = 0; |
1296 | 0 | for (size_t iFeat = 0; iFeat < nFeatCount; ++iFeat) |
1297 | 0 | { |
1298 | 0 | const size_t nLen = |
1299 | 0 | static_cast<size_t>(panOffsets[iFeat + 1] - panOffsets[iFeat]); |
1300 | 0 | if (nLen) |
1301 | 0 | { |
1302 | 0 | const auto psRawField = apoFeatures[iFeat]->GetRawFieldRef(i); |
1303 | 0 | memcpy(pachValues + nOffset, psRawField->String, nLen); |
1304 | 0 | nOffset += nLen; |
1305 | 0 | } |
1306 | 0 | } |
1307 | |
|
1308 | 0 | return nFeatCount; |
1309 | 0 | } |
1310 | | |
1311 | | /************************************************************************/ |
1312 | | /* FillStringListArray() */ |
1313 | | /************************************************************************/ |
1314 | | |
1315 | | template <class OffsetType> |
1316 | | static size_t |
1317 | | FillStringListArray(struct ArrowArray *psChild, |
1318 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1319 | | const size_t nFeatureCountLimit, const bool bIsNullable, |
1320 | | const int i, const size_t nMemLimit) |
1321 | 0 | { |
1322 | 0 | psChild->n_buffers = 2; |
1323 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
1324 | 0 | uint8_t *pabyValidity = nullptr; |
1325 | 0 | OffsetType *panOffsets = |
1326 | 0 | static_cast<OffsetType *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE( |
1327 | 0 | sizeof(OffsetType) * (1 + nFeatureCountLimit))); |
1328 | 0 | if (panOffsets == nullptr) |
1329 | 0 | return false; |
1330 | 0 | psChild->buffers[1] = panOffsets; |
1331 | |
|
1332 | 0 | OffsetType nStrings = 0; |
1333 | 0 | OffsetType nCountChars = 0; |
1334 | 0 | size_t nFeatCount = 0; |
1335 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat, ++nFeatCount) |
1336 | 0 | { |
1337 | 0 | panOffsets[iFeat] = nStrings; |
1338 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1339 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1340 | 0 | if (IsValidField(psRawField)) |
1341 | 0 | { |
1342 | 0 | const int nCount = psRawField->StringList.nCount; |
1343 | 0 | if (static_cast<size_t>(nCount) > |
1344 | 0 | static_cast<size_t>(nMemLimit - nStrings)) |
1345 | 0 | { |
1346 | 0 | if (nFeatCount == 0) |
1347 | 0 | return 0; |
1348 | 0 | goto after_loop; |
1349 | 0 | } |
1350 | 0 | for (int j = 0; j < nCount; ++j) |
1351 | 0 | { |
1352 | 0 | const size_t nLen = strlen(psRawField->StringList.paList[j]); |
1353 | 0 | if (nLen > static_cast<size_t>(nMemLimit - nCountChars)) |
1354 | 0 | { |
1355 | 0 | if (nFeatCount == 0) |
1356 | 0 | return 0; |
1357 | 0 | goto after_loop; |
1358 | 0 | } |
1359 | 0 | nCountChars += static_cast<OffsetType>(nLen); |
1360 | 0 | } |
1361 | 0 | nStrings += static_cast<OffsetType>(nCount); |
1362 | 0 | } |
1363 | 0 | else if (bIsNullable) |
1364 | 0 | { |
1365 | 0 | ++psChild->null_count; |
1366 | 0 | if (pabyValidity == nullptr) |
1367 | 0 | { |
1368 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1369 | 0 | psChild->buffers[0] = pabyValidity; |
1370 | 0 | if (pabyValidity == nullptr) |
1371 | 0 | return 0; |
1372 | 0 | } |
1373 | 0 | UnsetBit(pabyValidity, iFeat); |
1374 | 0 | } |
1375 | 0 | } |
1376 | 0 | after_loop: |
1377 | 0 | panOffsets[nFeatCount] = nStrings; |
1378 | |
|
1379 | 0 | psChild->n_children = 1; |
1380 | 0 | psChild->children = static_cast<struct ArrowArray **>( |
1381 | 0 | CPLCalloc(1, sizeof(struct ArrowArray *))); |
1382 | 0 | psChild->children[0] = static_cast<struct ArrowArray *>( |
1383 | 0 | CPLCalloc(1, sizeof(struct ArrowArray))); |
1384 | 0 | auto psValueChild = psChild->children[0]; |
1385 | |
|
1386 | 0 | psValueChild->release = OGRLayerDefaultReleaseArray; |
1387 | 0 | psValueChild->length = nStrings; |
1388 | 0 | psValueChild->n_buffers = 3; |
1389 | 0 | psValueChild->buffers = |
1390 | 0 | static_cast<const void **>(CPLCalloc(3, sizeof(void *))); |
1391 | |
|
1392 | 0 | OffsetType *panChildOffsets = static_cast<OffsetType *>( |
1393 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(sizeof(OffsetType) * (1 + nStrings))); |
1394 | 0 | if (panChildOffsets == nullptr) |
1395 | 0 | return 0; |
1396 | 0 | psValueChild->buffers[1] = panChildOffsets; |
1397 | |
|
1398 | 0 | char *pachValues = |
1399 | 0 | static_cast<char *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE(nCountChars + 1)); |
1400 | 0 | if (pachValues == nullptr) |
1401 | 0 | return 0; |
1402 | 0 | psValueChild->buffers[2] = pachValues; |
1403 | |
|
1404 | 0 | nStrings = 0; |
1405 | 0 | nCountChars = 0; |
1406 | 0 | for (size_t iFeat = 0; iFeat < nFeatCount; ++iFeat) |
1407 | 0 | { |
1408 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1409 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1410 | 0 | if (IsValidField(psRawField)) |
1411 | 0 | { |
1412 | 0 | const int nCount = psRawField->StringList.nCount; |
1413 | 0 | for (int j = 0; j < nCount; ++j) |
1414 | 0 | { |
1415 | 0 | panChildOffsets[nStrings] = nCountChars; |
1416 | 0 | ++nStrings; |
1417 | 0 | const size_t nLen = strlen(psRawField->StringList.paList[j]); |
1418 | 0 | memcpy(pachValues + nCountChars, |
1419 | 0 | psRawField->StringList.paList[j], nLen); |
1420 | 0 | nCountChars += static_cast<OffsetType>(nLen); |
1421 | 0 | } |
1422 | 0 | } |
1423 | 0 | } |
1424 | 0 | panChildOffsets[nStrings] = nCountChars; |
1425 | |
|
1426 | 0 | return nFeatCount; |
1427 | 0 | } |
1428 | | |
1429 | | /************************************************************************/ |
1430 | | /* FillBinaryArray() */ |
1431 | | /************************************************************************/ |
1432 | | |
1433 | | template <class T> |
1434 | | static size_t |
1435 | | FillBinaryArray(struct ArrowArray *psChild, |
1436 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1437 | | const size_t nFeatureCountLimit, const bool bIsNullable, |
1438 | | const int i, const size_t nMemLimit) |
1439 | 0 | { |
1440 | 0 | psChild->n_buffers = 3; |
1441 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(3, sizeof(void *))); |
1442 | 0 | uint8_t *pabyValidity = nullptr; |
1443 | 0 | T *panOffsets = static_cast<T *>( |
1444 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(sizeof(T) * (1 + nFeatureCountLimit))); |
1445 | 0 | if (panOffsets == nullptr) |
1446 | 0 | return 0; |
1447 | 0 | psChild->buffers[1] = panOffsets; |
1448 | |
|
1449 | 0 | T nOffset = 0; |
1450 | 0 | size_t nFeatCount = 0; |
1451 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat, ++nFeatCount) |
1452 | 0 | { |
1453 | 0 | panOffsets[iFeat] = nOffset; |
1454 | 0 | const auto psRawField = apoFeatures[iFeat]->GetRawFieldRef(i); |
1455 | 0 | if (IsValidField(psRawField)) |
1456 | 0 | { |
1457 | 0 | const size_t nLen = psRawField->Binary.nCount; |
1458 | 0 | if (nLen > static_cast<size_t>(nMemLimit - nOffset)) |
1459 | 0 | { |
1460 | 0 | if (iFeat == 0) |
1461 | 0 | return 0; |
1462 | 0 | break; |
1463 | 0 | } |
1464 | 0 | nOffset += static_cast<T>(nLen); |
1465 | 0 | } |
1466 | 0 | else if (bIsNullable) |
1467 | 0 | { |
1468 | 0 | ++psChild->null_count; |
1469 | 0 | if (pabyValidity == nullptr) |
1470 | 0 | { |
1471 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1472 | 0 | psChild->buffers[0] = pabyValidity; |
1473 | 0 | if (pabyValidity == nullptr) |
1474 | 0 | return 0; |
1475 | 0 | } |
1476 | 0 | UnsetBit(pabyValidity, iFeat); |
1477 | 0 | } |
1478 | 0 | } |
1479 | 0 | panOffsets[nFeatCount] = nOffset; |
1480 | |
|
1481 | 0 | GByte *pabyValues = |
1482 | 0 | static_cast<GByte *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE(nOffset + 1)); |
1483 | 0 | if (pabyValues == nullptr) |
1484 | 0 | return 0; |
1485 | 0 | psChild->buffers[2] = pabyValues; |
1486 | |
|
1487 | 0 | nOffset = 0; |
1488 | 0 | for (size_t iFeat = 0; iFeat < nFeatCount; ++iFeat) |
1489 | 0 | { |
1490 | 0 | const size_t nLen = |
1491 | 0 | static_cast<size_t>(panOffsets[iFeat + 1] - panOffsets[iFeat]); |
1492 | 0 | if (nLen) |
1493 | 0 | { |
1494 | 0 | const auto psRawField = apoFeatures[iFeat]->GetRawFieldRef(i); |
1495 | 0 | memcpy(pabyValues + nOffset, psRawField->Binary.paData, nLen); |
1496 | 0 | nOffset += static_cast<T>(nLen); |
1497 | 0 | } |
1498 | 0 | } |
1499 | |
|
1500 | 0 | return nFeatCount; |
1501 | 0 | } |
1502 | | |
1503 | | /************************************************************************/ |
1504 | | /* FillFixedWidthBinaryArray() */ |
1505 | | /************************************************************************/ |
1506 | | |
1507 | | static bool |
1508 | | FillFixedWidthBinaryArray(struct ArrowArray *psChild, |
1509 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1510 | | const size_t nFeatureCountLimit, |
1511 | | const bool bIsNullable, const int nWidth, const int i) |
1512 | 0 | { |
1513 | 0 | psChild->n_buffers = 2; |
1514 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(3, sizeof(void *))); |
1515 | 0 | uint8_t *pabyValidity = nullptr; |
1516 | |
|
1517 | 0 | assert(nFeatureCountLimit + 1 <= |
1518 | 0 | std::numeric_limits<size_t>::max() / nWidth); |
1519 | 0 | GByte *pabyValues = static_cast<GByte *>( |
1520 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE((nFeatureCountLimit + 1) * nWidth)); |
1521 | 0 | if (pabyValues == nullptr) |
1522 | 0 | return false; |
1523 | 0 | psChild->buffers[1] = pabyValues; |
1524 | |
|
1525 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat) |
1526 | 0 | { |
1527 | 0 | const auto psRawField = apoFeatures[iFeat]->GetRawFieldRef(i); |
1528 | 0 | if (IsValidField(psRawField)) |
1529 | 0 | { |
1530 | 0 | const auto nLen = psRawField->Binary.nCount; |
1531 | 0 | if (nLen < nWidth) |
1532 | 0 | { |
1533 | 0 | memcpy(pabyValues + iFeat * nWidth, psRawField->Binary.paData, |
1534 | 0 | nLen); |
1535 | 0 | memset(pabyValues + iFeat * nWidth + nLen, 0, nWidth - nLen); |
1536 | 0 | } |
1537 | 0 | else |
1538 | 0 | { |
1539 | 0 | memcpy(pabyValues + iFeat * nWidth, psRawField->Binary.paData, |
1540 | 0 | nWidth); |
1541 | 0 | } |
1542 | 0 | } |
1543 | 0 | else |
1544 | 0 | { |
1545 | 0 | memset(pabyValues + iFeat * nWidth, 0, nWidth); |
1546 | 0 | if (bIsNullable) |
1547 | 0 | { |
1548 | 0 | ++psChild->null_count; |
1549 | 0 | if (pabyValidity == nullptr) |
1550 | 0 | { |
1551 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1552 | 0 | psChild->buffers[0] = pabyValidity; |
1553 | 0 | if (pabyValidity == nullptr) |
1554 | 0 | return false; |
1555 | 0 | } |
1556 | 0 | UnsetBit(pabyValidity, iFeat); |
1557 | 0 | } |
1558 | 0 | } |
1559 | 0 | } |
1560 | | |
1561 | 0 | return true; |
1562 | 0 | } |
1563 | | |
1564 | | /************************************************************************/ |
1565 | | /* FillWKBGeometryArray() */ |
1566 | | /************************************************************************/ |
1567 | | |
1568 | | template <class T> |
1569 | | static size_t |
1570 | | FillWKBGeometryArray(struct ArrowArray *psChild, |
1571 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1572 | | const size_t nFeatureCountLimit, |
1573 | | const OGRGeomFieldDefn *poFieldDefn, const int i, |
1574 | | const size_t nMemLimit) |
1575 | 0 | { |
1576 | 0 | const bool bIsNullable = CPL_TO_BOOL(poFieldDefn->IsNullable()); |
1577 | 0 | psChild->n_buffers = 3; |
1578 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(3, sizeof(void *))); |
1579 | 0 | uint8_t *pabyValidity = nullptr; |
1580 | 0 | T *panOffsets = static_cast<T *>( |
1581 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(sizeof(T) * (1 + nFeatureCountLimit))); |
1582 | 0 | if (panOffsets == nullptr) |
1583 | 0 | return 0; |
1584 | 0 | psChild->buffers[1] = panOffsets; |
1585 | 0 | const auto eGeomType = poFieldDefn->GetType(); |
1586 | 0 | auto poEmptyGeom = |
1587 | 0 | std::unique_ptr<OGRGeometry>(OGRGeometryFactory::createGeometry( |
1588 | 0 | (eGeomType == wkbNone || wkbFlatten(eGeomType) == wkbUnknown) |
1589 | 0 | ? wkbGeometryCollection |
1590 | 0 | : eGeomType)); |
1591 | |
|
1592 | 0 | size_t nOffset = 0; |
1593 | 0 | size_t nFeatCount = 0; |
1594 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat, ++nFeatCount) |
1595 | 0 | { |
1596 | 0 | panOffsets[iFeat] = static_cast<T>(nOffset); |
1597 | 0 | const auto poGeom = apoFeatures[iFeat]->GetGeomFieldRef(i); |
1598 | 0 | if (poGeom != nullptr) |
1599 | 0 | { |
1600 | 0 | const size_t nLen = poGeom->WkbSize(); |
1601 | 0 | if (nLen > nMemLimit - nOffset) |
1602 | 0 | { |
1603 | 0 | if (nFeatCount == 0) |
1604 | 0 | return 0; |
1605 | 0 | break; |
1606 | 0 | } |
1607 | 0 | nOffset += static_cast<T>(nLen); |
1608 | 0 | } |
1609 | 0 | else if (bIsNullable) |
1610 | 0 | { |
1611 | 0 | ++psChild->null_count; |
1612 | 0 | if (pabyValidity == nullptr) |
1613 | 0 | { |
1614 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1615 | 0 | psChild->buffers[0] = pabyValidity; |
1616 | 0 | if (pabyValidity == nullptr) |
1617 | 0 | return 0; |
1618 | 0 | } |
1619 | 0 | UnsetBit(pabyValidity, iFeat); |
1620 | 0 | } |
1621 | 0 | else if (poEmptyGeom) |
1622 | 0 | { |
1623 | 0 | const size_t nLen = poEmptyGeom->WkbSize(); |
1624 | 0 | if (nLen > nMemLimit - nOffset) |
1625 | 0 | { |
1626 | 0 | if (nFeatCount == 0) |
1627 | 0 | return 0; |
1628 | 0 | break; |
1629 | 0 | } |
1630 | 0 | nOffset += static_cast<T>(nLen); |
1631 | 0 | } |
1632 | 0 | } |
1633 | 0 | panOffsets[nFeatCount] = static_cast<T>(nOffset); |
1634 | |
|
1635 | 0 | GByte *pabyValues = |
1636 | 0 | static_cast<GByte *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE(nOffset + 1)); |
1637 | 0 | if (pabyValues == nullptr) |
1638 | 0 | return 0; |
1639 | 0 | psChild->buffers[2] = pabyValues; |
1640 | |
|
1641 | 0 | nOffset = 0; |
1642 | 0 | for (size_t iFeat = 0; iFeat < nFeatCount; ++iFeat) |
1643 | 0 | { |
1644 | 0 | const size_t nLen = |
1645 | 0 | static_cast<size_t>(panOffsets[iFeat + 1] - panOffsets[iFeat]); |
1646 | 0 | if (nLen) |
1647 | 0 | { |
1648 | 0 | const auto poGeom = apoFeatures[iFeat]->GetGeomFieldRef(i); |
1649 | 0 | poGeom->exportToWkb(wkbNDR, pabyValues + nOffset, wkbVariantIso); |
1650 | 0 | nOffset += nLen; |
1651 | 0 | } |
1652 | 0 | else if (!bIsNullable && poEmptyGeom) |
1653 | 0 | { |
1654 | 0 | poEmptyGeom->exportToWkb(wkbNDR, pabyValues + nOffset, |
1655 | 0 | wkbVariantIso); |
1656 | 0 | nOffset += nLen; |
1657 | 0 | } |
1658 | 0 | } |
1659 | |
|
1660 | 0 | return nFeatCount; |
1661 | 0 | } |
1662 | | |
1663 | | /************************************************************************/ |
1664 | | /* FillDateArray() */ |
1665 | | /************************************************************************/ |
1666 | | |
1667 | | static bool FillDateArray(struct ArrowArray *psChild, |
1668 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1669 | | const size_t nFeatureCountLimit, |
1670 | | const bool bIsNullable, const int i) |
1671 | 0 | { |
1672 | 0 | psChild->n_buffers = 2; |
1673 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
1674 | 0 | uint8_t *pabyValidity = nullptr; |
1675 | 0 | int32_t *panValues = static_cast<int32_t *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE( |
1676 | 0 | sizeof(int32_t) * (nFeatureCountLimit + 1))); |
1677 | 0 | if (panValues == nullptr) |
1678 | 0 | return false; |
1679 | 0 | psChild->buffers[1] = panValues; |
1680 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat) |
1681 | 0 | { |
1682 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1683 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1684 | 0 | if (IsValidField(psRawField)) |
1685 | 0 | { |
1686 | 0 | struct tm brokenDown; |
1687 | 0 | memset(&brokenDown, 0, sizeof(brokenDown)); |
1688 | 0 | brokenDown.tm_year = psRawField->Date.Year - 1900; |
1689 | 0 | brokenDown.tm_mon = psRawField->Date.Month - 1; |
1690 | 0 | brokenDown.tm_mday = psRawField->Date.Day; |
1691 | 0 | panValues[iFeat] = |
1692 | 0 | static_cast<int>(CPLYMDHMSToUnixTime(&brokenDown) / 86400); |
1693 | 0 | } |
1694 | 0 | else if (bIsNullable) |
1695 | 0 | { |
1696 | 0 | panValues[iFeat] = 0; |
1697 | 0 | ++psChild->null_count; |
1698 | 0 | if (pabyValidity == nullptr) |
1699 | 0 | { |
1700 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1701 | 0 | psChild->buffers[0] = pabyValidity; |
1702 | 0 | if (pabyValidity == nullptr) |
1703 | 0 | return false; |
1704 | 0 | } |
1705 | 0 | UnsetBit(pabyValidity, iFeat); |
1706 | 0 | } |
1707 | 0 | else |
1708 | 0 | { |
1709 | 0 | panValues[iFeat] = 0; |
1710 | 0 | } |
1711 | 0 | } |
1712 | 0 | return true; |
1713 | 0 | } |
1714 | | |
1715 | | /************************************************************************/ |
1716 | | /* FillTimeArray() */ |
1717 | | /************************************************************************/ |
1718 | | |
1719 | | static bool FillTimeArray(struct ArrowArray *psChild, |
1720 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1721 | | const size_t nFeatureCountLimit, |
1722 | | const bool bIsNullable, const int i) |
1723 | 0 | { |
1724 | 0 | psChild->n_buffers = 2; |
1725 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
1726 | 0 | uint8_t *pabyValidity = nullptr; |
1727 | 0 | int32_t *panValues = static_cast<int32_t *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE( |
1728 | 0 | sizeof(int32_t) * (nFeatureCountLimit + 1))); |
1729 | 0 | if (panValues == nullptr) |
1730 | 0 | return false; |
1731 | 0 | psChild->buffers[1] = panValues; |
1732 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat) |
1733 | 0 | { |
1734 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1735 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1736 | 0 | if (IsValidField(psRawField)) |
1737 | 0 | { |
1738 | 0 | panValues[iFeat] = |
1739 | 0 | psRawField->Date.Hour * 3600000 + |
1740 | 0 | psRawField->Date.Minute * 60000 + |
1741 | 0 | static_cast<int>(psRawField->Date.Second * 1000 + 0.5); |
1742 | 0 | } |
1743 | 0 | else if (bIsNullable) |
1744 | 0 | { |
1745 | 0 | panValues[iFeat] = 0; |
1746 | 0 | ++psChild->null_count; |
1747 | 0 | if (pabyValidity == nullptr) |
1748 | 0 | { |
1749 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1750 | 0 | psChild->buffers[0] = pabyValidity; |
1751 | 0 | if (pabyValidity == nullptr) |
1752 | 0 | return false; |
1753 | 0 | } |
1754 | 0 | UnsetBit(pabyValidity, iFeat); |
1755 | 0 | } |
1756 | 0 | else |
1757 | 0 | { |
1758 | 0 | panValues[iFeat] = 0; |
1759 | 0 | } |
1760 | 0 | } |
1761 | 0 | return true; |
1762 | 0 | } |
1763 | | |
1764 | | /************************************************************************/ |
1765 | | /* FillDateTimeArray() */ |
1766 | | /************************************************************************/ |
1767 | | |
1768 | | static bool |
1769 | | FillDateTimeArray(struct ArrowArray *psChild, |
1770 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1771 | | const size_t nFeatureCountLimit, const bool bIsNullable, |
1772 | | const int i, int nFieldTZFlag) |
1773 | 0 | { |
1774 | 0 | psChild->n_buffers = 2; |
1775 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
1776 | 0 | uint8_t *pabyValidity = nullptr; |
1777 | 0 | int64_t *panValues = static_cast<int64_t *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE( |
1778 | 0 | sizeof(int64_t) * (nFeatureCountLimit + 1))); |
1779 | 0 | if (panValues == nullptr) |
1780 | 0 | return false; |
1781 | 0 | psChild->buffers[1] = panValues; |
1782 | 0 | struct tm brokenDown; |
1783 | 0 | memset(&brokenDown, 0, sizeof(brokenDown)); |
1784 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat) |
1785 | 0 | { |
1786 | 0 | auto &poFeature = apoFeatures[iFeat]; |
1787 | 0 | const auto psRawField = poFeature->GetRawFieldRef(i); |
1788 | 0 | if (IsValidField(psRawField)) |
1789 | 0 | { |
1790 | 0 | brokenDown.tm_year = psRawField->Date.Year - 1900; |
1791 | 0 | brokenDown.tm_mon = psRawField->Date.Month - 1; |
1792 | 0 | brokenDown.tm_mday = psRawField->Date.Day; |
1793 | 0 | brokenDown.tm_hour = psRawField->Date.Hour; |
1794 | 0 | brokenDown.tm_min = psRawField->Date.Minute; |
1795 | 0 | brokenDown.tm_sec = static_cast<int>(psRawField->Date.Second); |
1796 | 0 | auto nVal = |
1797 | 0 | CPLYMDHMSToUnixTime(&brokenDown) * 1000 + |
1798 | 0 | (static_cast<int>(psRawField->Date.Second * 1000 + 0.5) % 1000); |
1799 | 0 | if (nFieldTZFlag >= OGR_TZFLAG_MIXED_TZ && |
1800 | 0 | psRawField->Date.TZFlag > OGR_TZFLAG_MIXED_TZ) |
1801 | 0 | { |
1802 | | // Convert for psRawField->Date.TZFlag to UTC |
1803 | 0 | const int TZOffset = |
1804 | 0 | (psRawField->Date.TZFlag - OGR_TZFLAG_UTC) * 15; |
1805 | 0 | const int TZOffsetMS = TZOffset * 60 * 1000; |
1806 | 0 | nVal -= TZOffsetMS; |
1807 | 0 | } |
1808 | 0 | panValues[iFeat] = nVal; |
1809 | 0 | } |
1810 | 0 | else if (bIsNullable) |
1811 | 0 | { |
1812 | 0 | panValues[iFeat] = 0; |
1813 | 0 | ++psChild->null_count; |
1814 | 0 | if (pabyValidity == nullptr) |
1815 | 0 | { |
1816 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1817 | 0 | psChild->buffers[0] = pabyValidity; |
1818 | 0 | if (pabyValidity == nullptr) |
1819 | 0 | return false; |
1820 | 0 | } |
1821 | 0 | UnsetBit(pabyValidity, iFeat); |
1822 | 0 | } |
1823 | 0 | else |
1824 | 0 | { |
1825 | 0 | panValues[iFeat] = 0; |
1826 | 0 | } |
1827 | 0 | } |
1828 | 0 | return true; |
1829 | 0 | } |
1830 | | |
1831 | | /************************************************************************/ |
1832 | | /* FillDateTimeArrayAsString() */ |
1833 | | /************************************************************************/ |
1834 | | |
1835 | | static size_t |
1836 | | FillDateTimeArrayAsString(struct ArrowArray *psChild, |
1837 | | std::deque<std::unique_ptr<OGRFeature>> &apoFeatures, |
1838 | | const size_t nFeatureCountLimit, |
1839 | | const bool bIsNullable, const int i, |
1840 | | const size_t nMemLimit) |
1841 | 0 | { |
1842 | 0 | psChild->n_buffers = 3; |
1843 | 0 | psChild->buffers = static_cast<const void **>(CPLCalloc(3, sizeof(void *))); |
1844 | 0 | uint8_t *pabyValidity = nullptr; |
1845 | 0 | using T = uint32_t; |
1846 | 0 | T *panOffsets = static_cast<T *>( |
1847 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(sizeof(T) * (1 + nFeatureCountLimit))); |
1848 | 0 | if (panOffsets == nullptr) |
1849 | 0 | return 0; |
1850 | 0 | psChild->buffers[1] = panOffsets; |
1851 | |
|
1852 | 0 | size_t nOffset = 0; |
1853 | 0 | size_t nFeatCount = 0; |
1854 | 0 | for (size_t iFeat = 0; iFeat < nFeatureCountLimit; ++iFeat, ++nFeatCount) |
1855 | 0 | { |
1856 | 0 | panOffsets[iFeat] = static_cast<T>(nOffset); |
1857 | 0 | const auto psRawField = apoFeatures[iFeat]->GetRawFieldRef(i); |
1858 | 0 | if (IsValidField(psRawField)) |
1859 | 0 | { |
1860 | 0 | size_t nLen = strlen("YYYY-MM-DDTHH:MM:SS"); |
1861 | 0 | if (fmodf(psRawField->Date.Second, 1.0f) != 0) |
1862 | 0 | nLen += strlen(".sss"); |
1863 | 0 | if (psRawField->Date.TZFlag == OGR_TZFLAG_UTC) |
1864 | 0 | nLen += 1; // 'Z' |
1865 | 0 | else if (psRawField->Date.TZFlag > OGR_TZFLAG_MIXED_TZ) |
1866 | 0 | nLen += strlen("+hh:mm"); |
1867 | 0 | if (nLen > nMemLimit - nOffset) |
1868 | 0 | { |
1869 | 0 | if (nFeatCount == 0) |
1870 | 0 | return 0; |
1871 | 0 | break; |
1872 | 0 | } |
1873 | 0 | nOffset += static_cast<T>(nLen); |
1874 | 0 | } |
1875 | 0 | else if (bIsNullable) |
1876 | 0 | { |
1877 | 0 | ++psChild->null_count; |
1878 | 0 | if (pabyValidity == nullptr) |
1879 | 0 | { |
1880 | 0 | pabyValidity = AllocValidityBitmap(nFeatureCountLimit); |
1881 | 0 | psChild->buffers[0] = pabyValidity; |
1882 | 0 | if (pabyValidity == nullptr) |
1883 | 0 | return 0; |
1884 | 0 | } |
1885 | 0 | UnsetBit(pabyValidity, iFeat); |
1886 | 0 | } |
1887 | 0 | } |
1888 | 0 | panOffsets[nFeatCount] = static_cast<T>(nOffset); |
1889 | |
|
1890 | 0 | char *pachValues = |
1891 | 0 | static_cast<char *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE(nOffset + 1)); |
1892 | 0 | if (pachValues == nullptr) |
1893 | 0 | return 0; |
1894 | 0 | psChild->buffers[2] = pachValues; |
1895 | |
|
1896 | 0 | nOffset = 0; |
1897 | 0 | char szBuffer[OGR_SIZEOF_ISO8601_DATETIME_BUFFER]; |
1898 | 0 | OGRISO8601Format sFormat; |
1899 | 0 | sFormat.ePrecision = OGRISO8601Precision::AUTO; |
1900 | 0 | for (size_t iFeat = 0; iFeat < nFeatCount; ++iFeat) |
1901 | 0 | { |
1902 | 0 | const int nLen = |
1903 | 0 | static_cast<int>(panOffsets[iFeat + 1] - panOffsets[iFeat]); |
1904 | 0 | if (nLen) |
1905 | 0 | { |
1906 | 0 | const auto psRawField = apoFeatures[iFeat]->GetRawFieldRef(i); |
1907 | 0 | int nBufSize = OGRGetISO8601DateTime(psRawField, sFormat, szBuffer); |
1908 | 0 | if (nBufSize) |
1909 | 0 | { |
1910 | 0 | memcpy(pachValues + nOffset, szBuffer, |
1911 | 0 | std::min(nLen, nBufSize)); |
1912 | 0 | } |
1913 | 0 | if (nBufSize < nLen) |
1914 | 0 | { |
1915 | 0 | memset(pachValues + nOffset + nBufSize, 0, nLen - nBufSize); |
1916 | 0 | } |
1917 | 0 | nOffset += nLen; |
1918 | 0 | } |
1919 | 0 | } |
1920 | |
|
1921 | 0 | return nFeatCount; |
1922 | 0 | } |
1923 | | |
1924 | | /************************************************************************/ |
1925 | | /* GetNextArrowArray() */ |
1926 | | /************************************************************************/ |
1927 | | |
1928 | | /** Default implementation of the ArrowArrayStream::get_next() callback. |
1929 | | * |
1930 | | * To be used by driver implementations that have a custom GetArrowStream() |
1931 | | * implementation. |
1932 | | * |
1933 | | * @since GDAL 3.6 |
1934 | | */ |
1935 | | int OGRLayer::GetNextArrowArray(struct ArrowArrayStream *stream, |
1936 | | struct ArrowArray *out_array) |
1937 | 0 | { |
1938 | 0 | ArrowArrayStreamPrivateDataSharedDataWrapper *poPrivate = |
1939 | 0 | static_cast<ArrowArrayStreamPrivateDataSharedDataWrapper *>( |
1940 | 0 | stream->private_data); |
1941 | |
|
1942 | 0 | const bool bIncludeFID = CPLTestBool( |
1943 | 0 | m_aosArrowArrayStreamOptions.FetchNameValueDef("INCLUDE_FID", "YES")); |
1944 | 0 | const bool bDateTimeAsString = m_aosArrowArrayStreamOptions.FetchBool( |
1945 | 0 | GAS_OPT_DATETIME_AS_STRING, false); |
1946 | 0 | int nMaxBatchSize = atoi(m_aosArrowArrayStreamOptions.FetchNameValueDef( |
1947 | 0 | "MAX_FEATURES_IN_BATCH", "65536")); |
1948 | 0 | if (nMaxBatchSize <= 0) |
1949 | 0 | nMaxBatchSize = 1; |
1950 | 0 | if (nMaxBatchSize > INT_MAX - 1) |
1951 | 0 | nMaxBatchSize = INT_MAX - 1; |
1952 | |
|
1953 | 0 | auto &oFeatureQueue = |
1954 | 0 | m_poSharedArrowArrayStreamPrivateData->m_oFeatureQueue; |
1955 | |
|
1956 | 0 | memset(out_array, 0, sizeof(*out_array)); |
1957 | |
|
1958 | 0 | auto poLayerDefn = GetLayerDefn(); |
1959 | 0 | const int nFieldCount = poLayerDefn->GetFieldCount(); |
1960 | 0 | const int nGeomFieldCount = poLayerDefn->GetGeomFieldCount(); |
1961 | 0 | const int nMaxChildren = |
1962 | 0 | (bIncludeFID ? 1 : 0) + nFieldCount + nGeomFieldCount; |
1963 | 0 | int iSchemaChild = 0; |
1964 | |
|
1965 | 0 | if (!m_poSharedArrowArrayStreamPrivateData->m_anQueriedFIDs.empty()) |
1966 | 0 | { |
1967 | 0 | if (poPrivate->poShared->m_bEOF) |
1968 | 0 | { |
1969 | 0 | return 0; |
1970 | 0 | } |
1971 | 0 | if (m_poSharedArrowArrayStreamPrivateData->m_iQueriedFIDS == 0) |
1972 | 0 | { |
1973 | 0 | CPLDebug("OGR", "Using fast FID filtering"); |
1974 | 0 | } |
1975 | 0 | while ( |
1976 | 0 | oFeatureQueue.size() < static_cast<size_t>(nMaxBatchSize) && |
1977 | 0 | m_poSharedArrowArrayStreamPrivateData->m_iQueriedFIDS < |
1978 | 0 | m_poSharedArrowArrayStreamPrivateData->m_anQueriedFIDs.size()) |
1979 | 0 | { |
1980 | 0 | const auto nFID = |
1981 | 0 | m_poSharedArrowArrayStreamPrivateData->m_anQueriedFIDs |
1982 | 0 | [m_poSharedArrowArrayStreamPrivateData->m_iQueriedFIDS]; |
1983 | 0 | auto poFeature = std::unique_ptr<OGRFeature>(GetFeature(nFID)); |
1984 | 0 | ++m_poSharedArrowArrayStreamPrivateData->m_iQueriedFIDS; |
1985 | 0 | if (poFeature && (m_poFilterGeom == nullptr || |
1986 | 0 | FilterGeometry(poFeature->GetGeomFieldRef( |
1987 | 0 | m_iGeomFieldFilter)))) |
1988 | 0 | { |
1989 | 0 | oFeatureQueue.emplace_back(std::move(poFeature)); |
1990 | 0 | } |
1991 | 0 | } |
1992 | 0 | if (m_poSharedArrowArrayStreamPrivateData->m_iQueriedFIDS == |
1993 | 0 | m_poSharedArrowArrayStreamPrivateData->m_anQueriedFIDs.size()) |
1994 | 0 | { |
1995 | 0 | poPrivate->poShared->m_bEOF = true; |
1996 | 0 | } |
1997 | 0 | } |
1998 | 0 | else if (!poPrivate->poShared->m_bEOF) |
1999 | 0 | { |
2000 | 0 | while (oFeatureQueue.size() < static_cast<size_t>(nMaxBatchSize)) |
2001 | 0 | { |
2002 | 0 | auto poFeature = std::unique_ptr<OGRFeature>(GetNextFeature()); |
2003 | 0 | if (!poFeature) |
2004 | 0 | { |
2005 | 0 | poPrivate->poShared->m_bEOF = true; |
2006 | 0 | break; |
2007 | 0 | } |
2008 | 0 | oFeatureQueue.emplace_back(std::move(poFeature)); |
2009 | 0 | } |
2010 | 0 | } |
2011 | 0 | if (oFeatureQueue.empty()) |
2012 | 0 | { |
2013 | 0 | return 0; |
2014 | 0 | } |
2015 | | |
2016 | 0 | out_array->release = OGRLayerDefaultReleaseArray; |
2017 | 0 | out_array->null_count = 0; |
2018 | |
|
2019 | 0 | out_array->n_children = nMaxChildren; |
2020 | 0 | out_array->children = static_cast<struct ArrowArray **>( |
2021 | 0 | CPLCalloc(nMaxChildren, sizeof(struct ArrowArray *))); |
2022 | 0 | out_array->release = OGRLayerDefaultReleaseArray; |
2023 | 0 | out_array->n_buffers = 1; |
2024 | 0 | out_array->buffers = |
2025 | 0 | static_cast<const void **>(CPLCalloc(1, sizeof(void *))); |
2026 | |
|
2027 | 0 | size_t nFeatureCount = oFeatureQueue.size(); |
2028 | 0 | const uint32_t nMemLimit = OGRArrowArrayHelper::GetMemLimit(); |
2029 | 0 | if (bIncludeFID) |
2030 | 0 | { |
2031 | 0 | out_array->children[iSchemaChild] = static_cast<struct ArrowArray *>( |
2032 | 0 | CPLCalloc(1, sizeof(struct ArrowArray))); |
2033 | 0 | auto psChild = out_array->children[iSchemaChild]; |
2034 | 0 | ++iSchemaChild; |
2035 | 0 | psChild->release = OGRLayerDefaultReleaseArray; |
2036 | 0 | psChild->n_buffers = 2; |
2037 | 0 | psChild->buffers = |
2038 | 0 | static_cast<const void **>(CPLCalloc(2, sizeof(void *))); |
2039 | 0 | int64_t *panValues = |
2040 | 0 | static_cast<int64_t *>(VSI_MALLOC_ALIGNED_AUTO_VERBOSE( |
2041 | 0 | sizeof(int64_t) * (oFeatureQueue.size() + 1))); |
2042 | 0 | if (panValues == nullptr) |
2043 | 0 | goto error; |
2044 | 0 | psChild->buffers[1] = panValues; |
2045 | 0 | for (size_t iFeat = 0; iFeat < oFeatureQueue.size(); ++iFeat) |
2046 | 0 | { |
2047 | 0 | panValues[iFeat] = oFeatureQueue[iFeat]->GetFID(); |
2048 | 0 | } |
2049 | 0 | } |
2050 | | |
2051 | 0 | for (int i = 0; i < nFieldCount; ++i) |
2052 | 0 | { |
2053 | 0 | const auto poFieldDefn = poLayerDefn->GetFieldDefn(i); |
2054 | 0 | if (poFieldDefn->IsIgnored()) |
2055 | 0 | { |
2056 | 0 | continue; |
2057 | 0 | } |
2058 | | |
2059 | 0 | out_array->children[iSchemaChild] = static_cast<struct ArrowArray *>( |
2060 | 0 | CPLCalloc(1, sizeof(struct ArrowArray))); |
2061 | 0 | auto psChild = out_array->children[iSchemaChild]; |
2062 | 0 | ++iSchemaChild; |
2063 | 0 | psChild->release = OGRLayerDefaultReleaseArray; |
2064 | 0 | const bool bIsNullable = CPL_TO_BOOL(poFieldDefn->IsNullable()); |
2065 | 0 | const auto eSubType = poFieldDefn->GetSubType(); |
2066 | 0 | switch (poFieldDefn->GetType()) |
2067 | 0 | { |
2068 | 0 | case OFTInteger: |
2069 | 0 | { |
2070 | 0 | if (eSubType == OFSTBoolean) |
2071 | 0 | { |
2072 | 0 | if (!FillBoolArray(psChild, oFeatureQueue, nFeatureCount, |
2073 | 0 | bIsNullable, &OGRField::Integer, i)) |
2074 | 0 | goto error; |
2075 | 0 | } |
2076 | 0 | else if (eSubType == OFSTInt16) |
2077 | 0 | { |
2078 | 0 | if (!FillArray<int16_t>(psChild, oFeatureQueue, |
2079 | 0 | nFeatureCount, bIsNullable, |
2080 | 0 | &OGRField::Integer, i)) |
2081 | 0 | goto error; |
2082 | 0 | } |
2083 | 0 | else |
2084 | 0 | { |
2085 | 0 | if (!FillArray<int32_t>(psChild, oFeatureQueue, |
2086 | 0 | nFeatureCount, bIsNullable, |
2087 | 0 | &OGRField::Integer, i)) |
2088 | 0 | goto error; |
2089 | 0 | } |
2090 | | |
2091 | 0 | const auto &osDomainName = poFieldDefn->GetDomainName(); |
2092 | 0 | if (!osDomainName.empty()) |
2093 | 0 | { |
2094 | 0 | auto poDS = GetDataset(); |
2095 | 0 | if (poDS) |
2096 | 0 | { |
2097 | 0 | const auto poFieldDomain = |
2098 | 0 | poDS->GetFieldDomain(osDomainName); |
2099 | 0 | if (poFieldDomain && |
2100 | 0 | poFieldDomain->GetDomainType() == OFDT_CODED) |
2101 | 0 | { |
2102 | 0 | const OGRCodedFieldDomain *poCodedDomain = |
2103 | 0 | static_cast<const OGRCodedFieldDomain *>( |
2104 | 0 | poFieldDomain); |
2105 | 0 | OGRArrowArrayHelper::FillDict(psChild, |
2106 | 0 | poCodedDomain); |
2107 | 0 | } |
2108 | 0 | } |
2109 | 0 | } |
2110 | |
|
2111 | 0 | break; |
2112 | 0 | } |
2113 | | |
2114 | 0 | case OFTInteger64: |
2115 | 0 | { |
2116 | 0 | if (!FillArray<int64_t>(psChild, oFeatureQueue, nFeatureCount, |
2117 | 0 | bIsNullable, &OGRField::Integer64, i)) |
2118 | 0 | goto error; |
2119 | 0 | break; |
2120 | 0 | } |
2121 | | |
2122 | 0 | case OFTReal: |
2123 | 0 | { |
2124 | 0 | if (eSubType == OFSTFloat32) |
2125 | 0 | { |
2126 | 0 | if (!FillArray<float>(psChild, oFeatureQueue, nFeatureCount, |
2127 | 0 | bIsNullable, &OGRField::Real, i)) |
2128 | 0 | goto error; |
2129 | 0 | } |
2130 | 0 | else |
2131 | 0 | { |
2132 | 0 | if (!FillArray<double>(psChild, oFeatureQueue, |
2133 | 0 | nFeatureCount, bIsNullable, |
2134 | 0 | &OGRField::Real, i)) |
2135 | 0 | goto error; |
2136 | 0 | } |
2137 | 0 | break; |
2138 | 0 | } |
2139 | | |
2140 | 0 | case OFTString: |
2141 | 0 | case OFTWideString: |
2142 | 0 | { |
2143 | 0 | const size_t nThisFeatureCount = FillStringArray<int32_t>( |
2144 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, i, |
2145 | 0 | nMemLimit); |
2146 | 0 | if (nThisFeatureCount == 0) |
2147 | 0 | { |
2148 | 0 | goto error_max_mem; |
2149 | 0 | } |
2150 | 0 | if (nThisFeatureCount < nFeatureCount) |
2151 | 0 | nFeatureCount = nThisFeatureCount; |
2152 | 0 | break; |
2153 | 0 | } |
2154 | | |
2155 | 0 | case OFTBinary: |
2156 | 0 | { |
2157 | 0 | const int nWidth = poFieldDefn->GetWidth(); |
2158 | 0 | if (nWidth > 0) |
2159 | 0 | { |
2160 | 0 | if (nFeatureCount > nMemLimit / nWidth) |
2161 | 0 | { |
2162 | 0 | nFeatureCount = nMemLimit / nWidth; |
2163 | 0 | if (nFeatureCount == 0) |
2164 | 0 | goto error_max_mem; |
2165 | 0 | } |
2166 | 0 | if (!FillFixedWidthBinaryArray(psChild, oFeatureQueue, |
2167 | 0 | nFeatureCount, bIsNullable, |
2168 | 0 | nWidth, i)) |
2169 | 0 | goto error; |
2170 | 0 | } |
2171 | 0 | else |
2172 | 0 | { |
2173 | 0 | const size_t nThisFeatureCount = FillBinaryArray<int32_t>( |
2174 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, i, |
2175 | 0 | nMemLimit); |
2176 | 0 | if (nThisFeatureCount == 0) |
2177 | 0 | { |
2178 | 0 | goto error_max_mem; |
2179 | 0 | } |
2180 | 0 | if (nThisFeatureCount < nFeatureCount) |
2181 | 0 | nFeatureCount = nThisFeatureCount; |
2182 | 0 | } |
2183 | 0 | break; |
2184 | 0 | } |
2185 | | |
2186 | 0 | case OFTIntegerList: |
2187 | 0 | { |
2188 | 0 | size_t nThisFeatureCount; |
2189 | 0 | if (eSubType == OFSTBoolean) |
2190 | 0 | { |
2191 | 0 | nThisFeatureCount = |
2192 | 0 | FillListArrayBool<int32_t, GetFromIntegerList>( |
2193 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, |
2194 | 0 | i, nMemLimit); |
2195 | 0 | } |
2196 | 0 | else if (eSubType == OFSTInt16) |
2197 | 0 | { |
2198 | 0 | nThisFeatureCount = |
2199 | 0 | FillListArray<int32_t, int16_t, GetFromIntegerList>( |
2200 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, |
2201 | 0 | i, nMemLimit); |
2202 | 0 | } |
2203 | 0 | else |
2204 | 0 | { |
2205 | 0 | nThisFeatureCount = |
2206 | 0 | FillListArray<int32_t, int32_t, GetFromIntegerList>( |
2207 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, |
2208 | 0 | i, nMemLimit); |
2209 | 0 | } |
2210 | 0 | if (nThisFeatureCount == 0) |
2211 | 0 | { |
2212 | 0 | goto error_max_mem; |
2213 | 0 | } |
2214 | 0 | if (nThisFeatureCount < nFeatureCount) |
2215 | 0 | nFeatureCount = nThisFeatureCount; |
2216 | 0 | break; |
2217 | 0 | } |
2218 | | |
2219 | 0 | case OFTInteger64List: |
2220 | 0 | { |
2221 | 0 | const size_t nThisFeatureCount = |
2222 | 0 | FillListArray<int32_t, int64_t, GetFromInteger64List>( |
2223 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, i, |
2224 | 0 | nMemLimit); |
2225 | 0 | if (nThisFeatureCount == 0) |
2226 | 0 | { |
2227 | 0 | goto error_max_mem; |
2228 | 0 | } |
2229 | 0 | if (nThisFeatureCount < nFeatureCount) |
2230 | 0 | nFeatureCount = nThisFeatureCount; |
2231 | 0 | break; |
2232 | 0 | } |
2233 | | |
2234 | 0 | case OFTRealList: |
2235 | 0 | { |
2236 | 0 | size_t nThisFeatureCount; |
2237 | 0 | if (eSubType == OFSTFloat32) |
2238 | 0 | { |
2239 | 0 | nThisFeatureCount = |
2240 | 0 | FillListArray<int32_t, float, GetFromRealList>( |
2241 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, |
2242 | 0 | i, nMemLimit); |
2243 | 0 | } |
2244 | 0 | else |
2245 | 0 | { |
2246 | 0 | nThisFeatureCount = |
2247 | 0 | FillListArray<int32_t, double, GetFromRealList>( |
2248 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, |
2249 | 0 | i, nMemLimit); |
2250 | 0 | } |
2251 | 0 | if (nThisFeatureCount == 0) |
2252 | 0 | { |
2253 | 0 | goto error_max_mem; |
2254 | 0 | } |
2255 | 0 | if (nThisFeatureCount < nFeatureCount) |
2256 | 0 | nFeatureCount = nThisFeatureCount; |
2257 | 0 | break; |
2258 | 0 | } |
2259 | | |
2260 | 0 | case OFTStringList: |
2261 | 0 | case OFTWideStringList: |
2262 | 0 | { |
2263 | 0 | const size_t nThisFeatureCount = FillStringListArray<int32_t>( |
2264 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, i, |
2265 | 0 | nMemLimit); |
2266 | 0 | if (nThisFeatureCount == 0) |
2267 | 0 | { |
2268 | 0 | goto error_max_mem; |
2269 | 0 | } |
2270 | 0 | if (nThisFeatureCount < nFeatureCount) |
2271 | 0 | nFeatureCount = nThisFeatureCount; |
2272 | 0 | break; |
2273 | 0 | } |
2274 | | |
2275 | 0 | case OFTDate: |
2276 | 0 | { |
2277 | 0 | if (!FillDateArray(psChild, oFeatureQueue, nFeatureCount, |
2278 | 0 | bIsNullable, i)) |
2279 | 0 | goto error; |
2280 | 0 | break; |
2281 | 0 | } |
2282 | | |
2283 | 0 | case OFTTime: |
2284 | 0 | { |
2285 | 0 | if (!FillTimeArray(psChild, oFeatureQueue, nFeatureCount, |
2286 | 0 | bIsNullable, i)) |
2287 | 0 | goto error; |
2288 | 0 | break; |
2289 | 0 | } |
2290 | | |
2291 | 0 | case OFTDateTime: |
2292 | 0 | { |
2293 | 0 | if (bDateTimeAsString) |
2294 | 0 | { |
2295 | 0 | const size_t nThisFeatureCount = FillDateTimeArrayAsString( |
2296 | 0 | psChild, oFeatureQueue, nFeatureCount, bIsNullable, i, |
2297 | 0 | nMemLimit); |
2298 | 0 | if (nThisFeatureCount == 0) |
2299 | 0 | { |
2300 | 0 | goto error_max_mem; |
2301 | 0 | } |
2302 | 0 | if (nThisFeatureCount < nFeatureCount) |
2303 | 0 | nFeatureCount = nThisFeatureCount; |
2304 | 0 | } |
2305 | 0 | else |
2306 | 0 | { |
2307 | 0 | if (!FillDateTimeArray(psChild, oFeatureQueue, |
2308 | 0 | nFeatureCount, bIsNullable, i, |
2309 | 0 | poFieldDefn->GetTZFlag())) |
2310 | 0 | goto error; |
2311 | 0 | } |
2312 | 0 | break; |
2313 | 0 | } |
2314 | 0 | } |
2315 | 0 | } |
2316 | 0 | for (int i = 0; i < nGeomFieldCount; ++i) |
2317 | 0 | { |
2318 | 0 | const auto poFieldDefn = poLayerDefn->GetGeomFieldDefn(i); |
2319 | 0 | if (poFieldDefn->IsIgnored()) |
2320 | 0 | { |
2321 | 0 | continue; |
2322 | 0 | } |
2323 | | |
2324 | 0 | out_array->children[iSchemaChild] = static_cast<struct ArrowArray *>( |
2325 | 0 | CPLCalloc(1, sizeof(struct ArrowArray))); |
2326 | 0 | auto psChild = out_array->children[iSchemaChild]; |
2327 | 0 | ++iSchemaChild; |
2328 | 0 | psChild->release = OGRLayerDefaultReleaseArray; |
2329 | 0 | psChild->length = oFeatureQueue.size(); |
2330 | 0 | const size_t nThisFeatureCount = FillWKBGeometryArray<int32_t>( |
2331 | 0 | psChild, oFeatureQueue, nFeatureCount, poFieldDefn, i, nMemLimit); |
2332 | 0 | if (nThisFeatureCount == 0) |
2333 | 0 | { |
2334 | 0 | goto error_max_mem; |
2335 | 0 | } |
2336 | 0 | if (nThisFeatureCount < nFeatureCount) |
2337 | 0 | nFeatureCount = nThisFeatureCount; |
2338 | 0 | } |
2339 | | |
2340 | | // Remove consumed features from the queue |
2341 | 0 | if (nFeatureCount == oFeatureQueue.size()) |
2342 | 0 | oFeatureQueue.clear(); |
2343 | 0 | else |
2344 | 0 | { |
2345 | 0 | for (size_t i = 0; i < nFeatureCount; ++i) |
2346 | 0 | { |
2347 | 0 | oFeatureQueue.pop_front(); |
2348 | 0 | } |
2349 | 0 | } |
2350 | |
|
2351 | 0 | out_array->n_children = iSchemaChild; |
2352 | 0 | out_array->length = nFeatureCount; |
2353 | 0 | for (int i = 0; i < out_array->n_children; ++i) |
2354 | 0 | { |
2355 | 0 | out_array->children[i]->length = nFeatureCount; |
2356 | 0 | } |
2357 | |
|
2358 | 0 | return 0; |
2359 | | |
2360 | 0 | error_max_mem: |
2361 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2362 | 0 | "Too large feature: not even a single feature can be returned"); |
2363 | 0 | error: |
2364 | 0 | oFeatureQueue.clear(); |
2365 | 0 | poPrivate->poShared->m_bEOF = true; |
2366 | 0 | out_array->release(out_array); |
2367 | 0 | memset(out_array, 0, sizeof(*out_array)); |
2368 | 0 | return ENOMEM; |
2369 | 0 | } |
2370 | | |
2371 | | /************************************************************************/ |
2372 | | /* StaticGetNextArrowArray() */ |
2373 | | /************************************************************************/ |
2374 | | |
2375 | | /** Default implementation of the ArrowArrayStream::get_next() callback. |
2376 | | * |
2377 | | * To be used by driver implementations that have a custom GetArrowStream() |
2378 | | * implementation. |
2379 | | * |
2380 | | * @since GDAL 3.6 |
2381 | | */ |
2382 | | int OGRLayer::StaticGetNextArrowArray(struct ArrowArrayStream *stream, |
2383 | | struct ArrowArray *out_array) |
2384 | 0 | { |
2385 | 0 | auto poLayer = static_cast<ArrowArrayStreamPrivateDataSharedDataWrapper *>( |
2386 | 0 | stream->private_data) |
2387 | 0 | ->poShared->m_poLayer; |
2388 | 0 | if (poLayer == nullptr) |
2389 | 0 | { |
2390 | 0 | CPLError(CE_Failure, CPLE_NotSupported, |
2391 | 0 | "Calling get_next() on a freed OGRLayer is not supported"); |
2392 | 0 | return EINVAL; |
2393 | 0 | } |
2394 | 0 | return poLayer->GetNextArrowArray(stream, out_array); |
2395 | 0 | } |
2396 | | |
2397 | | /************************************************************************/ |
2398 | | /* ReleaseStream() */ |
2399 | | /************************************************************************/ |
2400 | | |
2401 | | /** Release a ArrowArrayStream. |
2402 | | * |
2403 | | * To be used by driver implementations that have a custom GetArrowStream() |
2404 | | * implementation. |
2405 | | * |
2406 | | * @param stream Arrow array stream to release. |
2407 | | * @since GDAL 3.6 |
2408 | | */ |
2409 | | void OGRLayer::ReleaseStream(struct ArrowArrayStream *stream) |
2410 | 0 | { |
2411 | 0 | assert(stream->release == OGRLayer::ReleaseStream); |
2412 | 0 | ArrowArrayStreamPrivateDataSharedDataWrapper *poPrivate = |
2413 | 0 | static_cast<ArrowArrayStreamPrivateDataSharedDataWrapper *>( |
2414 | 0 | stream->private_data); |
2415 | 0 | poPrivate->poShared->m_bArrowArrayStreamInProgress = false; |
2416 | 0 | poPrivate->poShared->m_bEOF = false; |
2417 | 0 | if (poPrivate->poShared->m_poLayer) |
2418 | 0 | poPrivate->poShared->m_poLayer->ResetReading(); |
2419 | 0 | delete poPrivate; |
2420 | 0 | stream->private_data = nullptr; |
2421 | 0 | stream->release = nullptr; |
2422 | 0 | } |
2423 | | |
2424 | | /************************************************************************/ |
2425 | | /* GetLastErrorArrowArrayStream() */ |
2426 | | /************************************************************************/ |
2427 | | |
2428 | | /** Default implementation of the ArrowArrayStream::get_last_error() callback. |
2429 | | * |
2430 | | * To be used by driver implementations that have a custom GetArrowStream() |
2431 | | * implementation. |
2432 | | * |
2433 | | * @since GDAL 3.6 |
2434 | | */ |
2435 | | const char *OGRLayer::GetLastErrorArrowArrayStream(struct ArrowArrayStream *) |
2436 | 0 | { |
2437 | 0 | const char *pszLastErrorMsg = CPLGetLastErrorMsg(); |
2438 | 0 | return pszLastErrorMsg[0] != '\0' ? pszLastErrorMsg : nullptr; |
2439 | 0 | } |
2440 | | |
2441 | | /************************************************************************/ |
2442 | | /* GetArrowStream() */ |
2443 | | /************************************************************************/ |
2444 | | |
2445 | | /** Get a Arrow C stream. |
2446 | | * |
2447 | | * On successful return, and when the stream interfaces is no longer needed, it |
2448 | | * must must be freed with out_stream->release(out_stream). Please carefully |
2449 | | * read https://arrow.apache.org/docs/format/CStreamInterface.html for more |
2450 | | * details on using Arrow C stream. |
2451 | | * |
2452 | | * The method may take into account ignored fields set with SetIgnoredFields() |
2453 | | * (the default implementation does), and should take into account filters set |
2454 | | * with SetSpatialFilter() and SetAttributeFilter(). Note however that |
2455 | | * specialized implementations may fallback to the default (slower) |
2456 | | * implementation when filters are set. |
2457 | | * Drivers that have a specialized implementation should advertise the |
2458 | | * OLCFastGetArrowStream capability. |
2459 | | * |
2460 | | * There are extra precautions to take into account in a OGR context. Unless |
2461 | | * otherwise specified by a particular driver implementation, the get_schema(), |
2462 | | * get_next() and get_last_error() function pointers of the ArrowArrayStream |
2463 | | * structure should no longer be used after the OGRLayer, from which the |
2464 | | * ArrowArrayStream structure was initialized, has been destroyed (typically at |
2465 | | * dataset closing). The reason is that those function pointers will typically |
2466 | | * point to methods of the OGRLayer instance. |
2467 | | * However, the ArrowSchema and ArrowArray structures filled from those |
2468 | | * callbacks can be used and must be released independently from the |
2469 | | * ArrowArrayStream or the layer. |
2470 | | * |
2471 | | * Furthermore, unless otherwise specified by a particular driver |
2472 | | * implementation, only one ArrowArrayStream can be active at a time on |
2473 | | * a given layer (that is the last active one must be explicitly released before |
2474 | | * a next one is asked). Changing filter state, ignored columns, modifying the |
2475 | | * schema or using ResetReading()/GetNextFeature() while using a |
2476 | | * ArrowArrayStream is strongly discouraged and may lead to unexpected results. |
2477 | | * As a rule of thumb, no OGRLayer methods that affect the state of a layer |
2478 | | * should be called on a layer, while an ArrowArrayStream on it is active. |
2479 | | * |
2480 | | * Starting with GDAL 3.8, the ArrowSchema::metadata field filled by the |
2481 | | * get_schema() callback may be set with the potential following items: |
2482 | | * <ul> |
2483 | | * <li>"GDAL:OGR:type": value of OGRFieldDefn::GetType(): (added in 3.11) |
2484 | | * Only used for DateTime fields when the DATETIME_AS_STRING=YES option is |
2485 | | * specified.</li> |
2486 | | * <li>"GDAL:OGR:alternative_name": value of |
2487 | | * OGRFieldDefn::GetAlternativeNameRef()</li> |
2488 | | * <li>"GDAL:OGR:comment": value of OGRFieldDefn::GetComment()</li> |
2489 | | * <li>"GDAL:OGR:default": value of OGRFieldDefn::GetDefault()</li> |
2490 | | * <li>"GDAL:OGR:subtype": value of OGRFieldDefn::GetSubType()</li> |
2491 | | * <li>"GDAL:OGR:width": value of OGRFieldDefn::GetWidth() (serialized as a |
2492 | | * string)</li> |
2493 | | * <li>"GDAL:OGR:unique": value of OGRFieldDefn::IsUnique() (serialized as |
2494 | | * "true" or "false")</li> |
2495 | | * <li>"GDAL:OGR:domain_name": value of OGRFieldDefn::GetDomainName()</li> |
2496 | | * </ul> |
2497 | | * |
2498 | | * A potential usage can be: |
2499 | | \code{.cpp} |
2500 | | struct ArrowArrayStream stream; |
2501 | | if( !poLayer->GetArrowStream(&stream, nullptr)) |
2502 | | { |
2503 | | CPLError(CE_Failure, CPLE_AppDefined, "GetArrowStream() failed\n"); |
2504 | | exit(1); |
2505 | | } |
2506 | | struct ArrowSchema schema; |
2507 | | if( stream.get_schema(&stream, &schema) == 0 ) |
2508 | | { |
2509 | | // Do something useful |
2510 | | schema.release(schema); |
2511 | | } |
2512 | | while( true ) |
2513 | | { |
2514 | | struct ArrowArray array; |
2515 | | // Look for an error (get_next() returning a non-zero code), or |
2516 | | // end of iteration (array.release == nullptr) |
2517 | | if( stream.get_next(&stream, &array) != 0 || |
2518 | | array.release == nullptr ) |
2519 | | { |
2520 | | break; |
2521 | | } |
2522 | | // Do something useful |
2523 | | array.release(&array); |
2524 | | } |
2525 | | stream.release(&stream); |
2526 | | \endcode |
2527 | | * |
2528 | | * A full example is available in the |
2529 | | * <a |
2530 | | href="https://gdal.org/tutorials/vector_api_tut.html#reading-from-ogr-using-the-arrow-c-stream-data-interface">Reading |
2531 | | From OGR using the Arrow C Stream data interface</a> tutorial. |
2532 | | * |
2533 | | * Options may be driver specific. The default implementation recognizes the |
2534 | | * following options: |
2535 | | * <ul> |
2536 | | * <li>INCLUDE_FID=YES/NO. Whether to include the FID column. Defaults to YES. |
2537 | | * </li> |
2538 | | * <li>MAX_FEATURES_IN_BATCH=integer. Maximum number of features to retrieve in |
2539 | | * a ArrowArray batch. Defaults to 65 536.</li> |
2540 | | * <li>TIMEZONE="unknown", "UTC", "(+|:)HH:MM" or any other value supported by |
2541 | | * Arrow. (GDAL >= 3.8) |
2542 | | * Override the timezone flag nominally provided by |
2543 | | * OGRFieldDefn::GetTZFlag(), and used for the Arrow field timezone |
2544 | | * declaration, with a user specified timezone. |
2545 | | * Note that datetime values in Arrow arrays are always stored in UTC, and |
2546 | | * that the time zone flag used by GDAL to convert to UTC is the one of the |
2547 | | * OGRField::Date::TZFlag member at the OGRFeature level. The conversion |
2548 | | * to UTC of a OGRField::Date is only done if both the timezone indicated by |
2549 | | * OGRField::Date::TZFlag and the one at the OGRFieldDefn level (or set by |
2550 | | * this TIMEZONE option) are not unknown.</li> |
2551 | | * <li>DATETIME_AS_STRING=YES/NO. Defaults to NO. Added in GDAL 3.11. |
2552 | | * Whether DateTime fields should be returned as a (normally ISO-8601 |
2553 | | * formatted) string by drivers. The aim is to be able to handle mixed |
2554 | | * timezones (or timezone naive values) in the same column. |
2555 | | * All drivers must honour that option, and potentially fallback to the |
2556 | | * OGRLayer generic implementation if they cannot (which is the case for the |
2557 | | * Arrow, Parquet and ADBC drivers). |
2558 | | * When DATETIME_AS_STRING=YES, the TIMEZONE option is ignored. |
2559 | | * </li> |
2560 | | * <li>GEOMETRY_METADATA_ENCODING=OGC/GEOARROW (GDAL >= 3.8). |
2561 | | * The default is OGC, which will lead to setting |
2562 | | * the Arrow geometry column metadata to ARROW:extension:name=ogc.wkb. |
2563 | | * If setting to GEOMETRY_METADATA_ENCODING to GEOARROW, |
2564 | | * ARROW:extension:name=geoarrow.wkb and |
2565 | | * ARROW:extension:metadata={"crs": <projjson CRS representation>> are set. |
2566 | | * </li> |
2567 | | * </ul> |
2568 | | * |
2569 | | * The Arrow/Parquet drivers recognize the following option: |
2570 | | * <ul> |
2571 | | * <li>GEOMETRY_ENCODING=WKB. To force a fallback to the generic implementation |
2572 | | * when the native geometry encoding is not WKB. Otherwise the geometry |
2573 | | * will be returned with its native Arrow encoding |
2574 | | * (possibly using GeoArrow encoding).</li> |
2575 | | * </ul> |
2576 | | * |
2577 | | * @param out_stream Output stream. Must *not* be NULL. The content of the |
2578 | | * structure does not need to be initialized. |
2579 | | * @param papszOptions NULL terminated list of key=value options. |
2580 | | * @return true in case of success. |
2581 | | * @since GDAL 3.6 |
2582 | | */ |
2583 | | bool OGRLayer::GetArrowStream(struct ArrowArrayStream *out_stream, |
2584 | | CSLConstList papszOptions) |
2585 | 0 | { |
2586 | 0 | memset(out_stream, 0, sizeof(*out_stream)); |
2587 | 0 | if (m_poSharedArrowArrayStreamPrivateData && |
2588 | 0 | m_poSharedArrowArrayStreamPrivateData->m_bArrowArrayStreamInProgress) |
2589 | 0 | { |
2590 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
2591 | 0 | "An arrow Arrow Stream is in progress on that layer. Only " |
2592 | 0 | "one at a time is allowed in this implementation."); |
2593 | 0 | return false; |
2594 | 0 | } |
2595 | 0 | m_aosArrowArrayStreamOptions.Assign(CSLDuplicate(papszOptions), true); |
2596 | |
|
2597 | 0 | out_stream->get_schema = OGRLayer::StaticGetArrowSchema; |
2598 | 0 | out_stream->get_next = OGRLayer::StaticGetNextArrowArray; |
2599 | 0 | out_stream->get_last_error = OGRLayer::GetLastErrorArrowArrayStream; |
2600 | 0 | out_stream->release = OGRLayer::ReleaseStream; |
2601 | |
|
2602 | 0 | if (m_poSharedArrowArrayStreamPrivateData == nullptr) |
2603 | 0 | { |
2604 | 0 | m_poSharedArrowArrayStreamPrivateData = |
2605 | 0 | std::make_shared<ArrowArrayStreamPrivateData>(); |
2606 | 0 | m_poSharedArrowArrayStreamPrivateData->m_poLayer = this; |
2607 | 0 | } |
2608 | 0 | m_poSharedArrowArrayStreamPrivateData->m_bArrowArrayStreamInProgress = true; |
2609 | | |
2610 | | // Special case for "FID = constant", or "FID IN (constant1, ...., constantN)" |
2611 | 0 | m_poSharedArrowArrayStreamPrivateData->m_anQueriedFIDs.clear(); |
2612 | 0 | m_poSharedArrowArrayStreamPrivateData->m_iQueriedFIDS = 0; |
2613 | 0 | if (m_poAttrQuery) |
2614 | 0 | { |
2615 | 0 | swq_expr_node *poNode = |
2616 | 0 | static_cast<swq_expr_node *>(m_poAttrQuery->GetSWQExpr()); |
2617 | 0 | if (poNode->eNodeType == SNT_OPERATION && |
2618 | 0 | (poNode->nOperation == SWQ_IN || poNode->nOperation == SWQ_EQ) && |
2619 | 0 | poNode->papoSubExpr[0]->eNodeType == SNT_COLUMN && |
2620 | 0 | poNode->papoSubExpr[0]->field_index == |
2621 | 0 | GetLayerDefn()->GetFieldCount() + SPF_FID && |
2622 | 0 | TestCapability(OLCRandomRead)) |
2623 | 0 | { |
2624 | 0 | std::set<GIntBig> oSetAlreadyListed; |
2625 | 0 | for (int i = 1; i < poNode->nSubExprCount; ++i) |
2626 | 0 | { |
2627 | 0 | if (poNode->papoSubExpr[i]->eNodeType == SNT_CONSTANT && |
2628 | 0 | poNode->papoSubExpr[i]->field_type == SWQ_INTEGER64 && |
2629 | 0 | oSetAlreadyListed.find(poNode->papoSubExpr[i]->int_value) == |
2630 | 0 | oSetAlreadyListed.end()) |
2631 | 0 | { |
2632 | 0 | oSetAlreadyListed.insert(poNode->papoSubExpr[i]->int_value); |
2633 | 0 | m_poSharedArrowArrayStreamPrivateData->m_anQueriedFIDs |
2634 | 0 | .push_back(poNode->papoSubExpr[i]->int_value); |
2635 | 0 | } |
2636 | 0 | } |
2637 | 0 | } |
2638 | 0 | } |
2639 | |
|
2640 | 0 | auto poPrivateData = new ArrowArrayStreamPrivateDataSharedDataWrapper(); |
2641 | 0 | poPrivateData->poShared = m_poSharedArrowArrayStreamPrivateData; |
2642 | 0 | out_stream->private_data = poPrivateData; |
2643 | 0 | return true; |
2644 | 0 | } |
2645 | | |
2646 | | /************************************************************************/ |
2647 | | /* OGR_L_GetArrowStream() */ |
2648 | | /************************************************************************/ |
2649 | | |
2650 | | /** Get a Arrow C stream. |
2651 | | * |
2652 | | * On successful return, and when the stream interfaces is no longer needed, it |
2653 | | * must be freed with out_stream->release(out_stream). Please carefully read |
2654 | | * https://arrow.apache.org/docs/format/CStreamInterface.html for more details |
2655 | | * on using Arrow C stream. |
2656 | | * |
2657 | | * The method may take into account ignored fields set with SetIgnoredFields() |
2658 | | * (the default implementation does), and should take into account filters set |
2659 | | * with SetSpatialFilter() and SetAttributeFilter(). Note however that |
2660 | | * specialized implementations may fallback to the default (slower) |
2661 | | * implementation when filters are set. |
2662 | | * Drivers that have a specialized implementation should |
2663 | | * advertise the OLCFastGetArrowStream capability. |
2664 | | * |
2665 | | * There are extra precautions to take into account in a OGR context. Unless |
2666 | | * otherwise specified by a particular driver implementation, the get_schema(), |
2667 | | * get_next() and get_last_error() function pointers of the ArrowArrayStream |
2668 | | * structure should no longer be used after the OGRLayer, from which the |
2669 | | * ArrowArrayStream structure was initialized, has been destroyed (typically at |
2670 | | * dataset closing). The reason is that those function pointers will typically |
2671 | | * point to methods of the OGRLayer instance. |
2672 | | * However, the ArrowSchema and ArrowArray structures filled from those |
2673 | | * callbacks can be used and must be released independently from the |
2674 | | * ArrowArrayStream or the layer. |
2675 | | * |
2676 | | * Furthermore, unless otherwise specified by a particular driver |
2677 | | * implementation, only one ArrowArrayStream can be active at a time on |
2678 | | * a given layer (that is the last active one must be explicitly released before |
2679 | | * a next one is asked). Changing filter state, ignored columns, modifying the |
2680 | | * schema or using ResetReading()/GetNextFeature() while using a |
2681 | | * ArrowArrayStream is strongly discouraged and may lead to unexpected results. |
2682 | | * As a rule of thumb, no OGRLayer methods that affect the state of a layer |
2683 | | * should be called on a layer, while an ArrowArrayStream on it is active. |
2684 | | * |
2685 | | * Starting with GDAL 3.8, the ArrowSchema::metadata field filled by the |
2686 | | * get_schema() callback may be set with the potential following items: |
2687 | | * <ul> |
2688 | | * <li>"GDAL:OGR:type": value of OGRFieldDefn::GetType(): (added in 3.11) |
2689 | | * Only used for DateTime fields when the DATETIME_AS_STRING=YES option is |
2690 | | * specified.</li> |
2691 | | * <li>"GDAL:OGR:alternative_name": value of |
2692 | | * OGRFieldDefn::GetAlternativeNameRef()</li> |
2693 | | * <li>"GDAL:OGR:comment": value of OGRFieldDefn::GetComment()</li> |
2694 | | * <li>"GDAL:OGR:default": value of OGRFieldDefn::GetDefault()</li> |
2695 | | * <li>"GDAL:OGR:subtype": value of OGRFieldDefn::GetSubType()</li> |
2696 | | * <li>"GDAL:OGR:width": value of OGRFieldDefn::GetWidth() (serialized as a |
2697 | | * string)</li> |
2698 | | * <li>"GDAL:OGR:unique": value of OGRFieldDefn::IsUnique() (serialized as |
2699 | | * "true" or "false")</li> |
2700 | | * <li>"GDAL:OGR:domain_name": value of OGRFieldDefn::GetDomainName()</li> |
2701 | | * </ul> |
2702 | | * |
2703 | | * A potential usage can be: |
2704 | | \code{.cpp} |
2705 | | struct ArrowArrayStream stream; |
2706 | | if( !OGR_L_GetArrowStream(hLayer, &stream, nullptr)) |
2707 | | { |
2708 | | CPLError(CE_Failure, CPLE_AppDefined, |
2709 | | "OGR_L_GetArrowStream() failed\n"); |
2710 | | exit(1); |
2711 | | } |
2712 | | struct ArrowSchema schema; |
2713 | | if( stream.get_schema(&stream, &schema) == 0 ) |
2714 | | { |
2715 | | // Do something useful |
2716 | | schema.release(schema); |
2717 | | } |
2718 | | while( true ) |
2719 | | { |
2720 | | struct ArrowArray array; |
2721 | | // Look for an error (get_next() returning a non-zero code), or |
2722 | | // end of iteration (array.release == nullptr) |
2723 | | if( stream.get_next(&stream, &array) != 0 || |
2724 | | array.release == nullptr ) |
2725 | | { |
2726 | | break; |
2727 | | } |
2728 | | // Do something useful |
2729 | | array.release(&array); |
2730 | | } |
2731 | | stream.release(&stream); |
2732 | | \endcode |
2733 | | * |
2734 | | * A full example is available in the |
2735 | | * <a |
2736 | | href="https://gdal.org/tutorials/vector_api_tut.html#reading-from-ogr-using-the-arrow-c-stream-data-interface">Reading |
2737 | | From OGR using the Arrow C Stream data interface</a> tutorial. |
2738 | | * |
2739 | | * Options may be driver specific. The default implementation recognizes the |
2740 | | * following options: |
2741 | | * <ul> |
2742 | | * <li>INCLUDE_FID=YES/NO. Whether to include the FID column. Defaults to |
2743 | | YES.</li> |
2744 | | * <li>MAX_FEATURES_IN_BATCH=integer. Maximum number of features to retrieve in |
2745 | | * a ArrowArray batch. Defaults to 65 536.</li> |
2746 | | * <li>TIMEZONE="unknown", "UTC", "(+|:)HH:MM" or any other value supported by |
2747 | | * Arrow. (GDAL >= 3.8) |
2748 | | * Override the timezone flag nominally provided by |
2749 | | * OGRFieldDefn::GetTZFlag(), and used for the Arrow field timezone |
2750 | | * declaration, with a user specified timezone. |
2751 | | * Note that datetime values in Arrow arrays are always stored in UTC, and |
2752 | | * that the time zone flag used by GDAL to convert to UTC is the one of the |
2753 | | * OGRField::Date::TZFlag member at the OGRFeature level. The conversion |
2754 | | * to UTC of a OGRField::Date is only done if both the timezone indicated by |
2755 | | * OGRField::Date::TZFlag and the one at the OGRFieldDefn level (or set by |
2756 | | * this TIMEZONE option) are not unknown.</li> |
2757 | | * <li>DATETIME_AS_STRING=YES/NO. Defaults to NO. Added in GDAL 3.11. |
2758 | | * Whether DateTime fields should be returned as a (normally ISO-8601 |
2759 | | * formatted) string by drivers. The aim is to be able to handle mixed |
2760 | | * timezones (or timezone naive values) in the same column. |
2761 | | * All drivers must honour that option, and potentially fallback to the |
2762 | | * OGRLayer generic implementation if they cannot (which is the case for the |
2763 | | * Arrow, Parquet and ADBC drivers). |
2764 | | * When DATETIME_AS_STRING=YES, the TIMEZONE option is ignored. |
2765 | | * </li> |
2766 | | * <li>GEOMETRY_METADATA_ENCODING=OGC/GEOARROW (GDAL >= 3.8). |
2767 | | * The default is OGC, which will lead to setting |
2768 | | * the Arrow geometry column metadata to ARROW:extension:name=ogc.wkb. |
2769 | | * If setting to GEOMETRY_METADATA_ENCODING to GEOARROW, |
2770 | | * ARROW:extension:name=geoarrow.wkb and |
2771 | | * ARROW:extension:metadata={"crs": <projjson CRS representation>> are set. |
2772 | | * </li> |
2773 | | * </ul> |
2774 | | * |
2775 | | * The Arrow/Parquet drivers recognize the following option: |
2776 | | * <ul> |
2777 | | * <li>GEOMETRY_ENCODING=WKB. To force a fallback to the generic implementation |
2778 | | * when the native geometry encoding is not WKB. Otherwise the geometry |
2779 | | * will be returned with its native Arrow encoding |
2780 | | * (possibly using GeoArrow encoding).</li> |
2781 | | * </ul> |
2782 | | * |
2783 | | * @param hLayer Layer |
2784 | | * @param out_stream Output stream. Must *not* be NULL. The content of the |
2785 | | * structure does not need to be initialized. |
2786 | | * @param papszOptions NULL terminated list of key=value options. |
2787 | | * @return true in case of success. |
2788 | | * @since GDAL 3.6 |
2789 | | */ |
2790 | | bool OGR_L_GetArrowStream(OGRLayerH hLayer, struct ArrowArrayStream *out_stream, |
2791 | | char **papszOptions) |
2792 | 0 | { |
2793 | 0 | VALIDATE_POINTER1(hLayer, "OGR_L_GetArrowStream", false); |
2794 | 0 | VALIDATE_POINTER1(out_stream, "OGR_L_GetArrowStream", false); |
2795 | | |
2796 | 0 | return OGRLayer::FromHandle(hLayer)->GetArrowStream(out_stream, |
2797 | 0 | papszOptions); |
2798 | 0 | } |
2799 | | |
2800 | | /************************************************************************/ |
2801 | | /* OGRParseArrowMetadata() */ |
2802 | | /************************************************************************/ |
2803 | | |
2804 | | std::map<std::string, std::string> |
2805 | | OGRParseArrowMetadata(const char *pabyMetadata) |
2806 | 0 | { |
2807 | 0 | std::map<std::string, std::string> oMetadata; |
2808 | 0 | int32_t nKVP; |
2809 | 0 | memcpy(&nKVP, pabyMetadata, sizeof(int32_t)); |
2810 | 0 | pabyMetadata += sizeof(int32_t); |
2811 | 0 | for (int i = 0; i < nKVP; ++i) |
2812 | 0 | { |
2813 | 0 | int32_t nSizeKey; |
2814 | 0 | memcpy(&nSizeKey, pabyMetadata, sizeof(int32_t)); |
2815 | 0 | pabyMetadata += sizeof(int32_t); |
2816 | 0 | std::string osKey; |
2817 | 0 | osKey.assign(pabyMetadata, nSizeKey); |
2818 | 0 | pabyMetadata += nSizeKey; |
2819 | |
|
2820 | 0 | int32_t nSizeValue; |
2821 | 0 | memcpy(&nSizeValue, pabyMetadata, sizeof(int32_t)); |
2822 | 0 | pabyMetadata += sizeof(int32_t); |
2823 | 0 | std::string osValue; |
2824 | 0 | osValue.assign(pabyMetadata, nSizeValue); |
2825 | 0 | pabyMetadata += nSizeValue; |
2826 | |
|
2827 | 0 | oMetadata[osKey] = std::move(osValue); |
2828 | 0 | } |
2829 | |
|
2830 | 0 | return oMetadata; |
2831 | 0 | } |
2832 | | |
2833 | | /************************************************************************/ |
2834 | | /* ParseDecimalFormat() */ |
2835 | | /************************************************************************/ |
2836 | | |
2837 | | static bool ParseDecimalFormat(const char *format, int &nPrecision, int &nScale, |
2838 | | int &nWidthInBytes) |
2839 | 0 | { |
2840 | | // d:19,10 ==> decimal128 [precision 19, scale 10] |
2841 | | // d:19,10,NNN ==> decimal bitwidth = NNN [precision 19, scale 10] |
2842 | 0 | nPrecision = 0; |
2843 | 0 | nScale = 0; |
2844 | 0 | nWidthInBytes = 128 / 8; // 128 bit |
2845 | 0 | const char *pszFirstComma = strchr(format + 2, ','); |
2846 | 0 | if (pszFirstComma) |
2847 | 0 | { |
2848 | 0 | nPrecision = atoi(format + 2); |
2849 | 0 | nScale = atoi(pszFirstComma + 1); |
2850 | 0 | const char *pszSecondComma = strchr(pszFirstComma + 1, ','); |
2851 | 0 | if (pszSecondComma) |
2852 | 0 | { |
2853 | 0 | const int nWidthInBits = atoi(pszSecondComma + 1); |
2854 | 0 | if ((nWidthInBits % 8) != 0) |
2855 | 0 | { |
2856 | | // shouldn't happen for well-format schemas |
2857 | 0 | nWidthInBytes = 0; |
2858 | 0 | return false; |
2859 | 0 | } |
2860 | 0 | else |
2861 | 0 | { |
2862 | 0 | nWidthInBytes = nWidthInBits / 8; |
2863 | 0 | } |
2864 | 0 | } |
2865 | 0 | } |
2866 | 0 | else |
2867 | 0 | { |
2868 | | // shouldn't happen for well-format schemas |
2869 | 0 | nWidthInBytes = 0; |
2870 | 0 | return false; |
2871 | 0 | } |
2872 | 0 | return true; |
2873 | 0 | } |
2874 | | |
2875 | | /************************************************************************/ |
2876 | | /* GetErrorIfUnsupportedDecimal() */ |
2877 | | /************************************************************************/ |
2878 | | |
2879 | | static const char *GetErrorIfUnsupportedDecimal(int nWidthInBytes, |
2880 | | int nPrecision) |
2881 | 0 | { |
2882 | |
|
2883 | 0 | if (nWidthInBytes != 128 / 8 && nWidthInBytes != 256 / 8) |
2884 | 0 | { |
2885 | 0 | return "For decimal field, only width 128 and 256 are supported"; |
2886 | 0 | } |
2887 | | |
2888 | | // precision=19 fits on 64 bits |
2889 | 0 | if (nPrecision <= 0 || nPrecision > 19) |
2890 | 0 | { |
2891 | 0 | return "For decimal field, only precision up to 19 is supported"; |
2892 | 0 | } |
2893 | | |
2894 | 0 | return nullptr; |
2895 | 0 | } |
2896 | | |
2897 | | /************************************************************************/ |
2898 | | /* IsHandledSchema() */ |
2899 | | /************************************************************************/ |
2900 | | |
2901 | | static bool IsHandledSchema(bool bTopLevel, const struct ArrowSchema *schema, |
2902 | | const std::string &osPrefix, bool bHasAttrQuery, |
2903 | | const CPLStringList &aosUsedFields) |
2904 | 0 | { |
2905 | 0 | const char *format = schema->format; |
2906 | 0 | if (IsStructure(format)) |
2907 | 0 | { |
2908 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
2909 | 0 | { |
2910 | 0 | if (!IsHandledSchema(/* bTopLevel = */ false, |
2911 | 0 | schema->children[static_cast<size_t>(i)], |
2912 | 0 | bTopLevel ? std::string() |
2913 | 0 | : osPrefix + schema->name + ".", |
2914 | 0 | bHasAttrQuery, aosUsedFields)) |
2915 | 0 | { |
2916 | 0 | return false; |
2917 | 0 | } |
2918 | 0 | } |
2919 | 0 | return true; |
2920 | 0 | } |
2921 | | |
2922 | | // Lists or maps |
2923 | 0 | if (IsList(format) || IsLargeList(format) || IsFixedSizeList(format) || |
2924 | 0 | IsMap(format)) |
2925 | 0 | { |
2926 | 0 | if (!IsHandledSchema(/* bTopLevel = */ false, schema->children[0], |
2927 | 0 | osPrefix, bHasAttrQuery, aosUsedFields)) |
2928 | 0 | { |
2929 | 0 | return false; |
2930 | 0 | } |
2931 | | // For now, we can't filter on lists or maps |
2932 | 0 | if (aosUsedFields.FindString((osPrefix + schema->name).c_str()) >= 0) |
2933 | 0 | { |
2934 | 0 | CPLDebug("OGR", |
2935 | 0 | "Field %s has unhandled format '%s' for an " |
2936 | 0 | "attribute to filter on", |
2937 | 0 | (osPrefix + schema->name).c_str(), format); |
2938 | 0 | return false; |
2939 | 0 | } |
2940 | 0 | return true; |
2941 | 0 | } |
2942 | | |
2943 | 0 | const char *const apszHandledFormats[] = { |
2944 | 0 | "b", // boolean |
2945 | 0 | "c", // int8 |
2946 | 0 | "C", // uint8 |
2947 | 0 | "s", // int16 |
2948 | 0 | "S", // uint16 |
2949 | 0 | "i", // int32 |
2950 | 0 | "I", // uint32 |
2951 | 0 | "l", // int64 |
2952 | 0 | "L", // uint64 |
2953 | 0 | "e", // float16 |
2954 | 0 | "f", // float32 |
2955 | 0 | "g", // float64, |
2956 | 0 | "z", // binary |
2957 | 0 | "Z", // large binary |
2958 | 0 | "u", // UTF-8 string |
2959 | 0 | "U", // large UTF-8 string |
2960 | 0 | "tdD", // date32[days] |
2961 | 0 | "tdm", // date64[milliseconds] |
2962 | 0 | "tts", //time32 [seconds] |
2963 | 0 | "ttm", //time32 [milliseconds] |
2964 | 0 | "ttu", //time64 [microseconds] |
2965 | 0 | "ttn", //time64 [nanoseconds] |
2966 | 0 | }; |
2967 | |
|
2968 | 0 | for (const char *pszHandledFormat : apszHandledFormats) |
2969 | 0 | { |
2970 | 0 | if (strcmp(format, pszHandledFormat) == 0) |
2971 | 0 | { |
2972 | 0 | return true; |
2973 | 0 | } |
2974 | 0 | } |
2975 | | |
2976 | 0 | if (IsDecimal(format)) |
2977 | 0 | { |
2978 | 0 | if (bHasAttrQuery && |
2979 | 0 | aosUsedFields.FindString((osPrefix + schema->name).c_str()) >= 0) |
2980 | 0 | { |
2981 | 0 | int nPrecision = 0; |
2982 | 0 | int nScale = 0; |
2983 | 0 | int nWidthInBytes = 0; |
2984 | 0 | if (!ParseDecimalFormat(format, nPrecision, nScale, nWidthInBytes)) |
2985 | 0 | { |
2986 | 0 | CPLDebug("OGR", "%s", |
2987 | 0 | (std::string("Invalid field format ") + format + |
2988 | 0 | " for field " + osPrefix + schema->name) |
2989 | 0 | .c_str()); |
2990 | 0 | return false; |
2991 | 0 | } |
2992 | | |
2993 | 0 | const char *pszError = |
2994 | 0 | GetErrorIfUnsupportedDecimal(nWidthInBytes, nPrecision); |
2995 | 0 | if (pszError) |
2996 | 0 | { |
2997 | 0 | CPLDebug("OGR", "%s", pszError); |
2998 | 0 | return false; |
2999 | 0 | } |
3000 | 0 | } |
3001 | 0 | return true; |
3002 | 0 | } |
3003 | | |
3004 | 0 | if (IsFixedWidthBinary(format) || IsTimestamp(format)) |
3005 | 0 | { |
3006 | 0 | return true; |
3007 | 0 | } |
3008 | | |
3009 | 0 | CPLDebug("OGR", "Field %s has unhandled format '%s'", |
3010 | 0 | (osPrefix + schema->name).c_str(), format); |
3011 | 0 | return false; |
3012 | 0 | } |
3013 | | |
3014 | | /************************************************************************/ |
3015 | | /* OGRLayer::CanPostFilterArrowArray() */ |
3016 | | /************************************************************************/ |
3017 | | |
3018 | | /** Whether the PostFilterArrowArray() can work on the schema to remove |
3019 | | * rows that aren't selected by the spatial or attribute filter. |
3020 | | */ |
3021 | | bool OGRLayer::CanPostFilterArrowArray(const struct ArrowSchema *schema) const |
3022 | 0 | { |
3023 | 0 | if (!IsHandledSchema( |
3024 | 0 | /* bTopLevel=*/true, schema, std::string(), |
3025 | 0 | m_poAttrQuery != nullptr, |
3026 | 0 | m_poAttrQuery ? CPLStringList(m_poAttrQuery->GetUsedFields()) |
3027 | 0 | : CPLStringList())) |
3028 | 0 | { |
3029 | 0 | return false; |
3030 | 0 | } |
3031 | | |
3032 | 0 | if (m_poFilterGeom) |
3033 | 0 | { |
3034 | 0 | bool bFound = false; |
3035 | 0 | const char *pszGeomFieldName = |
3036 | 0 | const_cast<OGRLayer *>(this) |
3037 | 0 | ->GetLayerDefn() |
3038 | 0 | ->GetGeomFieldDefn(m_iGeomFieldFilter) |
3039 | 0 | ->GetNameRef(); |
3040 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
3041 | 0 | { |
3042 | 0 | const auto fieldSchema = schema->children[i]; |
3043 | 0 | if (strcmp(fieldSchema->name, pszGeomFieldName) == 0) |
3044 | 0 | { |
3045 | 0 | if (!IsBinary(fieldSchema->format) && |
3046 | 0 | !IsLargeBinary(fieldSchema->format)) |
3047 | 0 | { |
3048 | 0 | CPLDebug("OGR", "Geometry field %s has handled format '%s'", |
3049 | 0 | fieldSchema->name, fieldSchema->format); |
3050 | 0 | return false; |
3051 | 0 | } |
3052 | | |
3053 | | // Check if ARROW:extension:name = ogc.wkb |
3054 | 0 | const char *pabyMetadata = fieldSchema->metadata; |
3055 | 0 | if (!pabyMetadata) |
3056 | 0 | { |
3057 | 0 | CPLDebug( |
3058 | 0 | "OGR", |
3059 | 0 | "Geometry field %s lacks metadata in its schema field", |
3060 | 0 | fieldSchema->name); |
3061 | 0 | return false; |
3062 | 0 | } |
3063 | | |
3064 | 0 | const auto oMetadata = OGRParseArrowMetadata(pabyMetadata); |
3065 | 0 | auto oIter = oMetadata.find(ARROW_EXTENSION_NAME_KEY); |
3066 | 0 | if (oIter == oMetadata.end()) |
3067 | 0 | { |
3068 | 0 | CPLDebug("OGR", |
3069 | 0 | "Geometry field %s lacks " |
3070 | 0 | "%s metadata " |
3071 | 0 | "in its schema field", |
3072 | 0 | fieldSchema->name, ARROW_EXTENSION_NAME_KEY); |
3073 | 0 | return false; |
3074 | 0 | } |
3075 | 0 | if (oIter->second != EXTENSION_NAME_OGC_WKB && |
3076 | 0 | oIter->second != EXTENSION_NAME_GEOARROW_WKB) |
3077 | 0 | { |
3078 | 0 | CPLDebug("OGR", |
3079 | 0 | "Geometry field %s has unexpected " |
3080 | 0 | "%s = '%s' metadata " |
3081 | 0 | "in its schema field", |
3082 | 0 | fieldSchema->name, ARROW_EXTENSION_NAME_KEY, |
3083 | 0 | oIter->second.c_str()); |
3084 | 0 | return false; |
3085 | 0 | } |
3086 | | |
3087 | 0 | bFound = true; |
3088 | 0 | break; |
3089 | 0 | } |
3090 | 0 | } |
3091 | 0 | if (!bFound) |
3092 | 0 | { |
3093 | 0 | CPLDebug("OGR", "Cannot find geometry field %s in schema", |
3094 | 0 | pszGeomFieldName); |
3095 | 0 | return false; |
3096 | 0 | } |
3097 | 0 | } |
3098 | | |
3099 | 0 | return true; |
3100 | 0 | } |
3101 | | |
3102 | | #if 0 |
3103 | | /************************************************************************/ |
3104 | | /* CheckValidityBuffer() */ |
3105 | | /************************************************************************/ |
3106 | | |
3107 | | static void CheckValidityBuffer(const struct ArrowArray *array) |
3108 | | { |
3109 | | if (array->null_count < 0) |
3110 | | return; |
3111 | | const uint8_t *pabyValidity = |
3112 | | static_cast<const uint8_t *>(const_cast<const void *>(array->buffers[0])); |
3113 | | if( !pabyValidity ) |
3114 | | { |
3115 | | CPLAssert(array->null_count == 0); |
3116 | | return; |
3117 | | } |
3118 | | size_t null_count = 0; |
3119 | | const size_t nOffset = static_cast<size_t>(array->offset); |
3120 | | for(size_t i = 0; i < static_cast<size_t>(array->length); ++i ) |
3121 | | { |
3122 | | if (!TestBit(pabyValidity, i + nOffset)) |
3123 | | ++ null_count; |
3124 | | } |
3125 | | CPLAssert(static_cast<size_t>(array->null_count) == null_count); |
3126 | | } |
3127 | | #endif |
3128 | | |
3129 | | /************************************************************************/ |
3130 | | /* CompactValidityBuffer() */ |
3131 | | /************************************************************************/ |
3132 | | |
3133 | | static void CompactValidityBuffer( |
3134 | | const struct ArrowSchema *, struct ArrowArray *array, size_t iStart, |
3135 | | const std::vector<bool> &abyValidityFromFilters, size_t nNewLength) |
3136 | 0 | { |
3137 | | // Invalidate null_count as the same validity buffer may be used when |
3138 | | // scrolling batches, and this creates confusion if we try to set it |
3139 | | // to different values among the batches |
3140 | 0 | if (array->null_count <= 0) |
3141 | 0 | { |
3142 | 0 | array->null_count = -1; |
3143 | 0 | return; |
3144 | 0 | } |
3145 | 0 | array->null_count = -1; |
3146 | |
|
3147 | 0 | CPLAssert(static_cast<size_t>(array->length) >= |
3148 | 0 | iStart + abyValidityFromFilters.size()); |
3149 | 0 | uint8_t *pabyValidity = |
3150 | 0 | static_cast<uint8_t *>(const_cast<void *>(array->buffers[0])); |
3151 | 0 | const size_t nLength = abyValidityFromFilters.size(); |
3152 | 0 | const size_t nOffset = static_cast<size_t>(array->offset); |
3153 | 0 | size_t j = iStart + nOffset; |
3154 | 0 | for (size_t i = 0; i < nLength && j < nNewLength + nOffset; ++i) |
3155 | 0 | { |
3156 | 0 | if (abyValidityFromFilters[i]) |
3157 | 0 | { |
3158 | 0 | if (TestBit(pabyValidity, i + iStart + nOffset)) |
3159 | 0 | SetBit(pabyValidity, j); |
3160 | 0 | else |
3161 | 0 | UnsetBit(pabyValidity, j); |
3162 | 0 | ++j; |
3163 | 0 | } |
3164 | 0 | } |
3165 | 0 | } |
3166 | | |
3167 | | /************************************************************************/ |
3168 | | /* CompactBoolArray() */ |
3169 | | /************************************************************************/ |
3170 | | |
3171 | | static void CompactBoolArray(const struct ArrowSchema *schema, |
3172 | | struct ArrowArray *array, size_t iStart, |
3173 | | const std::vector<bool> &abyValidityFromFilters, |
3174 | | size_t nNewLength) |
3175 | 0 | { |
3176 | 0 | CPLAssert(array->n_children == 0); |
3177 | 0 | CPLAssert(array->n_buffers == 2); |
3178 | 0 | CPLAssert(static_cast<size_t>(array->length) >= |
3179 | 0 | iStart + abyValidityFromFilters.size()); |
3180 | | |
3181 | 0 | const size_t nLength = abyValidityFromFilters.size(); |
3182 | 0 | const size_t nOffset = static_cast<size_t>(array->offset); |
3183 | 0 | uint8_t *pabyData = |
3184 | 0 | static_cast<uint8_t *>(const_cast<void *>(array->buffers[1])); |
3185 | 0 | size_t j = iStart + nOffset; |
3186 | 0 | for (size_t i = 0; i < nLength; ++i) |
3187 | 0 | { |
3188 | 0 | if (abyValidityFromFilters[i]) |
3189 | 0 | { |
3190 | 0 | if (TestBit(pabyData, i + iStart + nOffset)) |
3191 | 0 | SetBit(pabyData, j); |
3192 | 0 | else |
3193 | 0 | UnsetBit(pabyData, j); |
3194 | |
|
3195 | 0 | ++j; |
3196 | 0 | } |
3197 | 0 | } |
3198 | |
|
3199 | 0 | if (schema->flags & ARROW_FLAG_NULLABLE) |
3200 | 0 | CompactValidityBuffer(schema, array, iStart, abyValidityFromFilters, |
3201 | 0 | nNewLength); |
3202 | |
|
3203 | 0 | array->length = nNewLength; |
3204 | 0 | } |
3205 | | |
3206 | | /************************************************************************/ |
3207 | | /* CompactPrimitiveArray() */ |
3208 | | /************************************************************************/ |
3209 | | |
3210 | | template <class T> |
3211 | | static void CompactPrimitiveArray( |
3212 | | const struct ArrowSchema *schema, struct ArrowArray *array, size_t iStart, |
3213 | | const std::vector<bool> &abyValidityFromFilters, size_t nNewLength) |
3214 | 0 | { |
3215 | 0 | CPLAssert(array->n_children == 0); |
3216 | 0 | CPLAssert(array->n_buffers == 2); |
3217 | 0 | CPLAssert(static_cast<size_t>(array->length) >= |
3218 | 0 | iStart + abyValidityFromFilters.size()); |
3219 | | |
3220 | 0 | const size_t nLength = abyValidityFromFilters.size(); |
3221 | 0 | const size_t nOffset = static_cast<size_t>(array->offset); |
3222 | 0 | T *paData = |
3223 | 0 | static_cast<T *>(const_cast<void *>(array->buffers[1])) + nOffset; |
3224 | 0 | size_t j = iStart; |
3225 | 0 | for (size_t i = 0; i < nLength; ++i) |
3226 | 0 | { |
3227 | 0 | if (abyValidityFromFilters[i]) |
3228 | 0 | { |
3229 | 0 | paData[j] = paData[i + iStart]; |
3230 | 0 | ++j; |
3231 | 0 | } |
3232 | 0 | } |
3233 | |
|
3234 | 0 | if (schema->flags & ARROW_FLAG_NULLABLE) |
3235 | 0 | CompactValidityBuffer(schema, array, iStart, abyValidityFromFilters, |
3236 | 0 | nNewLength); |
3237 | |
|
3238 | 0 | array->length = nNewLength; |
3239 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void CompactPrimitiveArray<unsigned char>(ArrowSchema const*, ArrowArray*, unsigned long, std::__1::vector<bool, std::__1::allocator<bool> > const&, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:void CompactPrimitiveArray<unsigned short>(ArrowSchema const*, ArrowArray*, unsigned long, std::__1::vector<bool, std::__1::allocator<bool> > const&, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:void CompactPrimitiveArray<unsigned int>(ArrowSchema const*, ArrowArray*, unsigned long, std::__1::vector<bool, std::__1::allocator<bool> > const&, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:void CompactPrimitiveArray<unsigned long>(ArrowSchema const*, ArrowArray*, unsigned long, std::__1::vector<bool, std::__1::allocator<bool> > const&, unsigned long) |
3240 | | |
3241 | | /************************************************************************/ |
3242 | | /* CompactStringOrBinaryArray() */ |
3243 | | /************************************************************************/ |
3244 | | |
3245 | | template <class OffsetType> |
3246 | | static void CompactStringOrBinaryArray( |
3247 | | const struct ArrowSchema *schema, struct ArrowArray *array, size_t iStart, |
3248 | | const std::vector<bool> &abyValidityFromFilters, size_t nNewLength) |
3249 | 0 | { |
3250 | 0 | CPLAssert(array->n_children == 0); |
3251 | 0 | CPLAssert(array->n_buffers == 3); |
3252 | 0 | CPLAssert(static_cast<size_t>(array->length) >= |
3253 | 0 | iStart + abyValidityFromFilters.size()); |
3254 | | |
3255 | 0 | const size_t nLength = abyValidityFromFilters.size(); |
3256 | 0 | const size_t nOffset = static_cast<size_t>(array->offset); |
3257 | 0 | OffsetType *panOffsets = |
3258 | 0 | static_cast<OffsetType *>(const_cast<void *>(array->buffers[1])) + |
3259 | 0 | nOffset; |
3260 | 0 | GByte *pabyData = |
3261 | 0 | static_cast<GByte *>(const_cast<void *>(array->buffers[2])); |
3262 | 0 | size_t j = iStart; |
3263 | 0 | OffsetType nCurOffset = panOffsets[iStart]; |
3264 | 0 | for (size_t i = 0; i < nLength; ++i) |
3265 | 0 | { |
3266 | 0 | if (abyValidityFromFilters[i]) |
3267 | 0 | { |
3268 | 0 | const auto nStartOffset = panOffsets[i + iStart]; |
3269 | 0 | const auto nEndOffset = panOffsets[i + iStart + 1]; |
3270 | 0 | panOffsets[j] = nCurOffset; |
3271 | 0 | const auto nSize = static_cast<size_t>(nEndOffset - nStartOffset); |
3272 | 0 | if (nSize) |
3273 | 0 | { |
3274 | 0 | if (nCurOffset < nStartOffset) |
3275 | 0 | { |
3276 | 0 | memmove(pabyData + nCurOffset, pabyData + nStartOffset, |
3277 | 0 | nSize); |
3278 | 0 | } |
3279 | 0 | nCurOffset += static_cast<OffsetType>(nSize); |
3280 | 0 | } |
3281 | 0 | ++j; |
3282 | 0 | } |
3283 | 0 | } |
3284 | 0 | panOffsets[j] = nCurOffset; |
3285 | |
|
3286 | 0 | if (schema->flags & ARROW_FLAG_NULLABLE) |
3287 | 0 | CompactValidityBuffer(schema, array, iStart, abyValidityFromFilters, |
3288 | 0 | nNewLength); |
3289 | |
|
3290 | 0 | array->length = nNewLength; |
3291 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void CompactStringOrBinaryArray<unsigned int>(ArrowSchema const*, ArrowArray*, unsigned long, std::__1::vector<bool, std::__1::allocator<bool> > const&, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:void CompactStringOrBinaryArray<unsigned long>(ArrowSchema const*, ArrowArray*, unsigned long, std::__1::vector<bool, std::__1::allocator<bool> > const&, unsigned long) |
3292 | | |
3293 | | /************************************************************************/ |
3294 | | /* CompactFixedWidthArray() */ |
3295 | | /************************************************************************/ |
3296 | | |
3297 | | static void |
3298 | | CompactFixedWidthArray(const struct ArrowSchema *schema, |
3299 | | struct ArrowArray *array, int nWidth, size_t iStart, |
3300 | | const std::vector<bool> &abyValidityFromFilters, |
3301 | | size_t nNewLength) |
3302 | 0 | { |
3303 | 0 | CPLAssert(array->n_children == 0); |
3304 | 0 | CPLAssert(array->n_buffers == 2); |
3305 | 0 | CPLAssert(static_cast<size_t>(array->length) >= |
3306 | 0 | iStart + abyValidityFromFilters.size()); |
3307 | | |
3308 | 0 | const size_t nLength = abyValidityFromFilters.size(); |
3309 | 0 | const size_t nOffset = static_cast<size_t>(array->offset); |
3310 | 0 | GByte *pabyData = |
3311 | 0 | static_cast<GByte *>(const_cast<void *>(array->buffers[1])); |
3312 | 0 | size_t nStartOffset = (iStart + nOffset) * nWidth; |
3313 | 0 | size_t nCurOffset = nStartOffset; |
3314 | 0 | for (size_t i = 0; i < nLength; ++i, nStartOffset += nWidth) |
3315 | 0 | { |
3316 | 0 | if (abyValidityFromFilters[i]) |
3317 | 0 | { |
3318 | 0 | if (nCurOffset < nStartOffset) |
3319 | 0 | { |
3320 | 0 | memcpy(pabyData + nCurOffset, pabyData + nStartOffset, nWidth); |
3321 | 0 | } |
3322 | 0 | nCurOffset += nWidth; |
3323 | 0 | } |
3324 | 0 | } |
3325 | |
|
3326 | 0 | if (schema->flags & ARROW_FLAG_NULLABLE) |
3327 | 0 | CompactValidityBuffer(schema, array, iStart, abyValidityFromFilters, |
3328 | 0 | nNewLength); |
3329 | |
|
3330 | 0 | array->length = nNewLength; |
3331 | 0 | } |
3332 | | |
3333 | | /************************************************************************/ |
3334 | | /* CompactStructArray() */ |
3335 | | /************************************************************************/ |
3336 | | |
3337 | | static bool CompactArray(const struct ArrowSchema *schema, |
3338 | | struct ArrowArray *array, size_t iStart, |
3339 | | const std::vector<bool> &abyValidityFromFilters, |
3340 | | size_t nNewLength); |
3341 | | |
3342 | | static bool CompactStructArray(const struct ArrowSchema *schema, |
3343 | | struct ArrowArray *array, size_t iStart, |
3344 | | const std::vector<bool> &abyValidityFromFilters, |
3345 | | size_t nNewLength) |
3346 | 0 | { |
3347 | | // The equality might not be strict in the case of when some sub-arrays |
3348 | | // are fully void ! |
3349 | 0 | CPLAssert(array->n_children <= schema->n_children); |
3350 | 0 | for (int64_t iField = 0; iField < array->n_children; ++iField) |
3351 | 0 | { |
3352 | 0 | const auto psChildSchema = schema->children[iField]; |
3353 | 0 | const auto psChildArray = array->children[iField]; |
3354 | | // To please Arrow validation... |
3355 | 0 | const size_t nChildNewLength = |
3356 | 0 | static_cast<size_t>(array->offset) + nNewLength; |
3357 | 0 | if (psChildArray->length > array->length) |
3358 | 0 | { |
3359 | 0 | std::vector<bool> abyChildValidity(abyValidityFromFilters); |
3360 | 0 | abyChildValidity.resize( |
3361 | 0 | abyValidityFromFilters.size() + |
3362 | 0 | static_cast<size_t>(psChildArray->length - array->length), |
3363 | 0 | false); |
3364 | 0 | if (!CompactArray(psChildSchema, psChildArray, iStart, |
3365 | 0 | abyChildValidity, nChildNewLength)) |
3366 | 0 | { |
3367 | 0 | return false; |
3368 | 0 | } |
3369 | 0 | } |
3370 | 0 | else |
3371 | 0 | { |
3372 | 0 | if (!CompactArray(psChildSchema, psChildArray, iStart, |
3373 | 0 | abyValidityFromFilters, nChildNewLength)) |
3374 | 0 | { |
3375 | 0 | return false; |
3376 | 0 | } |
3377 | 0 | } |
3378 | 0 | CPLAssert(psChildArray->length == |
3379 | 0 | static_cast<int64_t>(nChildNewLength)); |
3380 | 0 | } |
3381 | | |
3382 | 0 | if (schema->flags & ARROW_FLAG_NULLABLE) |
3383 | 0 | CompactValidityBuffer(schema, array, iStart, abyValidityFromFilters, |
3384 | 0 | nNewLength); |
3385 | |
|
3386 | 0 | array->length = nNewLength; |
3387 | |
|
3388 | 0 | return true; |
3389 | 0 | } |
3390 | | |
3391 | | /************************************************************************/ |
3392 | | /* InvalidateNullCountRec() */ |
3393 | | /************************************************************************/ |
3394 | | |
3395 | | static void InvalidateNullCountRec(const struct ArrowSchema *schema, |
3396 | | struct ArrowArray *array) |
3397 | 0 | { |
3398 | 0 | if (schema->flags & ARROW_FLAG_NULLABLE) |
3399 | 0 | array->null_count = -1; |
3400 | 0 | for (int i = 0; i < array->n_children; ++i) |
3401 | 0 | InvalidateNullCountRec(schema->children[i], array->children[i]); |
3402 | 0 | } |
3403 | | |
3404 | | /************************************************************************/ |
3405 | | /* CompactListArray() */ |
3406 | | /************************************************************************/ |
3407 | | |
3408 | | template <class OffsetType> |
3409 | | static bool CompactListArray(const struct ArrowSchema *schema, |
3410 | | struct ArrowArray *array, size_t iStart, |
3411 | | const std::vector<bool> &abyValidityFromFilters, |
3412 | | size_t nNewLength) |
3413 | 0 | { |
3414 | 0 | CPLAssert(static_cast<size_t>(array->length) >= |
3415 | 0 | iStart + abyValidityFromFilters.size()); |
3416 | 0 | CPLAssert(array->n_children == 1); |
3417 | 0 | CPLAssert(array->n_buffers == 2); |
3418 | | |
3419 | 0 | const auto psChildSchema = schema->children[0]; |
3420 | 0 | const auto psChildArray = array->children[0]; |
3421 | |
|
3422 | 0 | const size_t nLength = abyValidityFromFilters.size(); |
3423 | 0 | const size_t nOffset = static_cast<size_t>(array->offset); |
3424 | 0 | OffsetType *panOffsets = |
3425 | 0 | static_cast<OffsetType *>(const_cast<void *>(array->buffers[1])) + |
3426 | 0 | nOffset; |
3427 | |
|
3428 | 0 | if (panOffsets[iStart + nLength] > panOffsets[iStart]) |
3429 | 0 | { |
3430 | 0 | std::vector<bool> abyChildValidity( |
3431 | 0 | static_cast<size_t>(panOffsets[iStart + nLength] - |
3432 | 0 | panOffsets[iStart]), |
3433 | 0 | true); |
3434 | 0 | size_t j = iStart; |
3435 | 0 | OffsetType nCurOffset = panOffsets[iStart]; |
3436 | 0 | for (size_t i = 0; i < nLength; ++i) |
3437 | 0 | { |
3438 | 0 | if (abyValidityFromFilters[i]) |
3439 | 0 | { |
3440 | 0 | const auto nSize = |
3441 | 0 | panOffsets[i + iStart + 1] - panOffsets[i + iStart]; |
3442 | 0 | panOffsets[j] = nCurOffset; |
3443 | 0 | nCurOffset += nSize; |
3444 | 0 | ++j; |
3445 | 0 | } |
3446 | 0 | else |
3447 | 0 | { |
3448 | 0 | const auto nStartOffset = panOffsets[i + iStart]; |
3449 | 0 | const auto nEndOffset = panOffsets[i + iStart + 1]; |
3450 | 0 | if (nStartOffset != nEndOffset) |
3451 | 0 | { |
3452 | 0 | if (nStartOffset >= |
3453 | 0 | panOffsets[iStart] + abyChildValidity.size()) |
3454 | 0 | { |
3455 | | // shouldn't happen in sane arrays... |
3456 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
3457 | 0 | "nStartOffset >= panOffsets[iStart] + " |
3458 | 0 | "abyChildValidity.size()"); |
3459 | 0 | return false; |
3460 | 0 | } |
3461 | | // nEndOffset might be equal to abyChildValidity.size() |
3462 | 0 | if (nEndOffset > |
3463 | 0 | panOffsets[iStart] + abyChildValidity.size()) |
3464 | 0 | { |
3465 | | // shouldn't happen in sane arrays... |
3466 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
3467 | 0 | "nEndOffset > panOffsets[iStart] + " |
3468 | 0 | "abyChildValidity.size()"); |
3469 | 0 | return false; |
3470 | 0 | } |
3471 | 0 | for (auto k = nStartOffset - panOffsets[iStart]; |
3472 | 0 | k < nEndOffset - panOffsets[iStart]; ++k) |
3473 | 0 | abyChildValidity[static_cast<size_t>(k)] = false; |
3474 | 0 | } |
3475 | 0 | } |
3476 | 0 | } |
3477 | 0 | panOffsets[j] = nCurOffset; |
3478 | 0 | const size_t nChildNewLength = static_cast<size_t>(panOffsets[j]); |
3479 | | // To please Arrow validation |
3480 | 0 | for (; j < iStart + nLength; ++j) |
3481 | 0 | panOffsets[j] = nCurOffset; |
3482 | |
|
3483 | 0 | if (!CompactArray(psChildSchema, psChildArray, |
3484 | 0 | static_cast<size_t>(panOffsets[iStart]), |
3485 | 0 | abyChildValidity, nChildNewLength)) |
3486 | 0 | return false; |
3487 | | |
3488 | 0 | CPLAssert(psChildArray->length == |
3489 | 0 | static_cast<int64_t>(nChildNewLength)); |
3490 | 0 | } |
3491 | 0 | else |
3492 | 0 | { |
3493 | 0 | InvalidateNullCountRec(psChildSchema, psChildArray); |
3494 | 0 | } |
3495 | | |
3496 | 0 | if (schema->flags & ARROW_FLAG_NULLABLE) |
3497 | 0 | CompactValidityBuffer(schema, array, iStart, abyValidityFromFilters, |
3498 | 0 | nNewLength); |
3499 | |
|
3500 | 0 | array->length = nNewLength; |
3501 | |
|
3502 | 0 | return true; |
3503 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:bool CompactListArray<unsigned int>(ArrowSchema const*, ArrowArray*, unsigned long, std::__1::vector<bool, std::__1::allocator<bool> > const&, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:bool CompactListArray<unsigned long>(ArrowSchema const*, ArrowArray*, unsigned long, std::__1::vector<bool, std::__1::allocator<bool> > const&, unsigned long) |
3504 | | |
3505 | | /************************************************************************/ |
3506 | | /* CompactFixedSizeListArray() */ |
3507 | | /************************************************************************/ |
3508 | | |
3509 | | static bool |
3510 | | CompactFixedSizeListArray(const struct ArrowSchema *schema, |
3511 | | struct ArrowArray *array, size_t N, size_t iStart, |
3512 | | const std::vector<bool> &abyValidityFromFilters, |
3513 | | size_t nNewLength) |
3514 | 0 | { |
3515 | 0 | CPLAssert(static_cast<size_t>(array->length) >= |
3516 | 0 | iStart + abyValidityFromFilters.size()); |
3517 | 0 | CPLAssert(array->n_children == 1); |
3518 | | |
3519 | 0 | const auto psChildSchema = schema->children[0]; |
3520 | 0 | const auto psChildArray = array->children[0]; |
3521 | |
|
3522 | 0 | const size_t nLength = abyValidityFromFilters.size(); |
3523 | 0 | const size_t nOffset = static_cast<size_t>(array->offset); |
3524 | 0 | std::vector<bool> abyChildValidity(N * nLength, true); |
3525 | 0 | size_t nChildNewLength = (iStart + nOffset) * N; |
3526 | 0 | size_t nSrcLength = 0; |
3527 | 0 | for (size_t i = 0; i < nLength; ++i) |
3528 | 0 | { |
3529 | 0 | if (abyValidityFromFilters[i]) |
3530 | 0 | { |
3531 | 0 | nChildNewLength += N; |
3532 | 0 | nSrcLength++; |
3533 | 0 | } |
3534 | 0 | else |
3535 | 0 | { |
3536 | 0 | const size_t nStartOffset = i * N; |
3537 | 0 | const size_t nEndOffset = (i + 1) * N; |
3538 | 0 | for (size_t k = nStartOffset; k < nEndOffset; ++k) |
3539 | 0 | abyChildValidity[k] = false; |
3540 | 0 | } |
3541 | 0 | } |
3542 | 0 | CPL_IGNORE_RET_VAL(nSrcLength); |
3543 | 0 | CPLAssert(iStart + nSrcLength == nNewLength); |
3544 | | |
3545 | 0 | if (!CompactArray(psChildSchema, psChildArray, (iStart + nOffset) * N, |
3546 | 0 | abyChildValidity, nChildNewLength)) |
3547 | 0 | return false; |
3548 | | |
3549 | 0 | if (schema->flags & ARROW_FLAG_NULLABLE) |
3550 | 0 | CompactValidityBuffer(schema, array, iStart, abyValidityFromFilters, |
3551 | 0 | nNewLength); |
3552 | |
|
3553 | 0 | array->length = nNewLength; |
3554 | |
|
3555 | 0 | CPLAssert(psChildArray->length >= |
3556 | 0 | static_cast<int64_t>(N) * (array->length + array->offset)); |
3557 | | |
3558 | 0 | return true; |
3559 | 0 | } |
3560 | | |
3561 | | /************************************************************************/ |
3562 | | /* CompactMapArray() */ |
3563 | | /************************************************************************/ |
3564 | | |
3565 | | static bool CompactMapArray(const struct ArrowSchema *schema, |
3566 | | struct ArrowArray *array, size_t iStart, |
3567 | | const std::vector<bool> &abyValidityFromFilters, |
3568 | | size_t nNewLength) |
3569 | 0 | { |
3570 | 0 | return CompactListArray<uint32_t>(schema, array, iStart, |
3571 | 0 | abyValidityFromFilters, nNewLength); |
3572 | 0 | } |
3573 | | |
3574 | | /************************************************************************/ |
3575 | | /* CompactArray() */ |
3576 | | /************************************************************************/ |
3577 | | |
3578 | | static bool CompactArray(const struct ArrowSchema *schema, |
3579 | | struct ArrowArray *array, size_t iStart, |
3580 | | const std::vector<bool> &abyValidityFromFilters, |
3581 | | size_t nNewLength) |
3582 | 0 | { |
3583 | 0 | const char *format = schema->format; |
3584 | |
|
3585 | 0 | if (IsStructure(format)) |
3586 | 0 | { |
3587 | 0 | if (!CompactStructArray(schema, array, iStart, abyValidityFromFilters, |
3588 | 0 | nNewLength)) |
3589 | 0 | return false; |
3590 | 0 | } |
3591 | 0 | else if (IsList(format)) |
3592 | 0 | { |
3593 | 0 | if (!CompactListArray<uint32_t>(schema, array, iStart, |
3594 | 0 | abyValidityFromFilters, nNewLength)) |
3595 | 0 | return false; |
3596 | 0 | } |
3597 | 0 | else if (IsLargeList(format)) |
3598 | 0 | { |
3599 | 0 | if (!CompactListArray<uint64_t>(schema, array, iStart, |
3600 | 0 | abyValidityFromFilters, nNewLength)) |
3601 | 0 | return false; |
3602 | 0 | } |
3603 | 0 | else if (IsMap(format)) |
3604 | 0 | { |
3605 | 0 | if (!CompactMapArray(schema, array, iStart, abyValidityFromFilters, |
3606 | 0 | nNewLength)) |
3607 | 0 | return false; |
3608 | 0 | } |
3609 | 0 | else if (IsFixedSizeList(format)) |
3610 | 0 | { |
3611 | 0 | const int N = GetFixedSizeList(format); |
3612 | 0 | if (N <= 0) |
3613 | 0 | return false; |
3614 | 0 | if (!CompactFixedSizeListArray(schema, array, static_cast<size_t>(N), |
3615 | 0 | iStart, abyValidityFromFilters, |
3616 | 0 | nNewLength)) |
3617 | 0 | return false; |
3618 | 0 | } |
3619 | 0 | else if (IsBoolean(format)) |
3620 | 0 | { |
3621 | 0 | CompactBoolArray(schema, array, iStart, abyValidityFromFilters, |
3622 | 0 | nNewLength); |
3623 | 0 | } |
3624 | 0 | else if (IsInt8(format) || IsUInt8(format)) |
3625 | 0 | { |
3626 | 0 | CompactPrimitiveArray<uint8_t>(schema, array, iStart, |
3627 | 0 | abyValidityFromFilters, nNewLength); |
3628 | 0 | } |
3629 | 0 | else if (IsInt16(format) || IsUInt16(format) || IsFloat16(format)) |
3630 | 0 | { |
3631 | 0 | CompactPrimitiveArray<uint16_t>(schema, array, iStart, |
3632 | 0 | abyValidityFromFilters, nNewLength); |
3633 | 0 | } |
3634 | 0 | else if (IsInt32(format) || IsUInt32(format) || IsFloat32(format) || |
3635 | 0 | strcmp(format, "tdD") == 0 || strcmp(format, "tts") == 0 || |
3636 | 0 | strcmp(format, "ttm") == 0) |
3637 | 0 | { |
3638 | 0 | CompactPrimitiveArray<uint32_t>(schema, array, iStart, |
3639 | 0 | abyValidityFromFilters, nNewLength); |
3640 | 0 | } |
3641 | 0 | else if (IsInt64(format) || IsUInt64(format) || IsFloat64(format) || |
3642 | 0 | strcmp(format, "tdm") == 0 || strcmp(format, "ttu") == 0 || |
3643 | 0 | strcmp(format, "ttn") == 0 || strncmp(format, "ts", 2) == 0) |
3644 | 0 | { |
3645 | 0 | CompactPrimitiveArray<uint64_t>(schema, array, iStart, |
3646 | 0 | abyValidityFromFilters, nNewLength); |
3647 | 0 | } |
3648 | 0 | else if (IsString(format) || IsBinary(format)) |
3649 | 0 | { |
3650 | 0 | CompactStringOrBinaryArray<uint32_t>( |
3651 | 0 | schema, array, iStart, abyValidityFromFilters, nNewLength); |
3652 | 0 | } |
3653 | 0 | else if (IsLargeString(format) || IsLargeBinary(format)) |
3654 | 0 | { |
3655 | 0 | CompactStringOrBinaryArray<uint64_t>( |
3656 | 0 | schema, array, iStart, abyValidityFromFilters, nNewLength); |
3657 | 0 | } |
3658 | 0 | else if (IsFixedWidthBinary(format)) |
3659 | 0 | { |
3660 | 0 | const int nWidth = GetFixedWithBinary(format); |
3661 | 0 | CompactFixedWidthArray(schema, array, nWidth, iStart, |
3662 | 0 | abyValidityFromFilters, nNewLength); |
3663 | 0 | } |
3664 | 0 | else if (IsDecimal(format)) |
3665 | 0 | { |
3666 | 0 | int nPrecision = 0; |
3667 | 0 | int nScale = 0; |
3668 | 0 | int nWidthInBytes = 0; |
3669 | 0 | if (!ParseDecimalFormat(format, nPrecision, nScale, nWidthInBytes)) |
3670 | 0 | { |
3671 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
3672 | 0 | "Unexpected error in PostFilterArrowArray(): unhandled " |
3673 | 0 | "field format: %s", |
3674 | 0 | format); |
3675 | |
|
3676 | 0 | return false; |
3677 | 0 | } |
3678 | 0 | CompactFixedWidthArray(schema, array, nWidthInBytes, iStart, |
3679 | 0 | abyValidityFromFilters, nNewLength); |
3680 | 0 | } |
3681 | 0 | else |
3682 | 0 | { |
3683 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
3684 | 0 | "Unexpected error in CompactArray(): unhandled " |
3685 | 0 | "field format: %s", |
3686 | 0 | format); |
3687 | 0 | return false; |
3688 | 0 | } |
3689 | | |
3690 | 0 | return true; |
3691 | 0 | } |
3692 | | |
3693 | | /************************************************************************/ |
3694 | | /* FillValidityArrayFromWKBArray() */ |
3695 | | /************************************************************************/ |
3696 | | |
3697 | | template <class OffsetType> |
3698 | | static size_t |
3699 | | FillValidityArrayFromWKBArray(struct ArrowArray *array, const OGRLayer *poLayer, |
3700 | | std::vector<bool> &abyValidityFromFilters) |
3701 | 0 | { |
3702 | 0 | const size_t nLength = static_cast<size_t>(array->length); |
3703 | 0 | const uint8_t *pabyValidity = |
3704 | 0 | array->null_count == 0 |
3705 | 0 | ? nullptr |
3706 | 0 | : static_cast<const uint8_t *>(array->buffers[0]); |
3707 | 0 | const size_t nOffset = static_cast<size_t>(array->offset); |
3708 | 0 | const OffsetType *panOffsets = |
3709 | 0 | static_cast<const OffsetType *>(array->buffers[1]) + nOffset; |
3710 | 0 | const GByte *pabyData = static_cast<const GByte *>(array->buffers[2]); |
3711 | 0 | OGREnvelope sEnvelope; |
3712 | 0 | abyValidityFromFilters.resize(nLength); |
3713 | 0 | size_t nCountIntersecting = 0; |
3714 | 0 | for (size_t i = 0; i < nLength; ++i) |
3715 | 0 | { |
3716 | 0 | if (!pabyValidity || TestBit(pabyValidity, i + nOffset)) |
3717 | 0 | { |
3718 | 0 | const GByte *pabyWKB = pabyData + panOffsets[i]; |
3719 | 0 | const size_t nWKBSize = |
3720 | 0 | static_cast<size_t>(panOffsets[i + 1] - panOffsets[i]); |
3721 | 0 | if (poLayer->FilterWKBGeometry(pabyWKB, nWKBSize, |
3722 | 0 | /* bEnvelopeAlreadySet=*/false, |
3723 | 0 | sEnvelope)) |
3724 | 0 | { |
3725 | 0 | abyValidityFromFilters[i] = true; |
3726 | 0 | nCountIntersecting++; |
3727 | 0 | } |
3728 | 0 | } |
3729 | 0 | } |
3730 | 0 | return nCountIntersecting; |
3731 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:unsigned long FillValidityArrayFromWKBArray<unsigned int>(ArrowArray*, OGRLayer const*, std::__1::vector<bool, std::__1::allocator<bool> >&) Unexecuted instantiation: ogrlayerarrow.cpp:unsigned long FillValidityArrayFromWKBArray<unsigned long>(ArrowArray*, OGRLayer const*, std::__1::vector<bool, std::__1::allocator<bool> >&) |
3732 | | |
3733 | | /************************************************************************/ |
3734 | | /* ArrowTimestampToOGRDateTime() */ |
3735 | | /************************************************************************/ |
3736 | | |
3737 | | static void ArrowTimestampToOGRDateTime(int64_t nTimestamp, |
3738 | | int nInvFactorToSecond, |
3739 | | const char *pszTZ, OGRFeature &oFeature, |
3740 | | int iField) |
3741 | 0 | { |
3742 | 0 | double floatingPart = 0; |
3743 | 0 | if (nInvFactorToSecond) |
3744 | 0 | { |
3745 | 0 | floatingPart = |
3746 | 0 | (nTimestamp % nInvFactorToSecond) / double(nInvFactorToSecond); |
3747 | 0 | nTimestamp /= nInvFactorToSecond; |
3748 | 0 | } |
3749 | 0 | int nTZFlag = 0; |
3750 | 0 | const size_t nTZLen = strlen(pszTZ); |
3751 | 0 | if ((nTZLen == 3 && strcmp(pszTZ, "UTC") == 0) || |
3752 | 0 | (nTZLen == 7 && strcmp(pszTZ, "Etc/UTC") == 0)) |
3753 | 0 | { |
3754 | 0 | nTZFlag = 100; |
3755 | 0 | } |
3756 | 0 | else if (nTZLen == 6 && (pszTZ[0] == '+' || pszTZ[0] == '-') && |
3757 | 0 | pszTZ[3] == ':') |
3758 | 0 | { |
3759 | 0 | int nTZHour = atoi(pszTZ + 1); |
3760 | 0 | int nTZMin = atoi(pszTZ + 4); |
3761 | 0 | if (nTZHour >= 0 && nTZHour <= 14 && nTZMin >= 0 && nTZMin < 60 && |
3762 | 0 | (nTZMin % 15) == 0) |
3763 | 0 | { |
3764 | 0 | nTZFlag = (nTZHour * 4) + (nTZMin / 15); |
3765 | 0 | if (pszTZ[0] == '+') |
3766 | 0 | { |
3767 | 0 | nTZFlag = 100 + nTZFlag; |
3768 | 0 | nTimestamp += nTZHour * 3600 + nTZMin * 60; |
3769 | 0 | } |
3770 | 0 | else |
3771 | 0 | { |
3772 | 0 | nTZFlag = 100 - nTZFlag; |
3773 | 0 | nTimestamp -= nTZHour * 3600 + nTZMin * 60; |
3774 | 0 | } |
3775 | 0 | } |
3776 | 0 | } |
3777 | 0 | struct tm dt; |
3778 | 0 | CPLUnixTimeToYMDHMS(nTimestamp, &dt); |
3779 | 0 | oFeature.SetField(iField, dt.tm_year + 1900, dt.tm_mon + 1, dt.tm_mday, |
3780 | 0 | dt.tm_hour, dt.tm_min, |
3781 | 0 | static_cast<float>(dt.tm_sec + floatingPart), nTZFlag); |
3782 | 0 | } |
3783 | | |
3784 | | /************************************************************************/ |
3785 | | /* BuildMapFieldNameToArrowPath() */ |
3786 | | /************************************************************************/ |
3787 | | |
3788 | | static void |
3789 | | BuildMapFieldNameToArrowPath(const struct ArrowSchema *schema, |
3790 | | std::map<std::string, std::vector<int>> &oMap, |
3791 | | const std::string &osPrefix, |
3792 | | std::vector<int> &anArrowPath) |
3793 | 0 | { |
3794 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
3795 | 0 | { |
3796 | 0 | auto psChild = schema->children[i]; |
3797 | 0 | anArrowPath.push_back(static_cast<int>(i)); |
3798 | 0 | if (IsStructure(psChild->format)) |
3799 | 0 | { |
3800 | 0 | std::string osNewPrefix(osPrefix); |
3801 | 0 | osNewPrefix += psChild->name; |
3802 | 0 | osNewPrefix += "."; |
3803 | 0 | BuildMapFieldNameToArrowPath(psChild, oMap, osNewPrefix, |
3804 | 0 | anArrowPath); |
3805 | 0 | } |
3806 | 0 | else |
3807 | 0 | { |
3808 | 0 | oMap[osPrefix + psChild->name] = anArrowPath; |
3809 | 0 | } |
3810 | 0 | anArrowPath.pop_back(); |
3811 | 0 | } |
3812 | 0 | } |
3813 | | |
3814 | | /************************************************************************/ |
3815 | | /* FillFieldList() */ |
3816 | | /************************************************************************/ |
3817 | | |
3818 | | template <typename ListOffsetType, typename ArrowType, |
3819 | | typename OGRType = ArrowType> |
3820 | | inline static void FillFieldList(const struct ArrowArray *array, |
3821 | | int iOGRFieldIdx, size_t nOffsettedIndex, |
3822 | | const struct ArrowArray *childArray, |
3823 | | OGRFeature &oFeature) |
3824 | 0 | { |
3825 | 0 | const auto panOffsets = |
3826 | 0 | static_cast<const ListOffsetType *>(array->buffers[1]) + |
3827 | 0 | nOffsettedIndex; |
3828 | 0 | std::vector<OGRType> aValues; |
3829 | 0 | const auto *paValues = |
3830 | 0 | static_cast<const ArrowType *>(childArray->buffers[1]); |
3831 | 0 | for (size_t i = static_cast<size_t>(panOffsets[0]); |
3832 | 0 | i < static_cast<size_t>(panOffsets[1]); ++i) |
3833 | 0 | { |
3834 | 0 | aValues.push_back(static_cast<OGRType>(paValues[i])); |
3835 | 0 | } |
3836 | 0 | oFeature.SetField(iOGRFieldIdx, static_cast<int>(aValues.size()), |
3837 | 0 | aValues.data()); |
3838 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, signed char, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, signed char, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, unsigned char, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, unsigned char, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, short, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, short, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, unsigned short, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, unsigned short, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, int, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, int, int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, unsigned int, long long>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, unsigned int, long long>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, long, long long>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, long, long long>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, unsigned long, double>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, unsigned long, double>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, float, double>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, float, double>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned int, double, double>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldList<unsigned long, double, double>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) |
3839 | | |
3840 | | /************************************************************************/ |
3841 | | /* FillFieldListFromBool() */ |
3842 | | /************************************************************************/ |
3843 | | |
3844 | | template <typename ListOffsetType> |
3845 | | inline static void |
3846 | | FillFieldListFromBool(const struct ArrowArray *array, int iOGRFieldIdx, |
3847 | | size_t nOffsettedIndex, |
3848 | | const struct ArrowArray *childArray, OGRFeature &oFeature) |
3849 | 0 | { |
3850 | 0 | const auto panOffsets = |
3851 | 0 | static_cast<const ListOffsetType *>(array->buffers[1]) + |
3852 | 0 | nOffsettedIndex; |
3853 | 0 | std::vector<int> aValues; |
3854 | 0 | const auto *paValues = static_cast<const uint8_t *>(childArray->buffers[1]); |
3855 | 0 | for (size_t i = static_cast<size_t>(panOffsets[0]); |
3856 | 0 | i < static_cast<size_t>(panOffsets[1]); ++i) |
3857 | 0 | { |
3858 | 0 | aValues.push_back(TestBit(paValues, i) ? 1 : 0); |
3859 | 0 | } |
3860 | 0 | oFeature.SetField(iOGRFieldIdx, static_cast<int>(aValues.size()), |
3861 | 0 | aValues.data()); |
3862 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldListFromBool<unsigned int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldListFromBool<unsigned long>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) |
3863 | | |
3864 | | /************************************************************************/ |
3865 | | /* FillFieldListFromHalfFloat() */ |
3866 | | /************************************************************************/ |
3867 | | |
3868 | | template <typename ListOffsetType> |
3869 | | inline static void FillFieldListFromHalfFloat( |
3870 | | const struct ArrowArray *array, int iOGRFieldIdx, size_t nOffsettedIndex, |
3871 | | const struct ArrowArray *childArray, OGRFeature &oFeature) |
3872 | 0 | { |
3873 | 0 | const auto panOffsets = |
3874 | 0 | static_cast<const ListOffsetType *>(array->buffers[1]) + |
3875 | 0 | nOffsettedIndex; |
3876 | 0 | std::vector<double> aValues; |
3877 | 0 | const auto *paValues = |
3878 | 0 | static_cast<const uint16_t *>(childArray->buffers[1]); |
3879 | 0 | for (size_t i = static_cast<size_t>(panOffsets[0]); |
3880 | 0 | i < static_cast<size_t>(panOffsets[1]); ++i) |
3881 | 0 | { |
3882 | 0 | const auto nFloat16AsUInt32 = CPLHalfToFloat(paValues[i]); |
3883 | 0 | float f; |
3884 | 0 | memcpy(&f, &nFloat16AsUInt32, sizeof(f)); |
3885 | 0 | aValues.push_back(f); |
3886 | 0 | } |
3887 | 0 | oFeature.SetField(iOGRFieldIdx, static_cast<int>(aValues.size()), |
3888 | 0 | aValues.data()); |
3889 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldListFromHalfFloat<unsigned int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldListFromHalfFloat<unsigned long>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) |
3890 | | |
3891 | | /************************************************************************/ |
3892 | | /* FillFieldListFromString() */ |
3893 | | /************************************************************************/ |
3894 | | |
3895 | | template <typename ListOffsetType, typename StringOffsetType> |
3896 | | inline static void FillFieldListFromString(const struct ArrowArray *array, |
3897 | | int iOGRFieldIdx, |
3898 | | size_t nOffsettedIndex, |
3899 | | const struct ArrowArray *childArray, |
3900 | | OGRFeature &oFeature) |
3901 | 0 | { |
3902 | 0 | const auto panOffsets = |
3903 | 0 | static_cast<const ListOffsetType *>(array->buffers[1]) + |
3904 | 0 | nOffsettedIndex; |
3905 | 0 | CPLStringList aosVals; |
3906 | 0 | const auto panSubOffsets = |
3907 | 0 | static_cast<const StringOffsetType *>(childArray->buffers[1]); |
3908 | 0 | const char *pszValues = static_cast<const char *>(childArray->buffers[2]); |
3909 | 0 | std::string osTmp; |
3910 | 0 | for (size_t i = static_cast<size_t>(panOffsets[0]); |
3911 | 0 | i < static_cast<size_t>(panOffsets[1]); ++i) |
3912 | 0 | { |
3913 | 0 | osTmp.assign( |
3914 | 0 | pszValues + panSubOffsets[i], |
3915 | 0 | static_cast<size_t>(panSubOffsets[i + 1] - panSubOffsets[i])); |
3916 | 0 | aosVals.AddString(osTmp.c_str()); |
3917 | 0 | } |
3918 | 0 | oFeature.SetField(iOGRFieldIdx, aosVals.List()); |
3919 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldListFromString<unsigned int, unsigned int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldListFromString<unsigned long, unsigned int>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldListFromString<unsigned int, unsigned long>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldListFromString<unsigned long, unsigned long>(ArrowArray const*, int, unsigned long, ArrowArray const*, OGRFeature&) |
3920 | | |
3921 | | /************************************************************************/ |
3922 | | /* FillFieldFixedSizeList() */ |
3923 | | /************************************************************************/ |
3924 | | |
3925 | | template <typename ArrowType, typename OGRType = ArrowType> |
3926 | | inline static void FillFieldFixedSizeList( |
3927 | | const struct ArrowArray *, int iOGRFieldIdx, size_t nOffsettedIndex, |
3928 | | const int nItems, const struct ArrowArray *childArray, OGRFeature &oFeature) |
3929 | 0 | { |
3930 | 0 | std::vector<OGRType> aValues; |
3931 | 0 | const auto *paValues = |
3932 | 0 | static_cast<const ArrowType *>(childArray->buffers[1]) + |
3933 | 0 | childArray->offset + nOffsettedIndex * nItems; |
3934 | 0 | for (int i = 0; i < nItems; ++i) |
3935 | 0 | { |
3936 | 0 | aValues.push_back(static_cast<OGRType>(paValues[i])); |
3937 | 0 | } |
3938 | 0 | oFeature.SetField(iOGRFieldIdx, static_cast<int>(aValues.size()), |
3939 | 0 | aValues.data()); |
3940 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<signed char, int>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<unsigned char, int>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<short, int>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<unsigned short, int>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<int, int>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<unsigned int, long long>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<long, long long>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<unsigned long, double>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<float, double>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeList<double, double>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) |
3941 | | |
3942 | | /************************************************************************/ |
3943 | | /* FillFieldFixedSizeListString() */ |
3944 | | /************************************************************************/ |
3945 | | |
3946 | | template <typename StringOffsetType> |
3947 | | inline static void FillFieldFixedSizeListString( |
3948 | | const struct ArrowArray *, int iOGRFieldIdx, size_t nOffsettedIndex, |
3949 | | const int nItems, const struct ArrowArray *childArray, OGRFeature &oFeature) |
3950 | 0 | { |
3951 | 0 | CPLStringList aosVals; |
3952 | 0 | const auto panSubOffsets = |
3953 | 0 | static_cast<const StringOffsetType *>(childArray->buffers[1]) + |
3954 | 0 | childArray->offset + nOffsettedIndex * nItems; |
3955 | 0 | const char *pszValues = static_cast<const char *>(childArray->buffers[2]); |
3956 | 0 | std::string osTmp; |
3957 | 0 | for (int i = 0; i < nItems; ++i) |
3958 | 0 | { |
3959 | 0 | osTmp.assign( |
3960 | 0 | pszValues + panSubOffsets[i], |
3961 | 0 | static_cast<size_t>(panSubOffsets[i + 1] - panSubOffsets[i])); |
3962 | 0 | aosVals.AddString(osTmp.c_str()); |
3963 | 0 | } |
3964 | 0 | oFeature.SetField(iOGRFieldIdx, aosVals.List()); |
3965 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeListString<unsigned int>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldFixedSizeListString<unsigned long>(ArrowArray const*, int, unsigned long, int, ArrowArray const*, OGRFeature&) |
3966 | | |
3967 | | /************************************************************************/ |
3968 | | /* GetValue() */ |
3969 | | /************************************************************************/ |
3970 | | |
3971 | | template <typename ArrowType> |
3972 | | inline static ArrowType GetValue(const struct ArrowArray *array, |
3973 | | size_t iFeature) |
3974 | 0 | { |
3975 | 0 | const auto *panValues = static_cast<const ArrowType *>(array->buffers[1]); |
3976 | 0 | return panValues[iFeature + array->offset]; |
3977 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:unsigned char GetValue<unsigned char>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:signed char GetValue<signed char>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:unsigned short GetValue<unsigned short>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:short GetValue<short>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:unsigned int GetValue<unsigned int>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:int GetValue<int>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:unsigned long GetValue<unsigned long>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:long GetValue<long>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:float GetValue<float>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:double GetValue<double>(ArrowArray const*, unsigned long) |
3978 | | |
3979 | | template <> bool GetValue<bool>(const struct ArrowArray *array, size_t iFeature) |
3980 | 0 | { |
3981 | 0 | const auto *pabyValues = static_cast<const uint8_t *>(array->buffers[1]); |
3982 | 0 | return TestBit(pabyValues, iFeature + static_cast<size_t>(array->offset)); |
3983 | 0 | } |
3984 | | |
3985 | | /************************************************************************/ |
3986 | | /* GetValueFloat16() */ |
3987 | | /************************************************************************/ |
3988 | | |
3989 | | static float GetValueFloat16(const struct ArrowArray *array, const size_t nIdx) |
3990 | 0 | { |
3991 | 0 | const auto *panValues = static_cast<const uint16_t *>(array->buffers[1]); |
3992 | 0 | const auto nFloat16AsUInt32 = |
3993 | 0 | CPLHalfToFloat(panValues[nIdx + array->offset]); |
3994 | 0 | float f; |
3995 | 0 | memcpy(&f, &nFloat16AsUInt32, sizeof(f)); |
3996 | 0 | return f; |
3997 | 0 | } |
3998 | | |
3999 | | /************************************************************************/ |
4000 | | /* GetValueDecimal() */ |
4001 | | /************************************************************************/ |
4002 | | |
4003 | | static double GetValueDecimal(const struct ArrowArray *array, |
4004 | | const int nWidthIn64BitWord, const int nScale, |
4005 | | const size_t nIdx) |
4006 | 0 | { |
4007 | 0 | #ifdef CPL_LSB |
4008 | 0 | const auto nIdxIn64BitWord = nIdx * nWidthIn64BitWord; |
4009 | | #else |
4010 | | const auto nIdxIn64BitWord = |
4011 | | nIdx * nWidthIn64BitWord + nWidthIn64BitWord - 1; |
4012 | | #endif |
4013 | 0 | const auto *panValues = static_cast<const int64_t *>(array->buffers[1]); |
4014 | 0 | const auto nVal = |
4015 | 0 | panValues[nIdxIn64BitWord + array->offset * nWidthIn64BitWord]; |
4016 | 0 | return static_cast<double>(nVal) * std::pow(10.0, -nScale); |
4017 | 0 | } |
4018 | | |
4019 | | /************************************************************************/ |
4020 | | /* GetString() */ |
4021 | | /************************************************************************/ |
4022 | | |
4023 | | template <class OffsetType> |
4024 | | static std::string GetString(const struct ArrowArray *array, const size_t nIdx) |
4025 | 0 | { |
4026 | 0 | const OffsetType *panOffsets = |
4027 | 0 | static_cast<const OffsetType *>(array->buffers[1]) + |
4028 | 0 | static_cast<size_t>(array->offset) + nIdx; |
4029 | 0 | const char *pabyStr = static_cast<const char *>(array->buffers[2]); |
4030 | 0 | std::string osStr; |
4031 | 0 | osStr.assign(pabyStr + static_cast<size_t>(panOffsets[0]), |
4032 | 0 | static_cast<size_t>(panOffsets[1] - panOffsets[0])); |
4033 | 0 | return osStr; |
4034 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > GetString<unsigned int>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > GetString<unsigned long>(ArrowArray const*, unsigned long) |
4035 | | |
4036 | | /************************************************************************/ |
4037 | | /* GetBinaryAsBase64() */ |
4038 | | /************************************************************************/ |
4039 | | |
4040 | | template <class OffsetType> |
4041 | | static std::string GetBinaryAsBase64(const struct ArrowArray *array, |
4042 | | const size_t nIdx) |
4043 | 0 | { |
4044 | 0 | const OffsetType *panOffsets = |
4045 | 0 | static_cast<const OffsetType *>(array->buffers[1]) + |
4046 | 0 | static_cast<size_t>(array->offset) + nIdx; |
4047 | 0 | const GByte *pabyData = static_cast<const GByte *>(array->buffers[2]); |
4048 | 0 | const size_t nLen = static_cast<size_t>(panOffsets[1] - panOffsets[0]); |
4049 | 0 | if (nLen > static_cast<size_t>(std::numeric_limits<int>::max())) |
4050 | 0 | { |
4051 | 0 | CPLError(CE_Failure, CPLE_AppDefined, "Too large binary"); |
4052 | 0 | return std::string(); |
4053 | 0 | } |
4054 | 0 | char *pszVal = CPLBase64Encode( |
4055 | 0 | static_cast<int>(nLen), pabyData + static_cast<size_t>(panOffsets[0])); |
4056 | 0 | std::string osStr(pszVal); |
4057 | 0 | CPLFree(pszVal); |
4058 | 0 | return osStr; |
4059 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > GetBinaryAsBase64<unsigned int>(ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > GetBinaryAsBase64<unsigned long>(ArrowArray const*, unsigned long) |
4060 | | |
4061 | | /************************************************************************/ |
4062 | | /* GetValueFixedWithBinaryAsBase64() */ |
4063 | | /************************************************************************/ |
4064 | | |
4065 | | static std::string |
4066 | | GetValueFixedWithBinaryAsBase64(const struct ArrowArray *array, |
4067 | | const int nWidth, const size_t nIdx) |
4068 | 0 | { |
4069 | 0 | const GByte *pabyData = static_cast<const GByte *>(array->buffers[1]); |
4070 | 0 | char *pszVal = CPLBase64Encode( |
4071 | 0 | nWidth, |
4072 | 0 | pabyData + (static_cast<size_t>(array->offset) + nIdx) * nWidth); |
4073 | 0 | std::string osStr(pszVal); |
4074 | 0 | CPLFree(pszVal); |
4075 | 0 | return osStr; |
4076 | 0 | } |
4077 | | |
4078 | | static CPLJSONObject GetObjectAsJSON(const struct ArrowSchema *schema, |
4079 | | const struct ArrowArray *array, |
4080 | | const size_t nIdx); |
4081 | | |
4082 | | /************************************************************************/ |
4083 | | /* AddToArray() */ |
4084 | | /************************************************************************/ |
4085 | | |
4086 | | static void AddToArray(CPLJSONArray &oArray, const struct ArrowSchema *schema, |
4087 | | const struct ArrowArray *array, const size_t nIdx) |
4088 | 0 | { |
4089 | 0 | if (IsBoolean(schema->format)) |
4090 | 0 | oArray.Add(GetValue<bool>(array, nIdx)); |
4091 | 0 | else if (IsUInt8(schema->format)) |
4092 | 0 | oArray.Add(GetValue<uint8_t>(array, nIdx)); |
4093 | 0 | else if (IsInt8(schema->format)) |
4094 | 0 | oArray.Add(GetValue<int8_t>(array, nIdx)); |
4095 | 0 | else if (IsUInt16(schema->format)) |
4096 | 0 | oArray.Add(GetValue<uint16_t>(array, nIdx)); |
4097 | 0 | else if (IsInt16(schema->format)) |
4098 | 0 | oArray.Add(GetValue<int16_t>(array, nIdx)); |
4099 | 0 | else if (IsUInt32(schema->format)) |
4100 | 0 | oArray.Add(static_cast<GIntBig>(GetValue<uint32_t>(array, nIdx))); |
4101 | 0 | else if (IsInt32(schema->format)) |
4102 | 0 | oArray.Add(GetValue<int32_t>(array, nIdx)); |
4103 | 0 | else if (IsUInt64(schema->format)) |
4104 | 0 | oArray.Add(GetValue<uint64_t>(array, nIdx)); |
4105 | 0 | else if (IsInt64(schema->format)) |
4106 | 0 | oArray.Add(static_cast<GIntBig>(GetValue<int64_t>(array, nIdx))); |
4107 | 0 | else if (IsFloat16(schema->format)) |
4108 | 0 | oArray.Add(GetValueFloat16(array, nIdx)); |
4109 | 0 | else if (IsFloat32(schema->format)) |
4110 | 0 | oArray.Add(GetValue<float>(array, nIdx)); |
4111 | 0 | else if (IsFloat64(schema->format)) |
4112 | 0 | oArray.Add(GetValue<double>(array, nIdx)); |
4113 | 0 | else if (IsString(schema->format)) |
4114 | 0 | oArray.Add(GetString<uint32_t>(array, nIdx)); |
4115 | 0 | else if (IsLargeString(schema->format)) |
4116 | 0 | oArray.Add(GetString<uint64_t>(array, nIdx)); |
4117 | 0 | else if (IsBinary(schema->format)) |
4118 | 0 | oArray.Add(GetBinaryAsBase64<uint32_t>(array, nIdx)); |
4119 | 0 | else if (IsLargeBinary(schema->format)) |
4120 | 0 | oArray.Add(GetBinaryAsBase64<uint64_t>(array, nIdx)); |
4121 | 0 | else if (IsFixedWidthBinary(schema->format)) |
4122 | 0 | oArray.Add(GetValueFixedWithBinaryAsBase64( |
4123 | 0 | array, GetFixedWithBinary(schema->format), nIdx)); |
4124 | 0 | else if (IsDecimal(schema->format)) |
4125 | 0 | { |
4126 | 0 | int nPrecision = 0; |
4127 | 0 | int nScale = 0; |
4128 | 0 | int nWidthInBytes = 0; |
4129 | 0 | const bool bOK = ParseDecimalFormat(schema->format, nPrecision, nScale, |
4130 | 0 | nWidthInBytes); |
4131 | | // Already validated |
4132 | 0 | CPLAssert(bOK); |
4133 | 0 | CPL_IGNORE_RET_VAL(bOK); |
4134 | 0 | oArray.Add(GetValueDecimal(array, nWidthInBytes / 8, nScale, nIdx)); |
4135 | 0 | } |
4136 | 0 | else |
4137 | 0 | oArray.Add(GetObjectAsJSON(schema, array, nIdx)); |
4138 | 0 | } |
4139 | | |
4140 | | /************************************************************************/ |
4141 | | /* GetListAsJSON() */ |
4142 | | /************************************************************************/ |
4143 | | |
4144 | | template <class OffsetType> |
4145 | | static CPLJSONArray GetListAsJSON(const struct ArrowSchema *schema, |
4146 | | const struct ArrowArray *array, |
4147 | | const size_t nIdx) |
4148 | 0 | { |
4149 | 0 | CPLJSONArray oArray; |
4150 | 0 | const auto panOffsets = static_cast<const OffsetType *>(array->buffers[1]) + |
4151 | 0 | array->offset + nIdx; |
4152 | 0 | const auto childSchema = schema->children[0]; |
4153 | 0 | const auto childArray = array->children[0]; |
4154 | 0 | const uint8_t *pabyValidity = |
4155 | 0 | childArray->null_count == 0 |
4156 | 0 | ? nullptr |
4157 | 0 | : static_cast<const uint8_t *>(childArray->buffers[0]); |
4158 | 0 | for (size_t k = static_cast<size_t>(panOffsets[0]); |
4159 | 0 | k < static_cast<size_t>(panOffsets[1]); k++) |
4160 | 0 | { |
4161 | 0 | if (!pabyValidity || |
4162 | 0 | TestBit(pabyValidity, k + static_cast<size_t>(childArray->offset))) |
4163 | 0 | { |
4164 | 0 | AddToArray(oArray, childSchema, childArray, k); |
4165 | 0 | } |
4166 | 0 | else |
4167 | 0 | { |
4168 | 0 | oArray.AddNull(); |
4169 | 0 | } |
4170 | 0 | } |
4171 | 0 | return oArray; |
4172 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:CPLJSONArray GetListAsJSON<unsigned int>(ArrowSchema const*, ArrowArray const*, unsigned long) Unexecuted instantiation: ogrlayerarrow.cpp:CPLJSONArray GetListAsJSON<unsigned long>(ArrowSchema const*, ArrowArray const*, unsigned long) |
4173 | | |
4174 | | /************************************************************************/ |
4175 | | /* GetFixedSizeListAsJSON() */ |
4176 | | /************************************************************************/ |
4177 | | |
4178 | | static CPLJSONArray GetFixedSizeListAsJSON(const struct ArrowSchema *schema, |
4179 | | const struct ArrowArray *array, |
4180 | | const size_t nIdx) |
4181 | 0 | { |
4182 | 0 | CPLJSONArray oArray; |
4183 | 0 | const int nVals = GetFixedSizeList(schema->format); |
4184 | 0 | const auto childSchema = schema->children[0]; |
4185 | 0 | const auto childArray = array->children[0]; |
4186 | 0 | const uint8_t *pabyValidity = |
4187 | 0 | childArray->null_count == 0 |
4188 | 0 | ? nullptr |
4189 | 0 | : static_cast<const uint8_t *>(childArray->buffers[0]); |
4190 | 0 | for (size_t k = nIdx * nVals; k < (nIdx + 1) * nVals; k++) |
4191 | 0 | { |
4192 | 0 | if (!pabyValidity || |
4193 | 0 | TestBit(pabyValidity, k + static_cast<size_t>(childArray->offset))) |
4194 | 0 | { |
4195 | 0 | AddToArray(oArray, childSchema, childArray, k); |
4196 | 0 | } |
4197 | 0 | else |
4198 | 0 | { |
4199 | 0 | oArray.AddNull(); |
4200 | 0 | } |
4201 | 0 | } |
4202 | 0 | return oArray; |
4203 | 0 | } |
4204 | | |
4205 | | /************************************************************************/ |
4206 | | /* AddToDict() */ |
4207 | | /************************************************************************/ |
4208 | | |
4209 | | static void AddToDict(CPLJSONObject &oDict, const std::string &osKey, |
4210 | | const struct ArrowSchema *schema, |
4211 | | const struct ArrowArray *array, const size_t nIdx) |
4212 | 0 | { |
4213 | 0 | if (IsBoolean(schema->format)) |
4214 | 0 | oDict.Add(osKey, GetValue<bool>(array, nIdx)); |
4215 | 0 | else if (IsUInt8(schema->format)) |
4216 | 0 | oDict.Add(osKey, GetValue<uint8_t>(array, nIdx)); |
4217 | 0 | else if (IsInt8(schema->format)) |
4218 | 0 | oDict.Add(osKey, GetValue<int8_t>(array, nIdx)); |
4219 | 0 | else if (IsUInt16(schema->format)) |
4220 | 0 | oDict.Add(osKey, GetValue<uint16_t>(array, nIdx)); |
4221 | 0 | else if (IsInt16(schema->format)) |
4222 | 0 | oDict.Add(osKey, GetValue<int16_t>(array, nIdx)); |
4223 | 0 | else if (IsUInt32(schema->format)) |
4224 | 0 | oDict.Add(osKey, static_cast<GIntBig>(GetValue<uint32_t>(array, nIdx))); |
4225 | 0 | else if (IsInt32(schema->format)) |
4226 | 0 | oDict.Add(osKey, GetValue<int32_t>(array, nIdx)); |
4227 | 0 | else if (IsUInt64(schema->format)) |
4228 | 0 | oDict.Add(osKey, GetValue<uint64_t>(array, nIdx)); |
4229 | 0 | else if (IsInt64(schema->format)) |
4230 | 0 | oDict.Add(osKey, static_cast<GIntBig>(GetValue<int64_t>(array, nIdx))); |
4231 | 0 | else if (IsFloat16(schema->format)) |
4232 | 0 | oDict.Add(osKey, GetValueFloat16(array, nIdx)); |
4233 | 0 | else if (IsFloat32(schema->format)) |
4234 | 0 | oDict.Add(osKey, GetValue<float>(array, nIdx)); |
4235 | 0 | else if (IsFloat64(schema->format)) |
4236 | 0 | oDict.Add(osKey, GetValue<double>(array, nIdx)); |
4237 | 0 | else if (IsString(schema->format)) |
4238 | 0 | oDict.Add(osKey, GetString<uint32_t>(array, nIdx)); |
4239 | 0 | else if (IsLargeString(schema->format)) |
4240 | 0 | oDict.Add(osKey, GetString<uint64_t>(array, nIdx)); |
4241 | 0 | else if (IsBinary(schema->format)) |
4242 | 0 | oDict.Add(osKey, GetBinaryAsBase64<uint32_t>(array, nIdx)); |
4243 | 0 | else if (IsLargeBinary(schema->format)) |
4244 | 0 | oDict.Add(osKey, GetBinaryAsBase64<uint64_t>(array, nIdx)); |
4245 | 0 | else if (IsFixedWidthBinary(schema->format)) |
4246 | 0 | oDict.Add(osKey, GetValueFixedWithBinaryAsBase64( |
4247 | 0 | array, GetFixedWithBinary(schema->format), nIdx)); |
4248 | 0 | else if (IsDecimal(schema->format)) |
4249 | 0 | { |
4250 | 0 | int nPrecision = 0; |
4251 | 0 | int nScale = 0; |
4252 | 0 | int nWidthInBytes = 0; |
4253 | 0 | const bool bOK = ParseDecimalFormat(schema->format, nPrecision, nScale, |
4254 | 0 | nWidthInBytes); |
4255 | | // Already validated |
4256 | 0 | CPLAssert(bOK); |
4257 | 0 | CPL_IGNORE_RET_VAL(bOK); |
4258 | 0 | oDict.Add(osKey, |
4259 | 0 | GetValueDecimal(array, nWidthInBytes / 8, nScale, nIdx)); |
4260 | 0 | } |
4261 | 0 | else |
4262 | 0 | oDict.Add(osKey, GetObjectAsJSON(schema, array, nIdx)); |
4263 | 0 | } |
4264 | | |
4265 | | /************************************************************************/ |
4266 | | /* GetMapAsJSON() */ |
4267 | | /************************************************************************/ |
4268 | | |
4269 | | static CPLJSONObject GetMapAsJSON(const struct ArrowSchema *schema, |
4270 | | const struct ArrowArray *array, |
4271 | | const size_t nIdx) |
4272 | 0 | { |
4273 | 0 | const auto schemaStruct = schema->children[0]; |
4274 | 0 | if (!IsStructure(schemaStruct->format)) |
4275 | 0 | { |
4276 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
4277 | 0 | "GetMapAsJSON(): !IsStructure(schemaStruct->format))"); |
4278 | 0 | return CPLJSONObject(); |
4279 | 0 | } |
4280 | 0 | const auto schemaKey = schemaStruct->children[0]; |
4281 | 0 | const auto schemaValues = schemaStruct->children[1]; |
4282 | 0 | if (!IsString(schemaKey->format)) |
4283 | 0 | { |
4284 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
4285 | 0 | "GetMapAsJSON(): !IsString(schemaKey->format))"); |
4286 | 0 | return CPLJSONObject(); |
4287 | 0 | } |
4288 | 0 | const auto arrayKeys = array->children[0]->children[0]; |
4289 | 0 | const auto arrayValues = array->children[0]->children[1]; |
4290 | |
|
4291 | 0 | CPLJSONObject oDict; |
4292 | 0 | const auto panOffsets = |
4293 | 0 | static_cast<const uint32_t *>(array->buffers[1]) + array->offset + nIdx; |
4294 | 0 | const uint8_t *pabyValidityKeys = |
4295 | 0 | arrayKeys->null_count == 0 |
4296 | 0 | ? nullptr |
4297 | 0 | : static_cast<const uint8_t *>(arrayKeys->buffers[0]); |
4298 | 0 | const uint32_t *panOffsetsKeys = |
4299 | 0 | static_cast<const uint32_t *>(arrayKeys->buffers[1]) + |
4300 | 0 | arrayKeys->offset; |
4301 | 0 | const char *pabyKeys = static_cast<const char *>(arrayKeys->buffers[2]); |
4302 | 0 | const uint8_t *pabyValidityValues = |
4303 | 0 | arrayValues->null_count == 0 |
4304 | 0 | ? nullptr |
4305 | 0 | : static_cast<const uint8_t *>(arrayValues->buffers[0]); |
4306 | 0 | for (uint32_t k = panOffsets[0]; k < panOffsets[1]; k++) |
4307 | 0 | { |
4308 | 0 | if (!pabyValidityKeys || |
4309 | 0 | TestBit(pabyValidityKeys, |
4310 | 0 | k + static_cast<size_t>(arrayKeys->offset))) |
4311 | 0 | { |
4312 | 0 | std::string osKey; |
4313 | 0 | osKey.assign(pabyKeys + panOffsetsKeys[k], |
4314 | 0 | panOffsetsKeys[k + 1] - panOffsetsKeys[k]); |
4315 | |
|
4316 | 0 | if (!pabyValidityValues || |
4317 | 0 | TestBit(pabyValidityValues, |
4318 | 0 | k + static_cast<size_t>(arrayValues->offset))) |
4319 | 0 | { |
4320 | 0 | AddToDict(oDict, osKey, schemaValues, arrayValues, k); |
4321 | 0 | } |
4322 | 0 | else |
4323 | 0 | { |
4324 | 0 | oDict.AddNull(osKey); |
4325 | 0 | } |
4326 | 0 | } |
4327 | 0 | } |
4328 | 0 | return oDict; |
4329 | 0 | } |
4330 | | |
4331 | | /************************************************************************/ |
4332 | | /* GetStructureAsJSON() */ |
4333 | | /************************************************************************/ |
4334 | | |
4335 | | static CPLJSONObject GetStructureAsJSON(const struct ArrowSchema *schema, |
4336 | | const struct ArrowArray *array, |
4337 | | const size_t nIdx) |
4338 | 0 | { |
4339 | 0 | CPLJSONObject oDict; |
4340 | 0 | for (int64_t k = 0; k < array->n_children; k++) |
4341 | 0 | { |
4342 | 0 | const uint8_t *pabyValidityValues = |
4343 | 0 | array->children[k]->null_count == 0 |
4344 | 0 | ? nullptr |
4345 | 0 | : static_cast<const uint8_t *>(array->children[k]->buffers[0]); |
4346 | 0 | if (!pabyValidityValues || |
4347 | 0 | TestBit(pabyValidityValues, |
4348 | 0 | nIdx + static_cast<size_t>(array->children[k]->offset))) |
4349 | 0 | { |
4350 | 0 | AddToDict(oDict, schema->children[k]->name, schema->children[k], |
4351 | 0 | array->children[k], nIdx); |
4352 | 0 | } |
4353 | 0 | else |
4354 | 0 | { |
4355 | 0 | oDict.AddNull(schema->children[k]->name); |
4356 | 0 | } |
4357 | 0 | } |
4358 | 0 | return oDict; |
4359 | 0 | } |
4360 | | |
4361 | | /************************************************************************/ |
4362 | | /* GetObjectAsJSON() */ |
4363 | | /************************************************************************/ |
4364 | | |
4365 | | static CPLJSONObject GetObjectAsJSON(const struct ArrowSchema *schema, |
4366 | | const struct ArrowArray *array, |
4367 | | const size_t nIdx) |
4368 | 0 | { |
4369 | 0 | if (IsMap(schema->format)) |
4370 | 0 | return GetMapAsJSON(schema, array, nIdx); |
4371 | 0 | else if (IsList(schema->format)) |
4372 | 0 | return GetListAsJSON<uint32_t>(schema, array, nIdx); |
4373 | 0 | else if (IsLargeList(schema->format)) |
4374 | 0 | return GetListAsJSON<uint64_t>(schema, array, nIdx); |
4375 | 0 | else if (IsFixedSizeList(schema->format)) |
4376 | 0 | return GetFixedSizeListAsJSON(schema, array, nIdx); |
4377 | 0 | else if (IsStructure(schema->format)) |
4378 | 0 | return GetStructureAsJSON(schema, array, nIdx); |
4379 | 0 | else |
4380 | 0 | { |
4381 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
4382 | 0 | "GetObjectAsJSON(): unhandled value format: %s", |
4383 | 0 | schema->format); |
4384 | 0 | return CPLJSONObject(); |
4385 | 0 | } |
4386 | 0 | } |
4387 | | |
4388 | | /************************************************************************/ |
4389 | | /* SetFieldForOtherFormats() */ |
4390 | | /************************************************************************/ |
4391 | | |
4392 | | static bool SetFieldForOtherFormats(OGRFeature &oFeature, |
4393 | | const int iOGRFieldIndex, |
4394 | | const size_t nOffsettedIndex, |
4395 | | const struct ArrowSchema *schema, |
4396 | | const struct ArrowArray *array) |
4397 | 0 | { |
4398 | 0 | const char *format = schema->format; |
4399 | 0 | if (IsFloat16(format)) |
4400 | 0 | { |
4401 | 0 | oFeature.SetField( |
4402 | 0 | iOGRFieldIndex, |
4403 | 0 | GetValueFloat16(array, nOffsettedIndex - |
4404 | 0 | static_cast<size_t>(array->offset))); |
4405 | 0 | } |
4406 | | |
4407 | 0 | else if (IsFixedWidthBinary(format)) |
4408 | 0 | { |
4409 | | // Fixed width binary |
4410 | 0 | const int nWidth = GetFixedWithBinary(format); |
4411 | 0 | oFeature.SetField(iOGRFieldIndex, nWidth, |
4412 | 0 | static_cast<const GByte *>(array->buffers[1]) + |
4413 | 0 | nOffsettedIndex * nWidth); |
4414 | 0 | } |
4415 | 0 | else if (format[0] == 't' && format[1] == 'd' && |
4416 | 0 | format[2] == 'D') // strcmp(format, "tdD") == 0 |
4417 | 0 | { |
4418 | | // date32[days] |
4419 | | // number of days since Epoch |
4420 | 0 | int64_t timestamp = static_cast<int64_t>(static_cast<const int32_t *>( |
4421 | 0 | array->buffers[1])[nOffsettedIndex]) * |
4422 | 0 | 3600 * 24; |
4423 | 0 | struct tm dt; |
4424 | 0 | CPLUnixTimeToYMDHMS(timestamp, &dt); |
4425 | 0 | oFeature.SetField(iOGRFieldIndex, dt.tm_year + 1900, dt.tm_mon + 1, |
4426 | 0 | dt.tm_mday, 0, 0, 0); |
4427 | 0 | return true; |
4428 | 0 | } |
4429 | 0 | else if (format[0] == 't' && format[1] == 'd' && |
4430 | 0 | format[2] == 'm') // strcmp(format, "tdm") == 0 |
4431 | 0 | { |
4432 | | // date64[milliseconds] |
4433 | | // number of milliseconds since Epoch |
4434 | 0 | int64_t timestamp = |
4435 | 0 | static_cast<const int64_t *>(array->buffers[1])[nOffsettedIndex] / |
4436 | 0 | 1000; |
4437 | 0 | struct tm dt; |
4438 | 0 | CPLUnixTimeToYMDHMS(timestamp, &dt); |
4439 | 0 | oFeature.SetField(iOGRFieldIndex, dt.tm_year + 1900, dt.tm_mon + 1, |
4440 | 0 | dt.tm_mday, 0, 0, 0); |
4441 | 0 | } |
4442 | 0 | else if (format[0] == 't' && format[1] == 't' && |
4443 | 0 | format[2] == 's') // strcmp(format, "tts") == 0 |
4444 | 0 | { |
4445 | | // time32 [seconds] |
4446 | 0 | int32_t value = |
4447 | 0 | static_cast<const int32_t *>(array->buffers[1])[nOffsettedIndex]; |
4448 | 0 | const int nHour = value / 3600; |
4449 | 0 | const int nMinute = (value / 60) % 60; |
4450 | 0 | const int nSecond = value % 60; |
4451 | 0 | oFeature.SetField(iOGRFieldIndex, 0, 0, 0, nHour, nMinute, |
4452 | 0 | static_cast<float>(nSecond)); |
4453 | 0 | } |
4454 | 0 | else if (format[0] == 't' && format[1] == 't' && |
4455 | 0 | format[2] == 'm') // strcmp(format, "ttm") == 0 |
4456 | 0 | { |
4457 | | // time32 [milliseconds] |
4458 | 0 | int32_t value = |
4459 | 0 | static_cast<const int32_t *>(array->buffers[1])[nOffsettedIndex]; |
4460 | 0 | double floatingPart = (value % 1000) / 1e3; |
4461 | 0 | value /= 1000; |
4462 | 0 | const int nHour = value / 3600; |
4463 | 0 | const int nMinute = (value / 60) % 60; |
4464 | 0 | const int nSecond = value % 60; |
4465 | 0 | oFeature.SetField(iOGRFieldIndex, 0, 0, 0, nHour, nMinute, |
4466 | 0 | static_cast<float>(nSecond + floatingPart)); |
4467 | 0 | } |
4468 | 0 | else if (format[0] == 't' && format[1] == 't' && |
4469 | 0 | (format[2] == 'u' || // time64 [microseconds] |
4470 | 0 | format[2] == 'n')) // time64 [nanoseconds] |
4471 | 0 | { |
4472 | 0 | oFeature.SetField(iOGRFieldIndex, |
4473 | 0 | static_cast<GIntBig>(static_cast<const int64_t *>( |
4474 | 0 | array->buffers[1])[nOffsettedIndex])); |
4475 | 0 | } |
4476 | 0 | else if (IsTimestampSeconds(format)) |
4477 | 0 | { |
4478 | 0 | ArrowTimestampToOGRDateTime( |
4479 | 0 | static_cast<const int64_t *>(array->buffers[1])[nOffsettedIndex], 1, |
4480 | 0 | GetTimestampTimezone(format), oFeature, iOGRFieldIndex); |
4481 | 0 | } |
4482 | 0 | else if (IsTimestampMilliseconds(format)) |
4483 | 0 | { |
4484 | 0 | ArrowTimestampToOGRDateTime( |
4485 | 0 | static_cast<const int64_t *>(array->buffers[1])[nOffsettedIndex], |
4486 | 0 | 1000, GetTimestampTimezone(format), oFeature, iOGRFieldIndex); |
4487 | 0 | } |
4488 | 0 | else if (IsTimestampMicroseconds(format)) |
4489 | 0 | { |
4490 | 0 | ArrowTimestampToOGRDateTime( |
4491 | 0 | static_cast<const int64_t *>(array->buffers[1])[nOffsettedIndex], |
4492 | 0 | 1000 * 1000, GetTimestampTimezone(format), oFeature, |
4493 | 0 | iOGRFieldIndex); |
4494 | 0 | } |
4495 | 0 | else if (IsTimestampNanoseconds(format)) |
4496 | 0 | { |
4497 | 0 | ArrowTimestampToOGRDateTime( |
4498 | 0 | static_cast<const int64_t *>(array->buffers[1])[nOffsettedIndex], |
4499 | 0 | 1000 * 1000 * 1000, GetTimestampTimezone(format), oFeature, |
4500 | 0 | iOGRFieldIndex); |
4501 | 0 | } |
4502 | 0 | else if (IsFixedSizeList(format)) |
4503 | 0 | { |
4504 | 0 | const int nItems = GetFixedSizeList(format); |
4505 | 0 | const auto childArray = array->children[0]; |
4506 | 0 | const char *childFormat = schema->children[0]->format; |
4507 | 0 | if (IsBoolean(childFormat)) |
4508 | 0 | { |
4509 | 0 | std::vector<int> aValues; |
4510 | 0 | const auto *paValues = |
4511 | 0 | static_cast<const uint8_t *>(childArray->buffers[1]); |
4512 | 0 | for (int i = 0; i < nItems; ++i) |
4513 | 0 | { |
4514 | 0 | aValues.push_back( |
4515 | 0 | TestBit(paValues, |
4516 | 0 | static_cast<size_t>(childArray->offset + |
4517 | 0 | nOffsettedIndex * nItems + i)) |
4518 | 0 | ? 1 |
4519 | 0 | : 0); |
4520 | 0 | } |
4521 | 0 | oFeature.SetField(iOGRFieldIndex, static_cast<int>(aValues.size()), |
4522 | 0 | aValues.data()); |
4523 | 0 | } |
4524 | 0 | else if (IsInt8(childFormat)) |
4525 | 0 | { |
4526 | 0 | FillFieldFixedSizeList<int8_t, int>(array, iOGRFieldIndex, |
4527 | 0 | nOffsettedIndex, nItems, |
4528 | 0 | childArray, oFeature); |
4529 | 0 | } |
4530 | 0 | else if (IsUInt8(childFormat)) |
4531 | 0 | { |
4532 | 0 | FillFieldFixedSizeList<uint8_t, int>(array, iOGRFieldIndex, |
4533 | 0 | nOffsettedIndex, nItems, |
4534 | 0 | childArray, oFeature); |
4535 | 0 | } |
4536 | 0 | else if (IsInt16(childFormat)) |
4537 | 0 | { |
4538 | 0 | FillFieldFixedSizeList<int16_t, int>(array, iOGRFieldIndex, |
4539 | 0 | nOffsettedIndex, nItems, |
4540 | 0 | childArray, oFeature); |
4541 | 0 | } |
4542 | 0 | else if (IsUInt16(childFormat)) |
4543 | 0 | { |
4544 | 0 | FillFieldFixedSizeList<uint16_t, int>(array, iOGRFieldIndex, |
4545 | 0 | nOffsettedIndex, nItems, |
4546 | 0 | childArray, oFeature); |
4547 | 0 | } |
4548 | 0 | else if (IsInt32(childFormat)) |
4549 | 0 | { |
4550 | 0 | FillFieldFixedSizeList<int32_t, int>(array, iOGRFieldIndex, |
4551 | 0 | nOffsettedIndex, nItems, |
4552 | 0 | childArray, oFeature); |
4553 | 0 | } |
4554 | 0 | else if (IsUInt32(childFormat)) |
4555 | 0 | { |
4556 | 0 | FillFieldFixedSizeList<uint32_t, GIntBig>(array, iOGRFieldIndex, |
4557 | 0 | nOffsettedIndex, nItems, |
4558 | 0 | childArray, oFeature); |
4559 | 0 | } |
4560 | 0 | else if (IsInt64(childFormat)) |
4561 | 0 | { |
4562 | 0 | FillFieldFixedSizeList<int64_t, GIntBig>(array, iOGRFieldIndex, |
4563 | 0 | nOffsettedIndex, nItems, |
4564 | 0 | childArray, oFeature); |
4565 | 0 | } |
4566 | 0 | else if (IsUInt64(childFormat)) |
4567 | 0 | { |
4568 | 0 | FillFieldFixedSizeList<uint64_t, double>(array, iOGRFieldIndex, |
4569 | 0 | nOffsettedIndex, nItems, |
4570 | 0 | childArray, oFeature); |
4571 | 0 | } |
4572 | 0 | else if (IsFloat16(childFormat)) |
4573 | 0 | { |
4574 | 0 | std::vector<double> aValues; |
4575 | 0 | for (int i = 0; i < nItems; ++i) |
4576 | 0 | { |
4577 | 0 | aValues.push_back( |
4578 | 0 | GetValueFloat16(childArray, nOffsettedIndex * nItems + i)); |
4579 | 0 | } |
4580 | 0 | oFeature.SetField(iOGRFieldIndex, static_cast<int>(aValues.size()), |
4581 | 0 | aValues.data()); |
4582 | 0 | } |
4583 | 0 | else if (IsFloat32(childFormat)) |
4584 | 0 | { |
4585 | 0 | FillFieldFixedSizeList<float, double>(array, iOGRFieldIndex, |
4586 | 0 | nOffsettedIndex, nItems, |
4587 | 0 | childArray, oFeature); |
4588 | 0 | } |
4589 | 0 | else if (IsFloat64(childFormat)) |
4590 | 0 | { |
4591 | 0 | FillFieldFixedSizeList<double, double>(array, iOGRFieldIndex, |
4592 | 0 | nOffsettedIndex, nItems, |
4593 | 0 | childArray, oFeature); |
4594 | 0 | } |
4595 | 0 | else if (IsString(childFormat)) |
4596 | 0 | { |
4597 | 0 | FillFieldFixedSizeListString<uint32_t>(array, iOGRFieldIndex, |
4598 | 0 | nOffsettedIndex, nItems, |
4599 | 0 | childArray, oFeature); |
4600 | 0 | } |
4601 | 0 | else if (IsLargeString(childFormat)) |
4602 | 0 | { |
4603 | 0 | FillFieldFixedSizeListString<uint64_t>(array, iOGRFieldIndex, |
4604 | 0 | nOffsettedIndex, nItems, |
4605 | 0 | childArray, oFeature); |
4606 | 0 | } |
4607 | 0 | } |
4608 | 0 | else if (IsList(format) || IsLargeList(format)) |
4609 | 0 | { |
4610 | 0 | const auto childArray = array->children[0]; |
4611 | 0 | const char *childFormat = schema->children[0]->format; |
4612 | 0 | if (IsBoolean(childFormat)) |
4613 | 0 | { |
4614 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4615 | 0 | FillFieldListFromBool<uint32_t>(array, iOGRFieldIndex, |
4616 | 0 | nOffsettedIndex, childArray, |
4617 | 0 | oFeature); |
4618 | 0 | else |
4619 | 0 | FillFieldListFromBool<uint64_t>(array, iOGRFieldIndex, |
4620 | 0 | nOffsettedIndex, childArray, |
4621 | 0 | oFeature); |
4622 | 0 | } |
4623 | 0 | else if (IsInt8(childFormat)) |
4624 | 0 | { |
4625 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4626 | 0 | FillFieldList<uint32_t, int8_t, int>(array, iOGRFieldIndex, |
4627 | 0 | nOffsettedIndex, |
4628 | 0 | childArray, oFeature); |
4629 | 0 | else |
4630 | 0 | FillFieldList<uint64_t, int8_t, int>(array, iOGRFieldIndex, |
4631 | 0 | nOffsettedIndex, |
4632 | 0 | childArray, oFeature); |
4633 | 0 | } |
4634 | 0 | else if (IsUInt8(childFormat)) |
4635 | 0 | { |
4636 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4637 | 0 | FillFieldList<uint32_t, uint8_t, int>(array, iOGRFieldIndex, |
4638 | 0 | nOffsettedIndex, |
4639 | 0 | childArray, oFeature); |
4640 | 0 | else |
4641 | 0 | FillFieldList<uint64_t, uint8_t, int>(array, iOGRFieldIndex, |
4642 | 0 | nOffsettedIndex, |
4643 | 0 | childArray, oFeature); |
4644 | 0 | } |
4645 | 0 | else if (IsInt16(childFormat)) |
4646 | 0 | { |
4647 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4648 | 0 | FillFieldList<uint32_t, int16_t, int>(array, iOGRFieldIndex, |
4649 | 0 | nOffsettedIndex, |
4650 | 0 | childArray, oFeature); |
4651 | 0 | else |
4652 | 0 | FillFieldList<uint64_t, int16_t, int>(array, iOGRFieldIndex, |
4653 | 0 | nOffsettedIndex, |
4654 | 0 | childArray, oFeature); |
4655 | 0 | } |
4656 | 0 | else if (IsUInt16(childFormat)) |
4657 | 0 | { |
4658 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4659 | 0 | FillFieldList<uint32_t, uint16_t, int>(array, iOGRFieldIndex, |
4660 | 0 | nOffsettedIndex, |
4661 | 0 | childArray, oFeature); |
4662 | 0 | else |
4663 | 0 | FillFieldList<uint64_t, uint16_t, int>(array, iOGRFieldIndex, |
4664 | 0 | nOffsettedIndex, |
4665 | 0 | childArray, oFeature); |
4666 | 0 | } |
4667 | 0 | else if (IsInt32(childFormat)) |
4668 | 0 | { |
4669 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4670 | 0 | FillFieldList<uint32_t, int32_t, int>(array, iOGRFieldIndex, |
4671 | 0 | nOffsettedIndex, |
4672 | 0 | childArray, oFeature); |
4673 | 0 | else |
4674 | 0 | FillFieldList<uint64_t, int32_t, int>(array, iOGRFieldIndex, |
4675 | 0 | nOffsettedIndex, |
4676 | 0 | childArray, oFeature); |
4677 | 0 | } |
4678 | 0 | else if (IsUInt32(childFormat)) |
4679 | 0 | { |
4680 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4681 | 0 | FillFieldList<uint32_t, uint32_t, GIntBig>( |
4682 | 0 | array, iOGRFieldIndex, nOffsettedIndex, childArray, |
4683 | 0 | oFeature); |
4684 | 0 | else |
4685 | 0 | FillFieldList<uint64_t, uint32_t, GIntBig>( |
4686 | 0 | array, iOGRFieldIndex, nOffsettedIndex, childArray, |
4687 | 0 | oFeature); |
4688 | 0 | } |
4689 | 0 | else if (IsInt64(childFormat)) |
4690 | 0 | { |
4691 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4692 | 0 | FillFieldList<uint32_t, int64_t, GIntBig>(array, iOGRFieldIndex, |
4693 | 0 | nOffsettedIndex, |
4694 | 0 | childArray, oFeature); |
4695 | 0 | else |
4696 | 0 | FillFieldList<uint64_t, int64_t, GIntBig>(array, iOGRFieldIndex, |
4697 | 0 | nOffsettedIndex, |
4698 | 0 | childArray, oFeature); |
4699 | 0 | } |
4700 | 0 | else if (IsUInt64(childFormat)) // (lossy conversion) |
4701 | 0 | { |
4702 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4703 | 0 | FillFieldList<uint32_t, uint64_t, double>(array, iOGRFieldIndex, |
4704 | 0 | nOffsettedIndex, |
4705 | 0 | childArray, oFeature); |
4706 | 0 | else |
4707 | 0 | FillFieldList<uint64_t, uint64_t, double>(array, iOGRFieldIndex, |
4708 | 0 | nOffsettedIndex, |
4709 | 0 | childArray, oFeature); |
4710 | 0 | } |
4711 | 0 | else if (IsFloat16(childFormat)) |
4712 | 0 | { |
4713 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4714 | 0 | FillFieldListFromHalfFloat<uint32_t>(array, iOGRFieldIndex, |
4715 | 0 | nOffsettedIndex, |
4716 | 0 | childArray, oFeature); |
4717 | 0 | else |
4718 | 0 | FillFieldListFromHalfFloat<uint64_t>(array, iOGRFieldIndex, |
4719 | 0 | nOffsettedIndex, |
4720 | 0 | childArray, oFeature); |
4721 | 0 | } |
4722 | 0 | else if (IsFloat32(childFormat)) |
4723 | 0 | { |
4724 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4725 | 0 | FillFieldList<uint32_t, float, double>(array, iOGRFieldIndex, |
4726 | 0 | nOffsettedIndex, |
4727 | 0 | childArray, oFeature); |
4728 | 0 | else |
4729 | 0 | FillFieldList<uint64_t, float, double>(array, iOGRFieldIndex, |
4730 | 0 | nOffsettedIndex, |
4731 | 0 | childArray, oFeature); |
4732 | 0 | } |
4733 | 0 | else if (IsFloat64(childFormat)) |
4734 | 0 | { |
4735 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4736 | 0 | FillFieldList<uint32_t, double, double>(array, iOGRFieldIndex, |
4737 | 0 | nOffsettedIndex, |
4738 | 0 | childArray, oFeature); |
4739 | 0 | else |
4740 | 0 | FillFieldList<uint64_t, double, double>(array, iOGRFieldIndex, |
4741 | 0 | nOffsettedIndex, |
4742 | 0 | childArray, oFeature); |
4743 | 0 | } |
4744 | 0 | else if (IsString(childFormat)) |
4745 | 0 | { |
4746 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4747 | 0 | FillFieldListFromString<uint32_t, uint32_t>( |
4748 | 0 | array, iOGRFieldIndex, nOffsettedIndex, childArray, |
4749 | 0 | oFeature); |
4750 | 0 | else |
4751 | 0 | FillFieldListFromString<uint64_t, uint32_t>( |
4752 | 0 | array, iOGRFieldIndex, nOffsettedIndex, childArray, |
4753 | 0 | oFeature); |
4754 | 0 | } |
4755 | 0 | else if (IsLargeString(childFormat)) |
4756 | 0 | { |
4757 | 0 | if (format[1] == ARROW_2ND_LETTER_LIST) |
4758 | 0 | FillFieldListFromString<uint32_t, uint64_t>( |
4759 | 0 | array, iOGRFieldIndex, nOffsettedIndex, childArray, |
4760 | 0 | oFeature); |
4761 | 0 | else |
4762 | 0 | FillFieldListFromString<uint64_t, uint64_t>( |
4763 | 0 | array, iOGRFieldIndex, nOffsettedIndex, childArray, |
4764 | 0 | oFeature); |
4765 | 0 | } |
4766 | 0 | else if (format[1] == ARROW_2ND_LETTER_LIST) |
4767 | 0 | { |
4768 | 0 | const size_t iFeature = |
4769 | 0 | static_cast<size_t>(nOffsettedIndex - array->offset); |
4770 | 0 | oFeature.SetField(iOGRFieldIndex, |
4771 | 0 | GetListAsJSON<uint32_t>(schema, array, iFeature) |
4772 | 0 | .Format(CPLJSONObject::PrettyFormat::Plain) |
4773 | 0 | .c_str()); |
4774 | 0 | } |
4775 | 0 | else |
4776 | 0 | { |
4777 | 0 | const size_t iFeature = |
4778 | 0 | static_cast<size_t>(nOffsettedIndex - array->offset); |
4779 | 0 | oFeature.SetField(iOGRFieldIndex, |
4780 | 0 | GetListAsJSON<uint64_t>(schema, array, iFeature) |
4781 | 0 | .Format(CPLJSONObject::PrettyFormat::Plain) |
4782 | 0 | .c_str()); |
4783 | 0 | } |
4784 | 0 | } |
4785 | 0 | else if (IsDecimal(format)) |
4786 | 0 | { |
4787 | 0 | int nPrecision = 0; |
4788 | 0 | int nScale = 0; |
4789 | 0 | int nWidthInBytes = 0; |
4790 | 0 | if (!ParseDecimalFormat(format, nPrecision, nScale, nWidthInBytes)) |
4791 | 0 | { |
4792 | 0 | CPLAssert(false); |
4793 | 0 | } |
4794 | | |
4795 | | // fits on a int64 |
4796 | 0 | CPLAssert(nPrecision <= 19); |
4797 | | // either 128 or 256 bits |
4798 | 0 | CPLAssert((nWidthInBytes % 8) == 0); |
4799 | 0 | const int nWidthIn64BitWord = nWidthInBytes / 8; |
4800 | 0 | const size_t iFeature = |
4801 | 0 | static_cast<size_t>(nOffsettedIndex - array->offset); |
4802 | 0 | oFeature.SetField( |
4803 | 0 | iOGRFieldIndex, |
4804 | 0 | GetValueDecimal(array, nWidthIn64BitWord, nScale, iFeature)); |
4805 | 0 | return true; |
4806 | 0 | } |
4807 | 0 | else if (IsMap(format)) |
4808 | 0 | { |
4809 | 0 | const size_t iFeature = |
4810 | 0 | static_cast<size_t>(nOffsettedIndex - array->offset); |
4811 | 0 | oFeature.SetField(iOGRFieldIndex, |
4812 | 0 | GetMapAsJSON(schema, array, iFeature) |
4813 | 0 | .Format(CPLJSONObject::PrettyFormat::Plain) |
4814 | 0 | .c_str()); |
4815 | 0 | } |
4816 | 0 | else |
4817 | 0 | { |
4818 | 0 | return false; |
4819 | 0 | } |
4820 | | |
4821 | 0 | return true; |
4822 | 0 | } |
4823 | | |
4824 | | /************************************************************************/ |
4825 | | /* FillValidityArrayFromAttrQuery() */ |
4826 | | /************************************************************************/ |
4827 | | |
4828 | | static size_t FillValidityArrayFromAttrQuery( |
4829 | | const OGRLayer *poLayer, OGRFeatureQuery *poAttrQuery, |
4830 | | const struct ArrowSchema *schema, struct ArrowArray *array, |
4831 | | std::vector<bool> &abyValidityFromFilters, CSLConstList papszOptions) |
4832 | 0 | { |
4833 | 0 | size_t nCountIntersecting = 0; |
4834 | 0 | auto poFeatureDefn = const_cast<OGRLayer *>(poLayer)->GetLayerDefn(); |
4835 | 0 | OGRFeature oFeature(poFeatureDefn); |
4836 | |
|
4837 | 0 | std::map<std::string, std::vector<int>> oMapFieldNameToArrowPath; |
4838 | 0 | std::vector<int> anArrowPathTmp; |
4839 | 0 | BuildMapFieldNameToArrowPath(schema, oMapFieldNameToArrowPath, |
4840 | 0 | std::string(), anArrowPathTmp); |
4841 | |
|
4842 | 0 | struct UsedFieldsInfo |
4843 | 0 | { |
4844 | 0 | int iOGRFieldIndex{}; |
4845 | 0 | std::vector<int> anArrowPath{}; |
4846 | 0 | }; |
4847 | |
|
4848 | 0 | std::vector<UsedFieldsInfo> aoUsedFieldsInfo; |
4849 | |
|
4850 | 0 | bool bNeedsFID = false; |
4851 | 0 | const CPLStringList aosUsedFields(poAttrQuery->GetUsedFields()); |
4852 | 0 | for (int i = 0; i < aosUsedFields.size(); ++i) |
4853 | 0 | { |
4854 | 0 | int iOGRFieldIndex = poFeatureDefn->GetFieldIndex(aosUsedFields[i]); |
4855 | 0 | if (iOGRFieldIndex >= 0) |
4856 | 0 | { |
4857 | 0 | const auto oIter = oMapFieldNameToArrowPath.find(aosUsedFields[i]); |
4858 | 0 | if (oIter != oMapFieldNameToArrowPath.end()) |
4859 | 0 | { |
4860 | 0 | UsedFieldsInfo info; |
4861 | 0 | info.iOGRFieldIndex = iOGRFieldIndex; |
4862 | 0 | info.anArrowPath = oIter->second; |
4863 | 0 | aoUsedFieldsInfo.push_back(std::move(info)); |
4864 | 0 | } |
4865 | 0 | else |
4866 | 0 | { |
4867 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
4868 | 0 | "Cannot find %s in oMapFieldNameToArrowPath", |
4869 | 0 | aosUsedFields[i]); |
4870 | 0 | } |
4871 | 0 | } |
4872 | 0 | else if (EQUAL(aosUsedFields[i], "FID")) |
4873 | 0 | { |
4874 | 0 | bNeedsFID = true; |
4875 | 0 | } |
4876 | 0 | else |
4877 | 0 | { |
4878 | 0 | CPLDebug("OGR", "Cannot find used field %s", aosUsedFields[i]); |
4879 | 0 | } |
4880 | 0 | } |
4881 | |
|
4882 | 0 | const size_t nLength = abyValidityFromFilters.size(); |
4883 | |
|
4884 | 0 | GIntBig nBaseSeqFID = -1; |
4885 | 0 | std::vector<int> anArrowPathToFIDColumn; |
4886 | 0 | if (bNeedsFID) |
4887 | 0 | { |
4888 | | // BASE_SEQUENTIAL_FID is set when there is no Arrow column for the FID |
4889 | | // and we assume sequential FID numbering |
4890 | 0 | const char *pszBaseSeqFID = |
4891 | 0 | CSLFetchNameValue(papszOptions, "BASE_SEQUENTIAL_FID"); |
4892 | 0 | if (pszBaseSeqFID) |
4893 | 0 | { |
4894 | 0 | nBaseSeqFID = CPLAtoGIntBig(pszBaseSeqFID); |
4895 | | |
4896 | | // Optimizimation for "FID = constant" |
4897 | 0 | swq_expr_node *poNode = |
4898 | 0 | static_cast<swq_expr_node *>(poAttrQuery->GetSWQExpr()); |
4899 | 0 | if (poNode->eNodeType == SNT_OPERATION && |
4900 | 0 | poNode->nOperation == SWQ_EQ && poNode->nSubExprCount == 2 && |
4901 | 0 | poNode->papoSubExpr[0]->eNodeType == SNT_COLUMN && |
4902 | 0 | poNode->papoSubExpr[1]->eNodeType == SNT_CONSTANT && |
4903 | 0 | poNode->papoSubExpr[0]->field_index == |
4904 | 0 | poFeatureDefn->GetFieldCount() + SPF_FID && |
4905 | 0 | poNode->papoSubExpr[1]->field_type == SWQ_INTEGER64) |
4906 | 0 | { |
4907 | 0 | if (nBaseSeqFID + static_cast<int64_t>(nLength) < |
4908 | 0 | poNode->papoSubExpr[1]->int_value || |
4909 | 0 | nBaseSeqFID > poNode->papoSubExpr[1]->int_value) |
4910 | 0 | { |
4911 | 0 | return 0; |
4912 | 0 | } |
4913 | 0 | } |
4914 | 0 | } |
4915 | 0 | else |
4916 | 0 | { |
4917 | 0 | const char *pszFIDColumn = |
4918 | 0 | const_cast<OGRLayer *>(poLayer)->GetFIDColumn(); |
4919 | 0 | if (pszFIDColumn && pszFIDColumn[0]) |
4920 | 0 | { |
4921 | 0 | const auto oIter = oMapFieldNameToArrowPath.find(pszFIDColumn); |
4922 | 0 | if (oIter != oMapFieldNameToArrowPath.end()) |
4923 | 0 | { |
4924 | 0 | anArrowPathToFIDColumn = oIter->second; |
4925 | 0 | } |
4926 | 0 | } |
4927 | 0 | if (anArrowPathToFIDColumn.empty()) |
4928 | 0 | { |
4929 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
4930 | 0 | "Filtering on FID requested but cannot associate a " |
4931 | 0 | "FID with Arrow records"); |
4932 | 0 | } |
4933 | 0 | } |
4934 | 0 | } |
4935 | | |
4936 | 0 | for (size_t iRow = 0; iRow < nLength; ++iRow) |
4937 | 0 | { |
4938 | 0 | if (!abyValidityFromFilters[iRow]) |
4939 | 0 | continue; |
4940 | | |
4941 | 0 | if (bNeedsFID) |
4942 | 0 | { |
4943 | 0 | if (nBaseSeqFID >= 0) |
4944 | 0 | { |
4945 | 0 | oFeature.SetFID(nBaseSeqFID + iRow); |
4946 | 0 | } |
4947 | 0 | else if (!anArrowPathToFIDColumn.empty()) |
4948 | 0 | { |
4949 | 0 | oFeature.SetFID(OGRNullFID); |
4950 | |
|
4951 | 0 | const struct ArrowSchema *psSchemaField = schema; |
4952 | 0 | const struct ArrowArray *psArray = array; |
4953 | 0 | bool bSkip = false; |
4954 | 0 | for (size_t i = 0; i < anArrowPathToFIDColumn.size(); ++i) |
4955 | 0 | { |
4956 | 0 | const int iChild = anArrowPathToFIDColumn[i]; |
4957 | 0 | if (i > 0) |
4958 | 0 | { |
4959 | 0 | const uint8_t *pabyValidity = |
4960 | 0 | psArray->null_count == 0 |
4961 | 0 | ? nullptr |
4962 | 0 | : static_cast<uint8_t *>( |
4963 | 0 | const_cast<void *>(psArray->buffers[0])); |
4964 | 0 | const size_t nOffsettedIndex = |
4965 | 0 | static_cast<size_t>(iRow + psArray->offset); |
4966 | 0 | if (pabyValidity && |
4967 | 0 | !TestBit(pabyValidity, nOffsettedIndex)) |
4968 | 0 | { |
4969 | 0 | bSkip = true; |
4970 | 0 | break; |
4971 | 0 | } |
4972 | 0 | } |
4973 | | |
4974 | 0 | psSchemaField = psSchemaField->children[iChild]; |
4975 | 0 | psArray = psArray->children[iChild]; |
4976 | 0 | } |
4977 | 0 | if (bSkip) |
4978 | 0 | continue; |
4979 | | |
4980 | 0 | const char *format = psSchemaField->format; |
4981 | 0 | const uint8_t *pabyValidity = |
4982 | 0 | psArray->null_count == 0 |
4983 | 0 | ? nullptr |
4984 | 0 | : static_cast<uint8_t *>( |
4985 | 0 | const_cast<void *>(psArray->buffers[0])); |
4986 | 0 | const size_t nOffsettedIndex = |
4987 | 0 | static_cast<size_t>(iRow + psArray->offset); |
4988 | 0 | if (pabyValidity && !TestBit(pabyValidity, nOffsettedIndex)) |
4989 | 0 | { |
4990 | | // do nothing |
4991 | 0 | } |
4992 | 0 | else if (IsInt32(format)) |
4993 | 0 | { |
4994 | 0 | oFeature.SetFID(static_cast<const int32_t *>( |
4995 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
4996 | 0 | } |
4997 | 0 | else if (IsInt64(format)) |
4998 | 0 | { |
4999 | 0 | oFeature.SetFID(static_cast<const int64_t *>( |
5000 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
5001 | 0 | } |
5002 | 0 | } |
5003 | 0 | } |
5004 | | |
5005 | 0 | for (const auto &sInfo : aoUsedFieldsInfo) |
5006 | 0 | { |
5007 | 0 | const int iOGRFieldIndex = sInfo.iOGRFieldIndex; |
5008 | 0 | const struct ArrowSchema *psSchemaField = schema; |
5009 | 0 | const struct ArrowArray *psArray = array; |
5010 | 0 | bool bSkip = false; |
5011 | 0 | for (size_t i = 0; i < sInfo.anArrowPath.size(); ++i) |
5012 | 0 | { |
5013 | 0 | const int iChild = sInfo.anArrowPath[i]; |
5014 | 0 | if (i > 0) |
5015 | 0 | { |
5016 | 0 | const uint8_t *pabyValidity = |
5017 | 0 | psArray->null_count == 0 |
5018 | 0 | ? nullptr |
5019 | 0 | : static_cast<uint8_t *>( |
5020 | 0 | const_cast<void *>(psArray->buffers[0])); |
5021 | 0 | const size_t nOffsettedIndex = |
5022 | 0 | static_cast<size_t>(iRow + psArray->offset); |
5023 | 0 | if (pabyValidity && !TestBit(pabyValidity, nOffsettedIndex)) |
5024 | 0 | { |
5025 | 0 | bSkip = true; |
5026 | 0 | oFeature.SetFieldNull(iOGRFieldIndex); |
5027 | 0 | break; |
5028 | 0 | } |
5029 | 0 | } |
5030 | | |
5031 | 0 | psSchemaField = psSchemaField->children[iChild]; |
5032 | 0 | psArray = psArray->children[iChild]; |
5033 | 0 | } |
5034 | 0 | if (bSkip) |
5035 | 0 | continue; |
5036 | | |
5037 | 0 | const char *format = psSchemaField->format; |
5038 | 0 | const uint8_t *pabyValidity = |
5039 | 0 | psArray->null_count == 0 |
5040 | 0 | ? nullptr |
5041 | 0 | : static_cast<uint8_t *>( |
5042 | 0 | const_cast<void *>(psArray->buffers[0])); |
5043 | 0 | const size_t nOffsettedIndex = |
5044 | 0 | static_cast<size_t>(iRow + psArray->offset); |
5045 | 0 | if (pabyValidity && !TestBit(pabyValidity, nOffsettedIndex)) |
5046 | 0 | { |
5047 | 0 | oFeature.SetFieldNull(iOGRFieldIndex); |
5048 | 0 | } |
5049 | 0 | else if (IsBoolean(format)) |
5050 | 0 | { |
5051 | 0 | oFeature.SetField( |
5052 | 0 | iOGRFieldIndex, |
5053 | 0 | TestBit(static_cast<const uint8_t *>(psArray->buffers[1]), |
5054 | 0 | nOffsettedIndex)); |
5055 | 0 | } |
5056 | 0 | else if (IsInt8(format)) |
5057 | 0 | { |
5058 | 0 | oFeature.SetField(iOGRFieldIndex, |
5059 | 0 | static_cast<const int8_t *>( |
5060 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
5061 | 0 | } |
5062 | 0 | else if (IsUInt8(format)) |
5063 | 0 | { |
5064 | 0 | oFeature.SetField(iOGRFieldIndex, |
5065 | 0 | static_cast<const uint8_t *>( |
5066 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
5067 | 0 | } |
5068 | 0 | else if (IsInt16(format)) |
5069 | 0 | { |
5070 | 0 | oFeature.SetField(iOGRFieldIndex, |
5071 | 0 | static_cast<const int16_t *>( |
5072 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
5073 | 0 | } |
5074 | 0 | else if (IsUInt16(format)) |
5075 | 0 | { |
5076 | 0 | oFeature.SetField(iOGRFieldIndex, |
5077 | 0 | static_cast<const uint16_t *>( |
5078 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
5079 | 0 | } |
5080 | 0 | else if (IsInt32(format)) |
5081 | 0 | { |
5082 | 0 | oFeature.SetField(iOGRFieldIndex, |
5083 | 0 | static_cast<const int32_t *>( |
5084 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
5085 | 0 | } |
5086 | 0 | else if (IsUInt32(format)) |
5087 | 0 | { |
5088 | 0 | oFeature.SetField( |
5089 | 0 | iOGRFieldIndex, |
5090 | 0 | static_cast<GIntBig>(static_cast<const uint32_t *>( |
5091 | 0 | psArray->buffers[1])[nOffsettedIndex])); |
5092 | 0 | } |
5093 | 0 | else if (IsInt64(format)) |
5094 | 0 | { |
5095 | 0 | oFeature.SetField( |
5096 | 0 | iOGRFieldIndex, |
5097 | 0 | static_cast<GIntBig>(static_cast<const int64_t *>( |
5098 | 0 | psArray->buffers[1])[nOffsettedIndex])); |
5099 | 0 | } |
5100 | 0 | else if (IsUInt64(format)) |
5101 | 0 | { |
5102 | 0 | oFeature.SetField( |
5103 | 0 | iOGRFieldIndex, |
5104 | 0 | static_cast<double>(static_cast<const uint64_t *>( |
5105 | 0 | psArray->buffers[1])[nOffsettedIndex])); |
5106 | 0 | } |
5107 | 0 | else if (IsFloat32(format)) |
5108 | 0 | { |
5109 | 0 | oFeature.SetField(iOGRFieldIndex, |
5110 | 0 | static_cast<const float *>( |
5111 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
5112 | 0 | } |
5113 | 0 | else if (IsFloat64(format)) |
5114 | 0 | { |
5115 | 0 | oFeature.SetField(iOGRFieldIndex, |
5116 | 0 | static_cast<const double *>( |
5117 | 0 | psArray->buffers[1])[nOffsettedIndex]); |
5118 | 0 | } |
5119 | 0 | else if (IsString(format)) |
5120 | 0 | { |
5121 | 0 | const auto nOffset = static_cast<const uint32_t *>( |
5122 | 0 | psArray->buffers[1])[nOffsettedIndex]; |
5123 | 0 | const auto nNextOffset = static_cast<const uint32_t *>( |
5124 | 0 | psArray->buffers[1])[nOffsettedIndex + 1]; |
5125 | 0 | const GByte *pabyData = |
5126 | 0 | static_cast<const GByte *>(psArray->buffers[2]); |
5127 | 0 | const uint32_t nSize = nNextOffset - nOffset; |
5128 | 0 | CPLAssert(oFeature.GetFieldDefnRef(iOGRFieldIndex)->GetType() == |
5129 | 0 | OFTString); |
5130 | 0 | char *pszStr = static_cast<char *>(CPLMalloc(nSize + 1)); |
5131 | 0 | memcpy(pszStr, pabyData + nOffset, nSize); |
5132 | 0 | pszStr[nSize] = 0; |
5133 | 0 | OGRField *psField = oFeature.GetRawFieldRef(iOGRFieldIndex); |
5134 | 0 | if (IsValidField(psField)) |
5135 | 0 | CPLFree(psField->String); |
5136 | 0 | psField->String = pszStr; |
5137 | 0 | } |
5138 | 0 | else if (IsLargeString(format)) |
5139 | 0 | { |
5140 | 0 | const auto nOffset = static_cast<const uint64_t *>( |
5141 | 0 | psArray->buffers[1])[nOffsettedIndex]; |
5142 | 0 | const auto nNextOffset = static_cast<const uint64_t *>( |
5143 | 0 | psArray->buffers[1])[nOffsettedIndex + 1]; |
5144 | 0 | const GByte *pabyData = |
5145 | 0 | static_cast<const GByte *>(psArray->buffers[2]); |
5146 | 0 | const size_t nSize = static_cast<size_t>(nNextOffset - nOffset); |
5147 | 0 | char *pszStr = static_cast<char *>(CPLMalloc(nSize + 1)); |
5148 | 0 | memcpy(pszStr, pabyData + static_cast<size_t>(nOffset), nSize); |
5149 | 0 | pszStr[nSize] = 0; |
5150 | 0 | OGRField *psField = oFeature.GetRawFieldRef(iOGRFieldIndex); |
5151 | 0 | if (IsValidField(psField)) |
5152 | 0 | CPLFree(psField->String); |
5153 | 0 | psField->String = pszStr; |
5154 | 0 | } |
5155 | 0 | else if (IsBinary(format)) |
5156 | 0 | { |
5157 | 0 | const auto nOffset = static_cast<const uint32_t *>( |
5158 | 0 | psArray->buffers[1])[nOffsettedIndex]; |
5159 | 0 | const auto nNextOffset = static_cast<const uint32_t *>( |
5160 | 0 | psArray->buffers[1])[nOffsettedIndex + 1]; |
5161 | 0 | const GByte *pabyData = |
5162 | 0 | static_cast<const GByte *>(psArray->buffers[2]); |
5163 | 0 | const uint32_t nSize = nNextOffset - nOffset; |
5164 | 0 | if (nSize > |
5165 | 0 | static_cast<size_t>(std::numeric_limits<int32_t>::max())) |
5166 | 0 | { |
5167 | 0 | abyValidityFromFilters.clear(); |
5168 | 0 | abyValidityFromFilters.resize(nLength); |
5169 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
5170 | 0 | "Unexpected error in PostFilterArrowArray(): too " |
5171 | 0 | "large binary"); |
5172 | 0 | return 0; |
5173 | 0 | } |
5174 | 0 | oFeature.SetField(iOGRFieldIndex, static_cast<int>(nSize), |
5175 | 0 | pabyData + nOffset); |
5176 | 0 | } |
5177 | 0 | else if (IsLargeBinary(format)) |
5178 | 0 | { |
5179 | 0 | const auto nOffset = static_cast<const uint64_t *>( |
5180 | 0 | psArray->buffers[1])[nOffsettedIndex]; |
5181 | 0 | const auto nNextOffset = static_cast<const uint64_t *>( |
5182 | 0 | psArray->buffers[1])[nOffsettedIndex + 1]; |
5183 | 0 | const GByte *pabyData = |
5184 | 0 | static_cast<const GByte *>(psArray->buffers[2]); |
5185 | 0 | const uint64_t nSize = nNextOffset - nOffset; |
5186 | 0 | if (nSize > |
5187 | 0 | static_cast<uint64_t>(std::numeric_limits<int32_t>::max())) |
5188 | 0 | { |
5189 | 0 | abyValidityFromFilters.clear(); |
5190 | 0 | abyValidityFromFilters.resize(nLength); |
5191 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
5192 | 0 | "Unexpected error in PostFilterArrowArray(): too " |
5193 | 0 | "large binary"); |
5194 | 0 | return 0; |
5195 | 0 | } |
5196 | 0 | oFeature.SetField(iOGRFieldIndex, static_cast<int>(nSize), |
5197 | 0 | pabyData + nOffset); |
5198 | 0 | } |
5199 | 0 | else if (!SetFieldForOtherFormats(oFeature, iOGRFieldIndex, |
5200 | 0 | nOffsettedIndex, psSchemaField, |
5201 | 0 | psArray)) |
5202 | 0 | { |
5203 | 0 | abyValidityFromFilters.clear(); |
5204 | 0 | abyValidityFromFilters.resize(nLength); |
5205 | 0 | CPLError( |
5206 | 0 | CE_Failure, CPLE_AppDefined, |
5207 | 0 | "Unexpected error in PostFilterArrowArray(): unhandled " |
5208 | 0 | "field format: %s", |
5209 | 0 | format); |
5210 | 0 | return 0; |
5211 | 0 | } |
5212 | 0 | } |
5213 | 0 | if (poAttrQuery->Evaluate(&oFeature)) |
5214 | 0 | { |
5215 | 0 | nCountIntersecting++; |
5216 | 0 | } |
5217 | 0 | else |
5218 | 0 | { |
5219 | 0 | abyValidityFromFilters[iRow] = false; |
5220 | 0 | } |
5221 | 0 | } |
5222 | 0 | return nCountIntersecting; |
5223 | 0 | } |
5224 | | |
5225 | | /************************************************************************/ |
5226 | | /* OGRLayer::PostFilterArrowArray() */ |
5227 | | /************************************************************************/ |
5228 | | |
5229 | | /** Remove rows that aren't selected by the spatial or attribute filter. |
5230 | | * |
5231 | | * Assumes that CanPostFilterArrowArray() has been called and returned true. |
5232 | | */ |
5233 | | void OGRLayer::PostFilterArrowArray(const struct ArrowSchema *schema, |
5234 | | struct ArrowArray *array, |
5235 | | CSLConstList papszOptions) const |
5236 | 0 | { |
5237 | 0 | if (!m_poFilterGeom && !m_poAttrQuery) |
5238 | 0 | return; |
5239 | | |
5240 | 0 | CPLAssert(schema->n_children == array->n_children); |
5241 | | |
5242 | 0 | int64_t iGeomField = -1; |
5243 | 0 | if (m_poFilterGeom) |
5244 | 0 | { |
5245 | 0 | const char *pszGeomFieldName = |
5246 | 0 | const_cast<OGRLayer *>(this) |
5247 | 0 | ->GetLayerDefn() |
5248 | 0 | ->GetGeomFieldDefn(m_iGeomFieldFilter) |
5249 | 0 | ->GetNameRef(); |
5250 | 0 | for (int64_t iField = 0; iField < schema->n_children; ++iField) |
5251 | 0 | { |
5252 | 0 | const auto fieldSchema = schema->children[iField]; |
5253 | 0 | if (strcmp(fieldSchema->name, pszGeomFieldName) == 0) |
5254 | 0 | { |
5255 | 0 | iGeomField = iField; |
5256 | 0 | break; |
5257 | 0 | } |
5258 | 0 | CPLAssert(array->children[iField]->length == |
5259 | 0 | array->children[0]->length); |
5260 | 0 | } |
5261 | | // Guaranteed if CanPostFilterArrowArray() returned true |
5262 | 0 | CPLAssert(iGeomField >= 0); |
5263 | 0 | CPLAssert(IsBinary(schema->children[iGeomField]->format) || |
5264 | 0 | IsLargeBinary(schema->children[iGeomField]->format)); |
5265 | 0 | CPLAssert(array->children[iGeomField]->n_buffers == 3); |
5266 | 0 | } |
5267 | | |
5268 | 0 | std::vector<bool> abyValidityFromFilters; |
5269 | 0 | const size_t nLength = static_cast<size_t>(array->length); |
5270 | 0 | const size_t nCountIntersectingGeom = |
5271 | 0 | m_poFilterGeom ? (IsBinary(schema->children[iGeomField]->format) |
5272 | 0 | ? FillValidityArrayFromWKBArray<uint32_t>( |
5273 | 0 | array->children[iGeomField], this, |
5274 | 0 | abyValidityFromFilters) |
5275 | 0 | : FillValidityArrayFromWKBArray<uint64_t>( |
5276 | 0 | array->children[iGeomField], this, |
5277 | 0 | abyValidityFromFilters)) |
5278 | 0 | : nLength; |
5279 | 0 | if (!m_poFilterGeom) |
5280 | 0 | abyValidityFromFilters.resize(nLength, true); |
5281 | 0 | const size_t nCountIntersecting = |
5282 | 0 | m_poAttrQuery && nCountIntersectingGeom > 0 |
5283 | 0 | ? FillValidityArrayFromAttrQuery(this, m_poAttrQuery, schema, array, |
5284 | 0 | abyValidityFromFilters, |
5285 | 0 | papszOptions) |
5286 | 0 | : m_poFilterGeom ? nCountIntersectingGeom |
5287 | 0 | : nLength; |
5288 | | // Nothing to do ? |
5289 | 0 | if (nCountIntersecting == nLength) |
5290 | 0 | { |
5291 | | // CPLDebug("OGR", "All rows match filter"); |
5292 | 0 | return; |
5293 | 0 | } |
5294 | | |
5295 | 0 | if (nCountIntersecting == 0) |
5296 | 0 | { |
5297 | 0 | array->length = 0; |
5298 | 0 | } |
5299 | 0 | else if (!CompactStructArray(schema, array, 0, abyValidityFromFilters, |
5300 | 0 | nCountIntersecting)) |
5301 | 0 | { |
5302 | 0 | array->release(array); |
5303 | 0 | memset(array, 0, sizeof(*array)); |
5304 | 0 | } |
5305 | 0 | } |
5306 | | |
5307 | | /************************************************************************/ |
5308 | | /* OGRCloneArrowArray */ |
5309 | | /************************************************************************/ |
5310 | | |
5311 | | static bool OGRCloneArrowArray(const struct ArrowSchema *schema, |
5312 | | const struct ArrowArray *src_array, |
5313 | | struct ArrowArray *out_array, |
5314 | | size_t nParentOffset) |
5315 | 0 | { |
5316 | 0 | memset(out_array, 0, sizeof(*out_array)); |
5317 | 0 | const size_t nLength = |
5318 | 0 | static_cast<size_t>(src_array->length) - nParentOffset; |
5319 | 0 | out_array->length = nLength; |
5320 | 0 | out_array->null_count = src_array->null_count; |
5321 | 0 | out_array->release = OGRLayerDefaultReleaseArray; |
5322 | |
|
5323 | 0 | bool bRet = true; |
5324 | |
|
5325 | 0 | out_array->n_buffers = src_array->n_buffers; |
5326 | 0 | out_array->buffers = static_cast<const void **>(CPLCalloc( |
5327 | 0 | static_cast<size_t>(src_array->n_buffers), sizeof(const void *))); |
5328 | 0 | CPLAssert(static_cast<size_t>(src_array->length) >= nParentOffset); |
5329 | 0 | const char *format = schema->format; |
5330 | 0 | const auto nOffset = static_cast<size_t>(src_array->offset) + nParentOffset; |
5331 | 0 | for (int64_t i = 0; i < src_array->n_buffers; ++i) |
5332 | 0 | { |
5333 | 0 | if (i == 0 || IsBoolean(format)) |
5334 | 0 | { |
5335 | 0 | if (i == 1) |
5336 | 0 | { |
5337 | 0 | CPLAssert(src_array->buffers[i]); |
5338 | 0 | } |
5339 | 0 | if (src_array->buffers[i]) |
5340 | 0 | { |
5341 | 0 | const size_t nBytes = nLength ? (nLength + 7) / 8 : 1; |
5342 | 0 | uint8_t *CPL_RESTRICT p = static_cast<uint8_t *>( |
5343 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(nBytes)); |
5344 | 0 | if (!p) |
5345 | 0 | { |
5346 | 0 | bRet = false; |
5347 | 0 | break; |
5348 | 0 | } |
5349 | 0 | const auto *CPL_RESTRICT pSrcArray = |
5350 | 0 | static_cast<const uint8_t *>(src_array->buffers[i]); |
5351 | 0 | if ((nOffset % 8) != 0) |
5352 | 0 | { |
5353 | | // Make sure last byte is fully initialized |
5354 | 0 | p[nBytes - 1] = 0; |
5355 | 0 | for (size_t iRow = 0; iRow < nLength; ++iRow) |
5356 | 0 | { |
5357 | 0 | if (TestBit(pSrcArray, nOffset + iRow)) |
5358 | 0 | SetBit(p, iRow); |
5359 | 0 | else |
5360 | 0 | UnsetBit(p, iRow); |
5361 | 0 | } |
5362 | 0 | } |
5363 | 0 | else |
5364 | 0 | { |
5365 | 0 | memcpy(p, pSrcArray + nOffset / 8, nBytes); |
5366 | 0 | } |
5367 | 0 | out_array->buffers[i] = p; |
5368 | 0 | } |
5369 | 0 | } |
5370 | 0 | else if (i == 1) |
5371 | 0 | { |
5372 | 0 | CPLAssert(src_array->buffers[i]); |
5373 | 0 | size_t nEltSize = 0; |
5374 | 0 | size_t nExtraElt = 0; |
5375 | 0 | if (IsUInt8(format) || IsInt8(format)) |
5376 | 0 | nEltSize = sizeof(uint8_t); |
5377 | 0 | else if (IsUInt16(format) || IsInt16(format) || IsFloat16(format)) |
5378 | 0 | nEltSize = sizeof(uint16_t); |
5379 | 0 | else if (IsUInt32(format) || IsInt32(format) || IsFloat32(format) || |
5380 | 0 | strcmp(format, "tdD") == 0 || strcmp(format, "tts") == 0 || |
5381 | 0 | strcmp(format, "ttm") == 0) |
5382 | 0 | { |
5383 | 0 | nEltSize = sizeof(uint32_t); |
5384 | 0 | } |
5385 | 0 | else if (IsString(format) || IsBinary(format) || IsList(format) || |
5386 | 0 | IsMap(format)) |
5387 | 0 | { |
5388 | 0 | nEltSize = sizeof(uint32_t); |
5389 | 0 | nExtraElt = 1; |
5390 | 0 | } |
5391 | 0 | else if (IsUInt64(format) || IsInt64(format) || IsFloat64(format) || |
5392 | 0 | strcmp(format, "tdm") == 0 || strcmp(format, "ttu") == 0 || |
5393 | 0 | strcmp(format, "ttn") == 0 || IsTimestamp(format)) |
5394 | 0 | { |
5395 | 0 | nEltSize = sizeof(uint64_t); |
5396 | 0 | } |
5397 | 0 | else if (IsLargeString(format) || IsLargeBinary(format) || |
5398 | 0 | IsLargeList(format)) |
5399 | 0 | { |
5400 | 0 | nEltSize = sizeof(uint64_t); |
5401 | 0 | nExtraElt = 1; |
5402 | 0 | } |
5403 | 0 | else if (IsFixedWidthBinary(format)) |
5404 | 0 | { |
5405 | 0 | nEltSize = GetFixedWithBinary(format); |
5406 | 0 | } |
5407 | 0 | else if (IsDecimal(format)) |
5408 | 0 | { |
5409 | 0 | int nPrecision = 0; |
5410 | 0 | int nScale = 0; |
5411 | 0 | int nWidthInBytes = 0; |
5412 | 0 | if (!ParseDecimalFormat(format, nPrecision, nScale, |
5413 | 0 | nWidthInBytes)) |
5414 | 0 | { |
5415 | 0 | CPLError( |
5416 | 0 | CE_Failure, CPLE_AppDefined, |
5417 | 0 | "Unexpected error in OGRCloneArrowArray(): unhandled " |
5418 | 0 | "field format: %s", |
5419 | 0 | format); |
5420 | |
|
5421 | 0 | return false; |
5422 | 0 | } |
5423 | 0 | nEltSize = nWidthInBytes; |
5424 | 0 | } |
5425 | 0 | if (nEltSize) |
5426 | 0 | { |
5427 | 0 | void *p = VSI_MALLOC_ALIGNED_AUTO_VERBOSE( |
5428 | 0 | nLength ? nEltSize * (nLength + nExtraElt) : 1); |
5429 | 0 | if (!p) |
5430 | 0 | { |
5431 | 0 | bRet = false; |
5432 | 0 | break; |
5433 | 0 | } |
5434 | 0 | if (nLength) |
5435 | 0 | { |
5436 | 0 | if ((IsString(format) || IsBinary(format)) && |
5437 | 0 | static_cast<const uint32_t *>( |
5438 | 0 | src_array->buffers[1])[nOffset] != 0) |
5439 | 0 | { |
5440 | 0 | const auto *CPL_RESTRICT pSrcOffsets = |
5441 | 0 | static_cast<const uint32_t *>( |
5442 | 0 | src_array->buffers[1]) + |
5443 | 0 | nOffset; |
5444 | 0 | const auto nShiftOffset = pSrcOffsets[0]; |
5445 | 0 | auto *CPL_RESTRICT pDstOffsets = |
5446 | 0 | static_cast<uint32_t *>(p); |
5447 | 0 | for (size_t iRow = 0; iRow <= nLength; ++iRow) |
5448 | 0 | { |
5449 | 0 | pDstOffsets[iRow] = |
5450 | 0 | pSrcOffsets[iRow] - nShiftOffset; |
5451 | 0 | } |
5452 | 0 | } |
5453 | 0 | else if ((IsLargeString(format) || IsLargeBinary(format)) && |
5454 | 0 | static_cast<const uint64_t *>( |
5455 | 0 | src_array->buffers[1])[nOffset] != 0) |
5456 | 0 | { |
5457 | 0 | const auto *CPL_RESTRICT pSrcOffsets = |
5458 | 0 | static_cast<const uint64_t *>( |
5459 | 0 | src_array->buffers[1]) + |
5460 | 0 | nOffset; |
5461 | 0 | const auto nShiftOffset = pSrcOffsets[0]; |
5462 | 0 | auto *CPL_RESTRICT pDstOffsets = |
5463 | 0 | static_cast<uint64_t *>(p); |
5464 | 0 | for (size_t iRow = 0; iRow <= nLength; ++iRow) |
5465 | 0 | { |
5466 | 0 | pDstOffsets[iRow] = |
5467 | 0 | pSrcOffsets[iRow] - nShiftOffset; |
5468 | 0 | } |
5469 | 0 | } |
5470 | 0 | else |
5471 | 0 | { |
5472 | 0 | memcpy( |
5473 | 0 | p, |
5474 | 0 | static_cast<const GByte *>(src_array->buffers[i]) + |
5475 | 0 | nEltSize * nOffset, |
5476 | 0 | nEltSize * (nLength + nExtraElt)); |
5477 | 0 | } |
5478 | 0 | } |
5479 | 0 | out_array->buffers[i] = p; |
5480 | 0 | } |
5481 | 0 | else |
5482 | 0 | { |
5483 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
5484 | 0 | "OGRCloneArrowArray(): unhandled case, array = %s, " |
5485 | 0 | "format = '%s', i = 1", |
5486 | 0 | schema->name, format); |
5487 | 0 | bRet = false; |
5488 | 0 | break; |
5489 | 0 | } |
5490 | 0 | } |
5491 | 0 | else if (i == 2) |
5492 | 0 | { |
5493 | 0 | CPLAssert(src_array->buffers[i]); |
5494 | 0 | size_t nSrcCharOffset = 0; |
5495 | 0 | size_t nCharCount = 0; |
5496 | 0 | if (IsString(format) || IsBinary(format)) |
5497 | 0 | { |
5498 | 0 | const auto *pSrcOffsets = |
5499 | 0 | static_cast<const uint32_t *>(src_array->buffers[1]) + |
5500 | 0 | nOffset; |
5501 | 0 | nSrcCharOffset = pSrcOffsets[0]; |
5502 | 0 | nCharCount = pSrcOffsets[nLength] - pSrcOffsets[0]; |
5503 | 0 | } |
5504 | 0 | else if (IsLargeString(format) || IsLargeBinary(format)) |
5505 | 0 | { |
5506 | 0 | const auto *pSrcOffsets = |
5507 | 0 | static_cast<const uint64_t *>(src_array->buffers[1]) + |
5508 | 0 | nOffset; |
5509 | 0 | nSrcCharOffset = static_cast<size_t>(pSrcOffsets[0]); |
5510 | 0 | nCharCount = |
5511 | 0 | static_cast<size_t>(pSrcOffsets[nLength] - pSrcOffsets[0]); |
5512 | 0 | } |
5513 | 0 | else |
5514 | 0 | { |
5515 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
5516 | 0 | "OGRCloneArrowArray(): unhandled case, array = %s, " |
5517 | 0 | "format = '%s', i = 2", |
5518 | 0 | schema->name, format); |
5519 | 0 | bRet = false; |
5520 | 0 | break; |
5521 | 0 | } |
5522 | 0 | void *p = |
5523 | 0 | VSI_MALLOC_ALIGNED_AUTO_VERBOSE(nCharCount ? nCharCount : 1); |
5524 | 0 | if (!p) |
5525 | 0 | { |
5526 | 0 | bRet = false; |
5527 | 0 | break; |
5528 | 0 | } |
5529 | 0 | if (nCharCount) |
5530 | 0 | { |
5531 | 0 | memcpy(p, |
5532 | 0 | static_cast<const GByte *>(src_array->buffers[i]) + |
5533 | 0 | nSrcCharOffset, |
5534 | 0 | nCharCount); |
5535 | 0 | } |
5536 | 0 | out_array->buffers[i] = p; |
5537 | 0 | } |
5538 | 0 | else |
5539 | 0 | { |
5540 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
5541 | 0 | "OGRCloneArrowArray(): unhandled case, array = %s, format " |
5542 | 0 | "= '%s', i = 3", |
5543 | 0 | schema->name, format); |
5544 | 0 | bRet = false; |
5545 | 0 | break; |
5546 | 0 | } |
5547 | 0 | } |
5548 | | |
5549 | 0 | if (bRet) |
5550 | 0 | { |
5551 | 0 | out_array->n_children = src_array->n_children; |
5552 | 0 | out_array->children = static_cast<struct ArrowArray **>( |
5553 | 0 | CPLCalloc(static_cast<size_t>(src_array->n_children), |
5554 | 0 | sizeof(struct ArrowArray *))); |
5555 | 0 | for (int64_t i = 0; i < src_array->n_children; ++i) |
5556 | 0 | { |
5557 | 0 | out_array->children[i] = static_cast<struct ArrowArray *>( |
5558 | 0 | CPLCalloc(1, sizeof(struct ArrowArray))); |
5559 | 0 | if (!OGRCloneArrowArray(schema->children[i], src_array->children[i], |
5560 | 0 | out_array->children[i], |
5561 | 0 | IsFixedSizeList(format) |
5562 | 0 | ? nOffset * GetFixedSizeList(format) |
5563 | 0 | : IsStructure(format) ? nOffset |
5564 | 0 | : 0)) |
5565 | 0 | { |
5566 | 0 | bRet = false; |
5567 | 0 | break; |
5568 | 0 | } |
5569 | 0 | } |
5570 | 0 | } |
5571 | |
|
5572 | 0 | if (bRet && src_array->dictionary) |
5573 | 0 | { |
5574 | 0 | out_array->dictionary = static_cast<struct ArrowArray *>( |
5575 | 0 | CPLCalloc(1, sizeof(struct ArrowArray))); |
5576 | 0 | bRet = OGRCloneArrowArray(schema->dictionary, src_array->dictionary, |
5577 | 0 | out_array->dictionary, 0); |
5578 | 0 | } |
5579 | |
|
5580 | 0 | if (!bRet) |
5581 | 0 | { |
5582 | 0 | out_array->release(out_array); |
5583 | 0 | memset(out_array, 0, sizeof(*out_array)); |
5584 | 0 | } |
5585 | 0 | return bRet; |
5586 | 0 | } |
5587 | | |
5588 | | /** Full/deep copy of an array. |
5589 | | * |
5590 | | * Renormalize the offset of the array (and its children) to 0. |
5591 | | * |
5592 | | * In case of failure, out_array will be let in a released state. |
5593 | | * |
5594 | | * @param schema Schema of the array. Must *NOT* be NULL. |
5595 | | * @param src_array Source array. Must *NOT* be NULL. |
5596 | | * @param out_array Output array. Must *NOT* be NULL (but its content may be random) |
5597 | | * @return true if success. |
5598 | | */ |
5599 | | bool OGRCloneArrowArray(const struct ArrowSchema *schema, |
5600 | | const struct ArrowArray *src_array, |
5601 | | struct ArrowArray *out_array) |
5602 | 0 | { |
5603 | 0 | return OGRCloneArrowArray(schema, src_array, out_array, 0); |
5604 | 0 | } |
5605 | | |
5606 | | /************************************************************************/ |
5607 | | /* OGRCloneArrowMetadata() */ |
5608 | | /************************************************************************/ |
5609 | | |
5610 | | static void *OGRCloneArrowMetadata(const void *pMetadata) |
5611 | 0 | { |
5612 | 0 | if (!pMetadata) |
5613 | 0 | return nullptr; |
5614 | 0 | std::vector<GByte> abyOut; |
5615 | 0 | const GByte *pabyMetadata = static_cast<const GByte *>(pMetadata); |
5616 | 0 | int32_t nKVP; |
5617 | 0 | abyOut.insert(abyOut.end(), pabyMetadata, pabyMetadata + sizeof(int32_t)); |
5618 | 0 | memcpy(&nKVP, pabyMetadata, sizeof(int32_t)); |
5619 | 0 | pabyMetadata += sizeof(int32_t); |
5620 | 0 | for (int i = 0; i < nKVP; ++i) |
5621 | 0 | { |
5622 | 0 | int32_t nSizeKey; |
5623 | 0 | abyOut.insert(abyOut.end(), pabyMetadata, |
5624 | 0 | pabyMetadata + sizeof(int32_t)); |
5625 | 0 | memcpy(&nSizeKey, pabyMetadata, sizeof(int32_t)); |
5626 | 0 | pabyMetadata += sizeof(int32_t); |
5627 | 0 | abyOut.insert(abyOut.end(), pabyMetadata, pabyMetadata + nSizeKey); |
5628 | 0 | pabyMetadata += nSizeKey; |
5629 | |
|
5630 | 0 | int32_t nSizeValue; |
5631 | 0 | abyOut.insert(abyOut.end(), pabyMetadata, |
5632 | 0 | pabyMetadata + sizeof(int32_t)); |
5633 | 0 | memcpy(&nSizeValue, pabyMetadata, sizeof(int32_t)); |
5634 | 0 | pabyMetadata += sizeof(int32_t); |
5635 | 0 | abyOut.insert(abyOut.end(), pabyMetadata, pabyMetadata + nSizeValue); |
5636 | 0 | pabyMetadata += nSizeValue; |
5637 | 0 | } |
5638 | |
|
5639 | 0 | GByte *pabyOut = static_cast<GByte *>(VSI_MALLOC_VERBOSE(abyOut.size())); |
5640 | 0 | if (pabyOut) |
5641 | 0 | memcpy(pabyOut, abyOut.data(), abyOut.size()); |
5642 | 0 | return pabyOut; |
5643 | 0 | } |
5644 | | |
5645 | | /************************************************************************/ |
5646 | | /* OGRCloneArrowSchema() */ |
5647 | | /************************************************************************/ |
5648 | | |
5649 | | /** Full/deep copy of a schema. |
5650 | | * |
5651 | | * In case of failure, out_schema will be let in a released state. |
5652 | | * |
5653 | | * @param schema Schema to clone. Must *NOT* be NULL. |
5654 | | * @param out_schema Output schema. Must *NOT* be NULL (but its content may be random) |
5655 | | * @return true if success. |
5656 | | */ |
5657 | | bool OGRCloneArrowSchema(const struct ArrowSchema *schema, |
5658 | | struct ArrowSchema *out_schema) |
5659 | 0 | { |
5660 | 0 | memset(out_schema, 0, sizeof(*out_schema)); |
5661 | 0 | out_schema->release = OGRLayerFullReleaseSchema; |
5662 | 0 | out_schema->format = CPLStrdup(schema->format); |
5663 | 0 | out_schema->name = CPLStrdup(schema->name); |
5664 | 0 | out_schema->metadata = static_cast<const char *>( |
5665 | 0 | const_cast<const void *>(OGRCloneArrowMetadata(schema->metadata))); |
5666 | 0 | out_schema->flags = schema->flags; |
5667 | 0 | if (schema->n_children) |
5668 | 0 | { |
5669 | 0 | out_schema->children = |
5670 | 0 | static_cast<struct ArrowSchema **>(VSI_CALLOC_VERBOSE( |
5671 | 0 | static_cast<int>(schema->n_children), sizeof(ArrowSchema *))); |
5672 | 0 | if (!out_schema->children) |
5673 | 0 | { |
5674 | 0 | out_schema->release(out_schema); |
5675 | 0 | return false; |
5676 | 0 | } |
5677 | 0 | out_schema->n_children = schema->n_children; |
5678 | 0 | for (int i = 0; i < static_cast<int>(schema->n_children); ++i) |
5679 | 0 | { |
5680 | 0 | out_schema->children[i] = static_cast<struct ArrowSchema *>( |
5681 | 0 | CPLMalloc(sizeof(ArrowSchema))); |
5682 | 0 | if (!OGRCloneArrowSchema(schema->children[i], |
5683 | 0 | out_schema->children[i])) |
5684 | 0 | { |
5685 | 0 | out_schema->release(out_schema); |
5686 | 0 | return false; |
5687 | 0 | } |
5688 | 0 | } |
5689 | 0 | } |
5690 | 0 | if (schema->dictionary) |
5691 | 0 | { |
5692 | 0 | out_schema->dictionary = |
5693 | 0 | static_cast<struct ArrowSchema *>(CPLMalloc(sizeof(ArrowSchema))); |
5694 | 0 | if (!OGRCloneArrowSchema(schema->dictionary, out_schema->dictionary)) |
5695 | 0 | { |
5696 | 0 | out_schema->release(out_schema); |
5697 | 0 | return false; |
5698 | 0 | } |
5699 | 0 | } |
5700 | 0 | return true; |
5701 | 0 | } |
5702 | | |
5703 | | /************************************************************************/ |
5704 | | /* OGRLayer::IsArrowSchemaSupported() */ |
5705 | | /************************************************************************/ |
5706 | | |
5707 | | const struct |
5708 | | { |
5709 | | const char *arrowType; |
5710 | | OGRFieldType eType; |
5711 | | OGRFieldSubType eSubType; |
5712 | | } gasArrowTypesToOGR[] = { |
5713 | | {"b", OFTInteger, OFSTBoolean}, {"c", OFTInteger, OFSTInt16}, // Int8 |
5714 | | {"C", OFTInteger, OFSTInt16}, // UInt8 |
5715 | | {"s", OFTInteger, OFSTInt16}, // Int16 |
5716 | | {"S", OFTInteger, OFSTNone}, // UInt16 |
5717 | | {"i", OFTInteger, OFSTNone}, // Int32 |
5718 | | {"I", OFTInteger64, OFSTNone}, // UInt32 |
5719 | | {"l", OFTInteger64, OFSTNone}, // Int64 |
5720 | | {"L", OFTReal, OFSTNone}, // UInt64 (potentially lossy conversion if going through OGRFeature) |
5721 | | {"e", OFTReal, OFSTFloat32}, // float16 |
5722 | | {"f", OFTReal, OFSTFloat32}, // float32 |
5723 | | {"g", OFTReal, OFSTNone}, // float64 |
5724 | | {"z", OFTBinary, OFSTNone}, // binary |
5725 | | {"Z", OFTBinary, OFSTNone}, // large binary (will be limited to 32 bit length though if going through OGRFeature!) |
5726 | | {"u", OFTString, OFSTNone}, // string |
5727 | | {"U", OFTString, OFSTNone}, // large string |
5728 | | {"tdD", OFTDate, OFSTNone}, // date32[days] |
5729 | | {"tdm", OFTDate, OFSTNone}, // date64[milliseconds] |
5730 | | {"tts", OFTTime, OFSTNone}, // time32 [seconds] |
5731 | | {"ttm", OFTTime, OFSTNone}, // time32 [milliseconds] |
5732 | | {"ttu", OFTTime, OFSTNone}, // time64 [microseconds] |
5733 | | {"ttn", OFTTime, OFSTNone}, // time64 [nanoseconds] |
5734 | | }; |
5735 | | |
5736 | | const struct |
5737 | | { |
5738 | | const char arrowLetter; |
5739 | | OGRFieldType eType; |
5740 | | OGRFieldSubType eSubType; |
5741 | | } gasListTypes[] = { |
5742 | | {ARROW_LETTER_BOOLEAN, OFTIntegerList, OFSTBoolean}, |
5743 | | {ARROW_LETTER_INT8, OFTIntegerList, OFSTInt16}, |
5744 | | {ARROW_LETTER_UINT8, OFTIntegerList, OFSTInt16}, |
5745 | | {ARROW_LETTER_INT16, OFTIntegerList, OFSTInt16}, |
5746 | | {ARROW_LETTER_UINT16, OFTIntegerList, OFSTNone}, |
5747 | | {ARROW_LETTER_INT32, OFTIntegerList, OFSTNone}, |
5748 | | {ARROW_LETTER_UINT32, OFTInteger64List, OFSTNone}, |
5749 | | {ARROW_LETTER_INT64, OFTInteger64List, OFSTNone}, |
5750 | | {ARROW_LETTER_UINT64, OFTRealList, |
5751 | | OFSTNone}, //(potentially lossy conversion if going through OGRFeature) |
5752 | | {ARROW_LETTER_FLOAT16, OFTRealList, OFSTFloat32}, |
5753 | | {ARROW_LETTER_FLOAT32, OFTRealList, OFSTFloat32}, |
5754 | | {ARROW_LETTER_FLOAT64, OFTRealList, OFSTNone}, |
5755 | | {ARROW_LETTER_STRING, OFTStringList, OFSTNone}, |
5756 | | {ARROW_LETTER_LARGE_STRING, OFTStringList, OFSTNone}, |
5757 | | }; |
5758 | | |
5759 | | static inline bool IsValidDictionaryIndexType(const char *format) |
5760 | 0 | { |
5761 | 0 | return (format[0] == ARROW_LETTER_INT8 || format[0] == ARROW_LETTER_UINT8 || |
5762 | 0 | format[0] == ARROW_LETTER_INT16 || |
5763 | 0 | format[0] == ARROW_LETTER_UINT16 || |
5764 | 0 | format[0] == ARROW_LETTER_INT32 || |
5765 | 0 | format[0] == ARROW_LETTER_UINT32 || |
5766 | 0 | format[0] == ARROW_LETTER_INT64 || |
5767 | 0 | format[0] == ARROW_LETTER_UINT64) && |
5768 | 0 | format[1] == 0; |
5769 | 0 | } |
5770 | | |
5771 | | static bool IsSupportForJSONObj(const struct ArrowSchema *schema) |
5772 | 0 | { |
5773 | 0 | const char *format = schema->format; |
5774 | 0 | if (IsStructure(format)) |
5775 | 0 | { |
5776 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
5777 | 0 | { |
5778 | 0 | if (!IsSupportForJSONObj(schema->children[i])) |
5779 | 0 | return false; |
5780 | 0 | } |
5781 | 0 | return true; |
5782 | 0 | } |
5783 | | |
5784 | 0 | for (const auto &sType : gasListTypes) |
5785 | 0 | { |
5786 | 0 | if (format[0] == sType.arrowLetter && format[1] == 0) |
5787 | 0 | { |
5788 | 0 | return true; |
5789 | 0 | } |
5790 | 0 | } |
5791 | | |
5792 | 0 | if (IsBinary(format) || IsLargeBinary(format) || IsFixedWidthBinary(format)) |
5793 | 0 | return true; |
5794 | | |
5795 | 0 | if (IsDecimal(format)) |
5796 | 0 | { |
5797 | 0 | int nPrecision = 0; |
5798 | 0 | int nScale = 0; |
5799 | 0 | int nWidthInBytes = 0; |
5800 | 0 | if (!ParseDecimalFormat(format, nPrecision, nScale, nWidthInBytes)) |
5801 | 0 | { |
5802 | 0 | CPLError(CE_Failure, CPLE_AppDefined, "Invalid field format %s", |
5803 | 0 | format); |
5804 | 0 | return false; |
5805 | 0 | } |
5806 | | |
5807 | 0 | return GetErrorIfUnsupportedDecimal(nWidthInBytes, nPrecision) == |
5808 | 0 | nullptr; |
5809 | 0 | } |
5810 | | |
5811 | 0 | if (IsMap(format)) |
5812 | 0 | { |
5813 | 0 | return IsStructure(schema->children[0]->format) && |
5814 | 0 | schema->children[0]->n_children == 2 && |
5815 | 0 | IsString(schema->children[0]->children[0]->format) && |
5816 | 0 | IsSupportForJSONObj(schema->children[0]->children[1]); |
5817 | 0 | } |
5818 | | |
5819 | 0 | if (IsList(format) || IsLargeList(format) || IsFixedSizeList(format)) |
5820 | 0 | { |
5821 | 0 | return IsSupportForJSONObj(schema->children[0]); |
5822 | 0 | } |
5823 | | |
5824 | 0 | return false; |
5825 | 0 | } |
5826 | | |
5827 | | static bool IsArrowSchemaSupportedInternal(const struct ArrowSchema *schema, |
5828 | | const std::string &osFieldPrefix, |
5829 | | std::string &osErrorMsg) |
5830 | 0 | { |
5831 | 0 | const auto AppendError = [&osErrorMsg](const std::string &osMsg) |
5832 | 0 | { |
5833 | 0 | if (!osErrorMsg.empty()) |
5834 | 0 | osErrorMsg += " "; |
5835 | 0 | osErrorMsg += osMsg; |
5836 | 0 | }; |
5837 | |
|
5838 | 0 | const char *fieldName = schema->name; |
5839 | 0 | const char *format = schema->format; |
5840 | 0 | if (IsStructure(format)) |
5841 | 0 | { |
5842 | 0 | bool bRet = true; |
5843 | 0 | const std::string osNewPrefix(osFieldPrefix + fieldName + "."); |
5844 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
5845 | 0 | { |
5846 | 0 | if (!IsArrowSchemaSupportedInternal(schema->children[i], |
5847 | 0 | osNewPrefix, osErrorMsg)) |
5848 | 0 | bRet = false; |
5849 | 0 | } |
5850 | 0 | return bRet; |
5851 | 0 | } |
5852 | | |
5853 | 0 | if (schema->dictionary) |
5854 | 0 | { |
5855 | 0 | if (!IsValidDictionaryIndexType(format)) |
5856 | 0 | { |
5857 | 0 | AppendError("Dictionary only supported if the parent is of " |
5858 | 0 | "type [U]Int[8|16|32|64]"); |
5859 | 0 | return false; |
5860 | 0 | } |
5861 | | |
5862 | 0 | schema = schema->dictionary; |
5863 | 0 | format = schema->format; |
5864 | 0 | } |
5865 | | |
5866 | 0 | if (IsList(format) || IsLargeList(format) || IsFixedSizeList(format)) |
5867 | 0 | { |
5868 | | // Only some subtypes supported |
5869 | 0 | const char *childFormat = schema->children[0]->format; |
5870 | 0 | for (const auto &sType : gasListTypes) |
5871 | 0 | { |
5872 | 0 | if (childFormat[0] == sType.arrowLetter && childFormat[1] == 0) |
5873 | 0 | { |
5874 | 0 | return true; |
5875 | 0 | } |
5876 | 0 | } |
5877 | | |
5878 | 0 | if (IsDecimal(childFormat)) |
5879 | 0 | { |
5880 | 0 | int nPrecision = 0; |
5881 | 0 | int nScale = 0; |
5882 | 0 | int nWidthInBytes = 0; |
5883 | 0 | if (!ParseDecimalFormat(childFormat, nPrecision, nScale, |
5884 | 0 | nWidthInBytes)) |
5885 | 0 | { |
5886 | 0 | AppendError(std::string("Invalid field format ") + childFormat + |
5887 | 0 | " for field " + osFieldPrefix + fieldName); |
5888 | 0 | return false; |
5889 | 0 | } |
5890 | | |
5891 | 0 | const char *pszError = |
5892 | 0 | GetErrorIfUnsupportedDecimal(nWidthInBytes, nPrecision); |
5893 | 0 | if (pszError) |
5894 | 0 | { |
5895 | 0 | AppendError(pszError); |
5896 | 0 | return false; |
5897 | 0 | } |
5898 | | |
5899 | 0 | return true; |
5900 | 0 | } |
5901 | | |
5902 | 0 | if (IsSupportForJSONObj(schema)) |
5903 | 0 | { |
5904 | 0 | return true; |
5905 | 0 | } |
5906 | | |
5907 | 0 | AppendError("Type list for field " + osFieldPrefix + fieldName + |
5908 | 0 | " is not supported."); |
5909 | 0 | return false; |
5910 | 0 | } |
5911 | | |
5912 | 0 | else if (IsMap(format)) |
5913 | 0 | { |
5914 | 0 | if (IsSupportForJSONObj(schema)) |
5915 | 0 | return true; |
5916 | | |
5917 | 0 | AppendError("Type map for field " + osFieldPrefix + fieldName + |
5918 | 0 | " is not supported."); |
5919 | 0 | return false; |
5920 | 0 | } |
5921 | 0 | else if (IsDecimal(format)) |
5922 | 0 | { |
5923 | 0 | int nPrecision = 0; |
5924 | 0 | int nScale = 0; |
5925 | 0 | int nWidthInBytes = 0; |
5926 | 0 | if (!ParseDecimalFormat(format, nPrecision, nScale, nWidthInBytes)) |
5927 | 0 | { |
5928 | 0 | AppendError(std::string("Invalid field format ") + format + |
5929 | 0 | " for field " + osFieldPrefix + fieldName); |
5930 | 0 | return false; |
5931 | 0 | } |
5932 | | |
5933 | 0 | const char *pszError = |
5934 | 0 | GetErrorIfUnsupportedDecimal(nWidthInBytes, nPrecision); |
5935 | 0 | if (pszError) |
5936 | 0 | { |
5937 | 0 | AppendError(pszError); |
5938 | 0 | return false; |
5939 | 0 | } |
5940 | | |
5941 | 0 | return true; |
5942 | 0 | } |
5943 | 0 | else |
5944 | 0 | { |
5945 | 0 | for (const auto &sType : gasArrowTypesToOGR) |
5946 | 0 | { |
5947 | 0 | if (strcmp(format, sType.arrowType) == 0) |
5948 | 0 | { |
5949 | 0 | return true; |
5950 | 0 | } |
5951 | 0 | } |
5952 | | |
5953 | 0 | if (IsFixedWidthBinary(format) || IsTimestamp(format)) |
5954 | 0 | return true; |
5955 | | |
5956 | 0 | AppendError("Type '" + std::string(format) + "' for field " + |
5957 | 0 | osFieldPrefix + fieldName + " is not supported."); |
5958 | 0 | return false; |
5959 | 0 | } |
5960 | 0 | } |
5961 | | |
5962 | | /** Returns whether the provided ArrowSchema is supported for writing. |
5963 | | * |
5964 | | * This method exists since not all drivers may support all Arrow data types. |
5965 | | * |
5966 | | * The ArrowSchema must be of type struct (format=+s) |
5967 | | * |
5968 | | * It is recommended to call this method before calling WriteArrowBatch(). |
5969 | | * |
5970 | | * This is the same as the C function OGR_L_IsArrowSchemaSupported(). |
5971 | | * |
5972 | | * @param schema Schema of type struct (format = '+s') |
5973 | | * @param papszOptions Options (none currently). Null terminated list, or nullptr. |
5974 | | * @param[out] osErrorMsg Reason of the failure, when this method returns false. |
5975 | | * @return true if the ArrowSchema is supported for writing. |
5976 | | * @since 3.8 |
5977 | | */ |
5978 | | bool OGRLayer::IsArrowSchemaSupported(const struct ArrowSchema *schema, |
5979 | | CPL_UNUSED CSLConstList papszOptions, |
5980 | | std::string &osErrorMsg) const |
5981 | 0 | { |
5982 | 0 | if (!IsStructure(schema->format)) |
5983 | 0 | { |
5984 | 0 | osErrorMsg = |
5985 | 0 | "IsArrowSchemaSupported() should be called on a schema that is a " |
5986 | 0 | "struct of fields"; |
5987 | 0 | return false; |
5988 | 0 | } |
5989 | | |
5990 | 0 | bool bRet = true; |
5991 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
5992 | 0 | { |
5993 | 0 | if (!IsArrowSchemaSupportedInternal(schema->children[i], std::string(), |
5994 | 0 | osErrorMsg)) |
5995 | 0 | bRet = false; |
5996 | 0 | } |
5997 | 0 | return bRet; |
5998 | 0 | } |
5999 | | |
6000 | | /************************************************************************/ |
6001 | | /* OGR_L_IsArrowSchemaSupported() */ |
6002 | | /************************************************************************/ |
6003 | | |
6004 | | /** Returns whether the provided ArrowSchema is supported for writing. |
6005 | | * |
6006 | | * This function exists since not all drivers may support all Arrow data types. |
6007 | | * |
6008 | | * The ArrowSchema must be of type struct (format=+s) |
6009 | | * |
6010 | | * It is recommended to call this function before calling OGR_L_WriteArrowBatch(). |
6011 | | * |
6012 | | * This is the same as the C++ method OGRLayer::IsArrowSchemaSupported(). |
6013 | | * |
6014 | | * @param hLayer Layer. |
6015 | | * @param schema Schema of type struct (format = '+s') |
6016 | | * @param papszOptions Options (none currently). Null terminated list, or nullptr. |
6017 | | * @param[out] ppszErrorMsg nullptr, or pointer to a string that will contain |
6018 | | * the reason of the failure, when this function returns false. |
6019 | | * @return true if the ArrowSchema is supported for writing. |
6020 | | * @since 3.8 |
6021 | | */ |
6022 | | bool OGR_L_IsArrowSchemaSupported(OGRLayerH hLayer, |
6023 | | const struct ArrowSchema *schema, |
6024 | | char **papszOptions, char **ppszErrorMsg) |
6025 | 0 | { |
6026 | 0 | VALIDATE_POINTER1(hLayer, __func__, false); |
6027 | 0 | VALIDATE_POINTER1(schema, __func__, false); |
6028 | | |
6029 | 0 | std::string osErrorMsg; |
6030 | 0 | if (!OGRLayer::FromHandle(hLayer)->IsArrowSchemaSupported( |
6031 | 0 | schema, papszOptions, osErrorMsg)) |
6032 | 0 | { |
6033 | 0 | if (ppszErrorMsg) |
6034 | 0 | *ppszErrorMsg = VSIStrdup(osErrorMsg.c_str()); |
6035 | 0 | return false; |
6036 | 0 | } |
6037 | 0 | else |
6038 | 0 | { |
6039 | 0 | if (ppszErrorMsg) |
6040 | 0 | *ppszErrorMsg = nullptr; |
6041 | 0 | return true; |
6042 | 0 | } |
6043 | 0 | } |
6044 | | |
6045 | | /************************************************************************/ |
6046 | | /* IsKnownCodedFieldDomain() */ |
6047 | | /************************************************************************/ |
6048 | | |
6049 | | static bool IsKnownCodedFieldDomain(OGRLayer *poLayer, |
6050 | | const char *arrowMetadata) |
6051 | 0 | { |
6052 | 0 | if (arrowMetadata) |
6053 | 0 | { |
6054 | 0 | const auto oMetadata = OGRParseArrowMetadata(arrowMetadata); |
6055 | 0 | for (const auto &oIter : oMetadata) |
6056 | 0 | { |
6057 | 0 | if (oIter.first == MD_GDAL_OGR_DOMAIN_NAME) |
6058 | 0 | { |
6059 | 0 | auto poDS = poLayer->GetDataset(); |
6060 | 0 | if (poDS) |
6061 | 0 | { |
6062 | 0 | const auto poFieldDomain = |
6063 | 0 | poDS->GetFieldDomain(oIter.second); |
6064 | 0 | if (poFieldDomain && |
6065 | 0 | poFieldDomain->GetDomainType() == OFDT_CODED) |
6066 | 0 | { |
6067 | 0 | return true; |
6068 | 0 | } |
6069 | 0 | } |
6070 | 0 | } |
6071 | 0 | } |
6072 | 0 | } |
6073 | 0 | return false; |
6074 | 0 | } |
6075 | | |
6076 | | /************************************************************************/ |
6077 | | /* OGRLayer::CreateFieldFromArrowSchema() */ |
6078 | | /************************************************************************/ |
6079 | | |
6080 | | //! @cond Doxygen_Suppress |
6081 | | bool OGRLayer::CreateFieldFromArrowSchemaInternal( |
6082 | | const struct ArrowSchema *schema, const std::string &osFieldPrefix, |
6083 | | CSLConstList papszOptions) |
6084 | 0 | { |
6085 | 0 | const char *fieldName = schema->name; |
6086 | 0 | const char *format = schema->format; |
6087 | 0 | if (IsStructure(format)) |
6088 | 0 | { |
6089 | 0 | const std::string osNewPrefix(osFieldPrefix + fieldName + "."); |
6090 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
6091 | 0 | { |
6092 | 0 | if (!CreateFieldFromArrowSchemaInternal(schema->children[i], |
6093 | 0 | osNewPrefix, papszOptions)) |
6094 | 0 | return false; |
6095 | 0 | } |
6096 | 0 | return true; |
6097 | 0 | } |
6098 | | |
6099 | 0 | CPLStringList aosNativeTypes; |
6100 | 0 | auto poLayer = const_cast<OGRLayer *>(this); |
6101 | 0 | auto poDS = poLayer->GetDataset(); |
6102 | 0 | if (poDS) |
6103 | 0 | { |
6104 | 0 | auto poDriver = poDS->GetDriver(); |
6105 | 0 | if (poDriver) |
6106 | 0 | { |
6107 | 0 | const char *pszMetadataItem = |
6108 | 0 | poDriver->GetMetadataItem(GDAL_DMD_CREATIONFIELDDATATYPES); |
6109 | 0 | if (pszMetadataItem) |
6110 | 0 | aosNativeTypes = CSLTokenizeString2(pszMetadataItem, " ", 0); |
6111 | 0 | } |
6112 | 0 | } |
6113 | |
|
6114 | 0 | if (schema->dictionary && |
6115 | 0 | !IsKnownCodedFieldDomain(poLayer, schema->metadata)) |
6116 | 0 | { |
6117 | 0 | if (!IsValidDictionaryIndexType(format)) |
6118 | 0 | { |
6119 | 0 | CPLError(CE_Failure, CPLE_NotSupported, |
6120 | 0 | "Dictionary only supported if the parent is of " |
6121 | 0 | "type [U]Int[8|16|32|64]"); |
6122 | 0 | return false; |
6123 | 0 | } |
6124 | | |
6125 | 0 | schema = schema->dictionary; |
6126 | 0 | format = schema->format; |
6127 | 0 | } |
6128 | | |
6129 | 0 | const auto AddField = [this, schema, fieldName, &aosNativeTypes, |
6130 | 0 | &osFieldPrefix, poDS](OGRFieldType eTypeIn, |
6131 | 0 | OGRFieldSubType eSubTypeIn, |
6132 | 0 | int nWidth, int nPrecision) |
6133 | 0 | { |
6134 | 0 | const char *pszTypeName = OGRFieldDefn::GetFieldTypeName(eTypeIn); |
6135 | 0 | auto eTypeOut = eTypeIn; |
6136 | 0 | auto eSubTypeOut = eSubTypeIn; |
6137 | 0 | if (!aosNativeTypes.empty() && |
6138 | 0 | aosNativeTypes.FindString(pszTypeName) < 0) |
6139 | 0 | { |
6140 | 0 | eTypeOut = OFTString; |
6141 | 0 | eSubTypeOut = |
6142 | 0 | (eTypeIn == OFTIntegerList || eTypeIn == OFTInteger64List || |
6143 | 0 | eTypeIn == OFTRealList || eTypeIn == OFTStringList) |
6144 | 0 | ? OFSTJSON |
6145 | 0 | : OFSTNone; |
6146 | 0 | } |
6147 | |
|
6148 | 0 | const std::string osWantedOGRFieldName = osFieldPrefix + fieldName; |
6149 | 0 | OGRFieldDefn oFieldDefn(osWantedOGRFieldName.c_str(), eTypeOut); |
6150 | 0 | oFieldDefn.SetSubType(eSubTypeOut); |
6151 | 0 | if (eTypeOut == eTypeIn && eSubTypeOut == eSubTypeIn) |
6152 | 0 | { |
6153 | 0 | oFieldDefn.SetWidth(nWidth); |
6154 | 0 | oFieldDefn.SetPrecision(nPrecision); |
6155 | 0 | } |
6156 | 0 | oFieldDefn.SetNullable((schema->flags & ARROW_FLAG_NULLABLE) != 0); |
6157 | |
|
6158 | 0 | if (schema->metadata) |
6159 | 0 | { |
6160 | 0 | const auto oMetadata = OGRParseArrowMetadata(schema->metadata); |
6161 | 0 | for (const auto &oIter : oMetadata) |
6162 | 0 | { |
6163 | 0 | if (oIter.first == MD_GDAL_OGR_TYPE) |
6164 | 0 | { |
6165 | 0 | const auto &osType = oIter.second; |
6166 | 0 | for (auto eType = OFTInteger; eType <= OFTMaxType;) |
6167 | 0 | { |
6168 | 0 | if (OGRFieldDefn::GetFieldTypeName(eType) == osType) |
6169 | 0 | { |
6170 | 0 | oFieldDefn.SetType(eType); |
6171 | 0 | break; |
6172 | 0 | } |
6173 | 0 | if (eType == OFTMaxType) |
6174 | 0 | break; |
6175 | 0 | else |
6176 | 0 | eType = static_cast<OGRFieldType>(eType + 1); |
6177 | 0 | } |
6178 | 0 | } |
6179 | 0 | else if (oIter.first == MD_GDAL_OGR_ALTERNATIVE_NAME) |
6180 | 0 | oFieldDefn.SetAlternativeName(oIter.second.c_str()); |
6181 | 0 | else if (oIter.first == MD_GDAL_OGR_COMMENT) |
6182 | 0 | oFieldDefn.SetComment(oIter.second); |
6183 | 0 | else if (oIter.first == MD_GDAL_OGR_DEFAULT) |
6184 | 0 | oFieldDefn.SetDefault(oIter.second.c_str()); |
6185 | 0 | else if (oIter.first == MD_GDAL_OGR_SUBTYPE) |
6186 | 0 | { |
6187 | 0 | if (eTypeIn == eTypeOut) |
6188 | 0 | { |
6189 | 0 | const auto &osSubType = oIter.second; |
6190 | 0 | for (auto eSubType = OFSTNone; |
6191 | 0 | eSubType <= OFSTMaxSubType;) |
6192 | 0 | { |
6193 | 0 | if (OGRFieldDefn::GetFieldSubTypeName(eSubType) == |
6194 | 0 | osSubType) |
6195 | 0 | { |
6196 | 0 | oFieldDefn.SetSubType(eSubType); |
6197 | 0 | break; |
6198 | 0 | } |
6199 | 0 | if (eSubType == OFSTMaxSubType) |
6200 | 0 | break; |
6201 | 0 | else |
6202 | 0 | eSubType = |
6203 | 0 | static_cast<OGRFieldSubType>(eSubType + 1); |
6204 | 0 | } |
6205 | 0 | } |
6206 | 0 | } |
6207 | 0 | else if (oIter.first == MD_GDAL_OGR_WIDTH) |
6208 | 0 | oFieldDefn.SetWidth(atoi(oIter.second.c_str())); |
6209 | 0 | else if (oIter.first == MD_GDAL_OGR_UNIQUE) |
6210 | 0 | oFieldDefn.SetUnique(oIter.second == "true"); |
6211 | 0 | else if (oIter.first == MD_GDAL_OGR_DOMAIN_NAME) |
6212 | 0 | { |
6213 | 0 | if (poDS && poDS->GetFieldDomain(oIter.second)) |
6214 | 0 | oFieldDefn.SetDomainName(oIter.second); |
6215 | 0 | } |
6216 | 0 | else if (oIter.first == ARROW_EXTENSION_NAME_KEY && |
6217 | 0 | oIter.second == EXTENSION_NAME_ARROW_JSON) |
6218 | 0 | { |
6219 | 0 | oFieldDefn.SetSubType(OFSTJSON); |
6220 | 0 | } |
6221 | 0 | else |
6222 | 0 | { |
6223 | 0 | CPLDebug("OGR", "Unknown field metadata: %s", |
6224 | 0 | oIter.first.c_str()); |
6225 | 0 | } |
6226 | 0 | } |
6227 | 0 | } |
6228 | 0 | auto poLayerDefn = GetLayerDefn(); |
6229 | 0 | const int nFieldCountBefore = poLayerDefn->GetFieldCount(); |
6230 | 0 | if (CreateField(&oFieldDefn) != OGRERR_NONE || |
6231 | 0 | nFieldCountBefore + 1 != poLayerDefn->GetFieldCount()) |
6232 | 0 | { |
6233 | 0 | return false; |
6234 | 0 | } |
6235 | 0 | const char *pszActualFieldName = |
6236 | 0 | poLayerDefn->GetFieldDefn(nFieldCountBefore)->GetNameRef(); |
6237 | 0 | if (pszActualFieldName != osWantedOGRFieldName) |
6238 | 0 | { |
6239 | 0 | m_poPrivate |
6240 | 0 | ->m_oMapArrowFieldNameToOGRFieldName[osWantedOGRFieldName] = |
6241 | 0 | pszActualFieldName; |
6242 | 0 | } |
6243 | 0 | return true; |
6244 | 0 | }; |
6245 | |
|
6246 | 0 | for (const auto &sType : gasArrowTypesToOGR) |
6247 | 0 | { |
6248 | 0 | if (strcmp(format, sType.arrowType) == 0) |
6249 | 0 | { |
6250 | 0 | return AddField(sType.eType, sType.eSubType, 0, 0); |
6251 | 0 | } |
6252 | 0 | } |
6253 | | |
6254 | 0 | if (IsMap(format)) |
6255 | 0 | { |
6256 | 0 | return AddField(OFTString, OFSTJSON, 0, 0); |
6257 | 0 | } |
6258 | | |
6259 | 0 | if (IsTimestamp(format)) |
6260 | 0 | { |
6261 | 0 | return AddField(OFTDateTime, OFSTNone, 0, 0); |
6262 | 0 | } |
6263 | | |
6264 | 0 | if (IsFixedWidthBinary(format)) |
6265 | 0 | { |
6266 | 0 | return AddField(OFTBinary, OFSTNone, GetFixedWithBinary(format), 0); |
6267 | 0 | } |
6268 | | |
6269 | 0 | if (IsList(format) || IsLargeList(format) || IsFixedSizeList(format)) |
6270 | 0 | { |
6271 | 0 | const char *childFormat = schema->children[0]->format; |
6272 | 0 | for (const auto &sType : gasListTypes) |
6273 | 0 | { |
6274 | 0 | if (childFormat[0] == sType.arrowLetter && childFormat[1] == 0) |
6275 | 0 | { |
6276 | 0 | return AddField(sType.eType, sType.eSubType, 0, 0); |
6277 | 0 | } |
6278 | 0 | } |
6279 | | |
6280 | 0 | if (IsDecimal(childFormat)) |
6281 | 0 | { |
6282 | 0 | int nPrecision = 0; |
6283 | 0 | int nScale = 0; |
6284 | 0 | int nWidthInBytes = 0; |
6285 | 0 | if (!ParseDecimalFormat(childFormat, nPrecision, nScale, |
6286 | 0 | nWidthInBytes)) |
6287 | 0 | { |
6288 | 0 | CPLError(CE_Failure, CPLE_AppDefined, "%s", |
6289 | 0 | (std::string("Invalid field format ") + format + |
6290 | 0 | " for field " + osFieldPrefix + fieldName) |
6291 | 0 | .c_str()); |
6292 | 0 | return false; |
6293 | 0 | } |
6294 | | |
6295 | 0 | const char *pszError = |
6296 | 0 | GetErrorIfUnsupportedDecimal(nWidthInBytes, nPrecision); |
6297 | 0 | if (pszError) |
6298 | 0 | { |
6299 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", pszError); |
6300 | 0 | return false; |
6301 | 0 | } |
6302 | | |
6303 | | // DBF convention: add space for negative sign and decimal separator |
6304 | 0 | return AddField(OFTRealList, OFSTNone, nPrecision + 2, nScale); |
6305 | 0 | } |
6306 | | |
6307 | 0 | if (IsSupportForJSONObj(schema->children[0])) |
6308 | 0 | { |
6309 | 0 | return AddField(OFTString, OFSTJSON, 0, 0); |
6310 | 0 | } |
6311 | | |
6312 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", |
6313 | 0 | ("List of type '" + std::string(childFormat) + "' for field " + |
6314 | 0 | osFieldPrefix + fieldName + " is not supported.") |
6315 | 0 | .c_str()); |
6316 | 0 | return false; |
6317 | 0 | } |
6318 | | |
6319 | 0 | if (IsDecimal(format)) |
6320 | 0 | { |
6321 | 0 | int nPrecision = 0; |
6322 | 0 | int nScale = 0; |
6323 | 0 | int nWidthInBytes = 0; |
6324 | 0 | if (!ParseDecimalFormat(format, nPrecision, nScale, nWidthInBytes)) |
6325 | 0 | { |
6326 | 0 | CPLError(CE_Failure, CPLE_AppDefined, "%s", |
6327 | 0 | (std::string("Invalid field format ") + format + |
6328 | 0 | " for field " + osFieldPrefix + fieldName) |
6329 | 0 | .c_str()); |
6330 | 0 | return false; |
6331 | 0 | } |
6332 | | |
6333 | 0 | const char *pszError = |
6334 | 0 | GetErrorIfUnsupportedDecimal(nWidthInBytes, nPrecision); |
6335 | 0 | if (pszError) |
6336 | 0 | { |
6337 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", pszError); |
6338 | 0 | return false; |
6339 | 0 | } |
6340 | | |
6341 | | // DBF convention: add space for negative sign and decimal separator |
6342 | 0 | return AddField(OFTReal, OFSTNone, nPrecision + 2, nScale); |
6343 | 0 | } |
6344 | | |
6345 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", |
6346 | 0 | ("Type '" + std::string(format) + "' for field " + osFieldPrefix + |
6347 | 0 | fieldName + " is not supported.") |
6348 | 0 | .c_str()); |
6349 | 0 | return false; |
6350 | 0 | } |
6351 | | |
6352 | | //! @endcond |
6353 | | |
6354 | | /** Creates a field from an ArrowSchema. |
6355 | | * |
6356 | | * This should only be used for attribute fields. Geometry fields should |
6357 | | * be created with CreateGeomField(). The FID field should also not be |
6358 | | * passed with this method. |
6359 | | * |
6360 | | * Contrary to the IsArrowSchemaSupported() and WriteArrowBatch() methods, the |
6361 | | * passed schema must be for an individual field, and thus, is *not* of type |
6362 | | * struct (format=+s) (unless writing a set of fields grouped together in the |
6363 | | * same structure). |
6364 | | * |
6365 | | * Additional field metadata can be speciffed through the ArrowSchema::metadata |
6366 | | * field with the potential following items: |
6367 | | * <ul> |
6368 | | * <li>"GDAL:OGR:alternative_name": value of |
6369 | | * OGRFieldDefn::GetAlternativeNameRef()</li> |
6370 | | * <li>"GDAL:OGR:comment": value of OGRFieldDefn::GetComment()</li> |
6371 | | * <li>"GDAL:OGR:default": value of OGRFieldDefn::GetDefault()</li> |
6372 | | * <li>"GDAL:OGR:subtype": value of OGRFieldDefn::GetSubType()</li> |
6373 | | * <li>"GDAL:OGR:width": value of OGRFieldDefn::GetWidth() (serialized as a |
6374 | | * string)</li> |
6375 | | * <li>"GDAL:OGR:unique": value of OGRFieldDefn::IsUnique() (serialized as |
6376 | | * "true" or "false")</li> |
6377 | | * <li>"GDAL:OGR:domain_name": value of OGRFieldDefn::GetDomainName()</li> |
6378 | | * </ul> |
6379 | | * |
6380 | | * This method and CreateField() are mutually exclusive in the same session. |
6381 | | * |
6382 | | * This method is the same as the C function OGR_L_CreateFieldFromArrowSchema(). |
6383 | | * |
6384 | | * @param schema Schema of the field to create. |
6385 | | * @param papszOptions Options (none currently). Null terminated list, or nullptr. |
6386 | | * @return true in case of success |
6387 | | * @since 3.8 |
6388 | | */ |
6389 | | bool OGRLayer::CreateFieldFromArrowSchema(const struct ArrowSchema *schema, |
6390 | | CSLConstList papszOptions) |
6391 | 0 | { |
6392 | 0 | return CreateFieldFromArrowSchemaInternal(schema, std::string(), |
6393 | 0 | papszOptions); |
6394 | 0 | } |
6395 | | |
6396 | | /************************************************************************/ |
6397 | | /* OGR_L_CreateFieldFromArrowSchema() */ |
6398 | | /************************************************************************/ |
6399 | | |
6400 | | /** Creates a field from an ArrowSchema. |
6401 | | * |
6402 | | * This should only be used for attribute fields. Geometry fields should |
6403 | | * be created with CreateGeomField(). The FID field should also not be |
6404 | | * passed with this method. |
6405 | | * |
6406 | | * Contrary to the IsArrowSchemaSupported() and WriteArrowBatch() methods, the |
6407 | | * passed schema must be for an individual field, and thus, is *not* of type |
6408 | | * struct (format=+s) (unless writing a set of fields grouped together in the |
6409 | | * same structure). |
6410 | | * |
6411 | | * Additional field metadata can be speciffed through the ArrowSchema::metadata |
6412 | | * field with the potential following items: |
6413 | | * <ul> |
6414 | | * <li>"GDAL:OGR:alternative_name": value of |
6415 | | * OGRFieldDefn::GetAlternativeNameRef()</li> |
6416 | | * <li>"GDAL:OGR:comment": value of OGRFieldDefn::GetComment()</li> |
6417 | | * <li>"GDAL:OGR:default": value of OGRFieldDefn::GetDefault()</li> |
6418 | | * <li>"GDAL:OGR:subtype": value of OGRFieldDefn::GetSubType()</li> |
6419 | | * <li>"GDAL:OGR:width": value of OGRFieldDefn::GetWidth() (serialized as a |
6420 | | * string)</li> |
6421 | | * <li>"GDAL:OGR:unique": value of OGRFieldDefn::IsUnique() (serialized as |
6422 | | * "true" or "false")</li> |
6423 | | * <li>"GDAL:OGR:domain_name": value of OGRFieldDefn::GetDomainName()</li> |
6424 | | * </ul> |
6425 | | * |
6426 | | * This method and CreateField() are mutually exclusive in the same session. |
6427 | | * |
6428 | | * This method is the same as the C++ method OGRLayer::CreateFieldFromArrowSchema(). |
6429 | | * |
6430 | | * @param hLayer Layer. |
6431 | | * @param schema Schema of the field to create. |
6432 | | * @param papszOptions Options (none currently). Null terminated list, or nullptr. |
6433 | | * @return true in case of success |
6434 | | * @since 3.8 |
6435 | | */ |
6436 | | bool OGR_L_CreateFieldFromArrowSchema(OGRLayerH hLayer, |
6437 | | const struct ArrowSchema *schema, |
6438 | | char **papszOptions) |
6439 | 0 | { |
6440 | 0 | VALIDATE_POINTER1(hLayer, __func__, false); |
6441 | 0 | VALIDATE_POINTER1(schema, __func__, false); |
6442 | | |
6443 | 0 | return OGRLayer::FromHandle(hLayer)->CreateFieldFromArrowSchema( |
6444 | 0 | schema, papszOptions); |
6445 | 0 | } |
6446 | | |
6447 | | /************************************************************************/ |
6448 | | /* BuildOGRFieldInfo() */ |
6449 | | /************************************************************************/ |
6450 | | |
6451 | | constexpr int FID_COLUMN_SPECIAL_OGR_FIELD_IDX = -2; |
6452 | | |
6453 | | struct FieldInfo |
6454 | | { |
6455 | | std::string osName{}; |
6456 | | int iOGRFieldIdx = -1; |
6457 | | const char *format = nullptr; |
6458 | | OGRFieldType eNominalFieldType = |
6459 | | OFTMaxType; // OGR data type that would best match the Arrow type |
6460 | | OGRFieldType eTargetFieldType = |
6461 | | OFTMaxType; // actual OGR data type of the layer field |
6462 | | // OGR data type of the feature passed to FillFeature() |
6463 | | OGRFieldType eSetFeatureFieldType = OFTMaxType; |
6464 | | bool bIsGeomCol = false; |
6465 | | bool bUseDictionary = false; |
6466 | | bool bUseStringOptim = false; |
6467 | | int nWidthInBytes = 0; // only used for decimal fields |
6468 | | int nPrecision = 0; // only used for decimal fields |
6469 | | int nScale = 0; // only used for decimal fields |
6470 | | }; |
6471 | | |
6472 | | static bool BuildOGRFieldInfo( |
6473 | | const struct ArrowSchema *schema, struct ArrowArray *array, |
6474 | | const OGRFeatureDefn *poFeatureDefn, const std::string &osFieldPrefix, |
6475 | | const CPLStringList &aosNativeTypes, bool &bFallbackTypesUsed, |
6476 | | std::vector<FieldInfo> &asFieldInfo, const char *pszFIDName, |
6477 | | const char *pszGeomFieldName, OGRLayer *poLayer, |
6478 | | const std::map<std::string, std::string> &oMapArrowFieldNameToOGRFieldName, |
6479 | | const struct ArrowSchema *&schemaFIDColumn, |
6480 | | struct ArrowArray *&arrayFIDColumn) |
6481 | 0 | { |
6482 | 0 | const char *fieldName = schema->name; |
6483 | 0 | const char *format = schema->format; |
6484 | 0 | if (IsStructure(format)) |
6485 | 0 | { |
6486 | 0 | const std::string osNewPrefix(osFieldPrefix + fieldName + "."); |
6487 | 0 | for (int64_t i = 0; i < array->n_children; ++i) |
6488 | 0 | { |
6489 | 0 | if (!BuildOGRFieldInfo(schema->children[i], array->children[i], |
6490 | 0 | poFeatureDefn, osNewPrefix, aosNativeTypes, |
6491 | 0 | bFallbackTypesUsed, asFieldInfo, pszFIDName, |
6492 | 0 | pszGeomFieldName, poLayer, |
6493 | 0 | oMapArrowFieldNameToOGRFieldName, |
6494 | 0 | schemaFIDColumn, arrayFIDColumn)) |
6495 | 0 | { |
6496 | 0 | return false; |
6497 | 0 | } |
6498 | 0 | } |
6499 | 0 | return true; |
6500 | 0 | } |
6501 | | |
6502 | 0 | FieldInfo sInfo; |
6503 | |
|
6504 | 0 | if (schema->dictionary && |
6505 | 0 | !IsKnownCodedFieldDomain(poLayer, schema->metadata)) |
6506 | 0 | { |
6507 | 0 | if (!IsValidDictionaryIndexType(format)) |
6508 | 0 | { |
6509 | 0 | CPLError(CE_Failure, CPLE_NotSupported, |
6510 | 0 | "Dictionary only supported if the parent is of " |
6511 | 0 | "type [U]Int[8|16|32|64]"); |
6512 | 0 | return false; |
6513 | 0 | } |
6514 | | |
6515 | 0 | sInfo.bUseDictionary = true; |
6516 | 0 | schema = schema->dictionary; |
6517 | 0 | format = schema->format; |
6518 | 0 | array = array->dictionary; |
6519 | 0 | } |
6520 | | |
6521 | 0 | sInfo.osName = osFieldPrefix + fieldName; |
6522 | 0 | sInfo.format = format; |
6523 | 0 | if (pszFIDName && sInfo.osName == pszFIDName) |
6524 | 0 | { |
6525 | 0 | if (IsInt32(format) || IsInt64(format)) |
6526 | 0 | { |
6527 | 0 | sInfo.iOGRFieldIdx = FID_COLUMN_SPECIAL_OGR_FIELD_IDX; |
6528 | 0 | schemaFIDColumn = schema; |
6529 | 0 | arrayFIDColumn = array; |
6530 | 0 | } |
6531 | 0 | else |
6532 | 0 | { |
6533 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6534 | 0 | "FID column '%s' should be of Arrow format 'i' " |
6535 | 0 | "(int32) or 'l' (int64)", |
6536 | 0 | sInfo.osName.c_str()); |
6537 | 0 | return false; |
6538 | 0 | } |
6539 | 0 | } |
6540 | 0 | else |
6541 | 0 | { |
6542 | 0 | const std::string &osExpectedOGRFieldName = |
6543 | 0 | [&oMapArrowFieldNameToOGRFieldName, &sInfo]() -> const std::string & |
6544 | 0 | { |
6545 | 0 | const auto oIter = |
6546 | 0 | oMapArrowFieldNameToOGRFieldName.find(sInfo.osName); |
6547 | 0 | if (oIter != oMapArrowFieldNameToOGRFieldName.end()) |
6548 | 0 | return oIter->second; |
6549 | 0 | return sInfo.osName; |
6550 | 0 | }(); |
6551 | 0 | sInfo.iOGRFieldIdx = |
6552 | 0 | poFeatureDefn->GetFieldIndex(osExpectedOGRFieldName.c_str()); |
6553 | 0 | if (sInfo.iOGRFieldIdx >= 0) |
6554 | 0 | { |
6555 | 0 | bool bTypeOK = false; |
6556 | 0 | const auto eOGRType = |
6557 | 0 | poFeatureDefn->GetFieldDefn(sInfo.iOGRFieldIdx)->GetType(); |
6558 | 0 | sInfo.eTargetFieldType = eOGRType; |
6559 | 0 | for (const auto &sType : gasArrowTypesToOGR) |
6560 | 0 | { |
6561 | 0 | if (strcmp(format, sType.arrowType) == 0) |
6562 | 0 | { |
6563 | 0 | sInfo.bUseStringOptim = sType.eType == OFTString; |
6564 | 0 | sInfo.eNominalFieldType = sType.eType; |
6565 | 0 | if (eOGRType == sInfo.eNominalFieldType) |
6566 | 0 | { |
6567 | 0 | bTypeOK = true; |
6568 | 0 | break; |
6569 | 0 | } |
6570 | 0 | else if (eOGRType == OFTString) |
6571 | 0 | { |
6572 | 0 | bFallbackTypesUsed = true; |
6573 | 0 | bTypeOK = true; |
6574 | 0 | break; |
6575 | 0 | } |
6576 | 0 | else if (eOGRType == OFTInteger && |
6577 | 0 | sType.eType == OFTInteger64) |
6578 | 0 | { |
6579 | | // Potentially lossy. |
6580 | 0 | CPLDebug("OGR", |
6581 | 0 | "For field %s, writing from Arrow array of " |
6582 | 0 | "type Int64 into OGR Int32 field. " |
6583 | 0 | "Potentially loss conversion can happen", |
6584 | 0 | sInfo.osName.c_str()); |
6585 | 0 | bFallbackTypesUsed = true; |
6586 | 0 | bTypeOK = true; |
6587 | 0 | break; |
6588 | 0 | } |
6589 | 0 | else if (eOGRType == OFTInteger && sType.eType == OFTReal) |
6590 | 0 | { |
6591 | | // Potentially lossy. |
6592 | 0 | CPLDebug("OGR", |
6593 | 0 | "For field %s, writing from Arrow array of " |
6594 | 0 | "type Real into OGR Int32 field. " |
6595 | 0 | "Potentially loss conversion can happen", |
6596 | 0 | sInfo.osName.c_str()); |
6597 | 0 | bFallbackTypesUsed = true; |
6598 | 0 | bTypeOK = true; |
6599 | 0 | break; |
6600 | 0 | } |
6601 | 0 | else if (eOGRType == OFTInteger64 && sType.eType == OFTReal) |
6602 | 0 | { |
6603 | | // Potentially lossy. |
6604 | 0 | CPLDebug("OGR", |
6605 | 0 | "For field %s, writing from Arrow array of " |
6606 | 0 | "type Real into OGR Int64 field. " |
6607 | 0 | "Potentially loss conversion can happen", |
6608 | 0 | sInfo.osName.c_str()); |
6609 | 0 | bFallbackTypesUsed = true; |
6610 | 0 | bTypeOK = true; |
6611 | 0 | break; |
6612 | 0 | } |
6613 | 0 | else if (eOGRType == OFTReal && sType.eType == OFTInteger64) |
6614 | 0 | { |
6615 | | // Potentially lossy. |
6616 | 0 | CPLDebug("OGR", |
6617 | 0 | "For field %s, writing from Arrow array of " |
6618 | 0 | "type Int64 into OGR Real field. " |
6619 | 0 | "Potentially loss conversion can happen", |
6620 | 0 | sInfo.osName.c_str()); |
6621 | 0 | bFallbackTypesUsed = true; |
6622 | 0 | bTypeOK = true; |
6623 | 0 | break; |
6624 | 0 | } |
6625 | 0 | else if ((eOGRType == OFTInteger64 || |
6626 | 0 | eOGRType == OFTReal) && |
6627 | 0 | sType.eType == OFTInteger) |
6628 | 0 | { |
6629 | | // Non-lossy |
6630 | 0 | bFallbackTypesUsed = true; |
6631 | 0 | bTypeOK = true; |
6632 | 0 | break; |
6633 | 0 | } |
6634 | 0 | else if (eOGRType == OFTDateTime && |
6635 | 0 | sType.eType == OFTString) |
6636 | 0 | { |
6637 | 0 | bFallbackTypesUsed = true; |
6638 | 0 | bTypeOK = true; |
6639 | 0 | break; |
6640 | 0 | } |
6641 | 0 | else |
6642 | 0 | { |
6643 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6644 | 0 | "For field %s, OGR field type is %s whereas " |
6645 | 0 | "Arrow type implies %s", |
6646 | 0 | sInfo.osName.c_str(), |
6647 | 0 | OGR_GetFieldTypeName(eOGRType), |
6648 | 0 | OGR_GetFieldTypeName(sType.eType)); |
6649 | 0 | return false; |
6650 | 0 | } |
6651 | 0 | } |
6652 | 0 | } |
6653 | | |
6654 | 0 | if (!bTypeOK && IsMap(format)) |
6655 | 0 | { |
6656 | 0 | sInfo.eNominalFieldType = OFTString; |
6657 | 0 | if (eOGRType == sInfo.eNominalFieldType) |
6658 | 0 | { |
6659 | 0 | bTypeOK = true; |
6660 | 0 | } |
6661 | 0 | else |
6662 | 0 | { |
6663 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6664 | 0 | "For field %s, OGR field type is %s whereas " |
6665 | 0 | "Arrow type implies %s", |
6666 | 0 | sInfo.osName.c_str(), |
6667 | 0 | OGR_GetFieldTypeName(eOGRType), |
6668 | 0 | OGR_GetFieldTypeName(OFTString)); |
6669 | 0 | return false; |
6670 | 0 | } |
6671 | 0 | } |
6672 | | |
6673 | 0 | if (!bTypeOK && IsTimestamp(format)) |
6674 | 0 | { |
6675 | 0 | sInfo.eNominalFieldType = OFTDateTime; |
6676 | 0 | if (eOGRType == sInfo.eNominalFieldType) |
6677 | 0 | { |
6678 | 0 | bTypeOK = true; |
6679 | 0 | } |
6680 | 0 | else if (eOGRType == OFTString) |
6681 | 0 | { |
6682 | 0 | bFallbackTypesUsed = true; |
6683 | 0 | bTypeOK = true; |
6684 | 0 | } |
6685 | 0 | else |
6686 | 0 | { |
6687 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6688 | 0 | "For field %s, OGR field type is %s whereas " |
6689 | 0 | "Arrow type implies %s", |
6690 | 0 | sInfo.osName.c_str(), |
6691 | 0 | OGR_GetFieldTypeName(eOGRType), |
6692 | 0 | OGR_GetFieldTypeName(OFTDateTime)); |
6693 | 0 | return false; |
6694 | 0 | } |
6695 | 0 | } |
6696 | | |
6697 | 0 | if (!bTypeOK && IsFixedWidthBinary(format)) |
6698 | 0 | { |
6699 | 0 | sInfo.eNominalFieldType = OFTBinary; |
6700 | 0 | if (eOGRType == sInfo.eNominalFieldType) |
6701 | 0 | { |
6702 | 0 | bTypeOK = true; |
6703 | 0 | } |
6704 | 0 | else if (eOGRType == OFTString) |
6705 | 0 | { |
6706 | 0 | bFallbackTypesUsed = true; |
6707 | 0 | bTypeOK = true; |
6708 | 0 | } |
6709 | 0 | else |
6710 | 0 | { |
6711 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6712 | 0 | "For field %s, OGR field type is %s whereas " |
6713 | 0 | "Arrow type implies %s", |
6714 | 0 | sInfo.osName.c_str(), |
6715 | 0 | OGR_GetFieldTypeName(eOGRType), |
6716 | 0 | OGR_GetFieldTypeName(OFTBinary)); |
6717 | 0 | return false; |
6718 | 0 | } |
6719 | 0 | } |
6720 | | |
6721 | 0 | if (!bTypeOK && (IsList(format) || IsLargeList(format) || |
6722 | 0 | IsFixedSizeList(format))) |
6723 | 0 | { |
6724 | 0 | const char *childFormat = schema->children[0]->format; |
6725 | 0 | for (const auto &sType : gasListTypes) |
6726 | 0 | { |
6727 | 0 | if (childFormat[0] == sType.arrowLetter && |
6728 | 0 | childFormat[1] == 0) |
6729 | 0 | { |
6730 | 0 | sInfo.eNominalFieldType = sType.eType; |
6731 | 0 | if (eOGRType == sInfo.eNominalFieldType) |
6732 | 0 | { |
6733 | 0 | bTypeOK = true; |
6734 | 0 | break; |
6735 | 0 | } |
6736 | 0 | else if (eOGRType == OFTString) |
6737 | 0 | { |
6738 | 0 | bFallbackTypesUsed = true; |
6739 | 0 | bTypeOK = true; |
6740 | 0 | break; |
6741 | 0 | } |
6742 | 0 | else |
6743 | 0 | { |
6744 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6745 | 0 | "For field %s, OGR field type is %s " |
6746 | 0 | "whereas " |
6747 | 0 | "Arrow type implies %s", |
6748 | 0 | sInfo.osName.c_str(), |
6749 | 0 | OGR_GetFieldTypeName(eOGRType), |
6750 | 0 | OGR_GetFieldTypeName(sType.eType)); |
6751 | 0 | return false; |
6752 | 0 | } |
6753 | 0 | } |
6754 | 0 | } |
6755 | | |
6756 | 0 | if (!bTypeOK && IsDecimal(childFormat)) |
6757 | 0 | { |
6758 | 0 | if (!ParseDecimalFormat(childFormat, sInfo.nPrecision, |
6759 | 0 | sInfo.nScale, sInfo.nWidthInBytes)) |
6760 | 0 | { |
6761 | 0 | CPLError(CE_Failure, CPLE_AppDefined, "%s", |
6762 | 0 | (std::string("Invalid field format ") + |
6763 | 0 | childFormat + " for field " + osFieldPrefix + |
6764 | 0 | fieldName) |
6765 | 0 | .c_str()); |
6766 | 0 | return false; |
6767 | 0 | } |
6768 | | |
6769 | 0 | const char *pszError = GetErrorIfUnsupportedDecimal( |
6770 | 0 | sInfo.nWidthInBytes, sInfo.nPrecision); |
6771 | 0 | if (pszError) |
6772 | 0 | { |
6773 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", pszError); |
6774 | 0 | return false; |
6775 | 0 | } |
6776 | | |
6777 | 0 | sInfo.eNominalFieldType = OFTRealList; |
6778 | 0 | if (eOGRType == sInfo.eNominalFieldType) |
6779 | 0 | { |
6780 | 0 | bTypeOK = true; |
6781 | 0 | } |
6782 | 0 | else if (eOGRType == OFTString) |
6783 | 0 | { |
6784 | 0 | bFallbackTypesUsed = true; |
6785 | 0 | bTypeOK = true; |
6786 | 0 | } |
6787 | 0 | else |
6788 | 0 | { |
6789 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6790 | 0 | "For field %s, OGR field type is %s whereas " |
6791 | 0 | "Arrow type implies %s", |
6792 | 0 | sInfo.osName.c_str(), |
6793 | 0 | OGR_GetFieldTypeName(eOGRType), |
6794 | 0 | OGR_GetFieldTypeName(OFTRealList)); |
6795 | 0 | return false; |
6796 | 0 | } |
6797 | 0 | } |
6798 | | |
6799 | 0 | if (!bTypeOK && IsSupportForJSONObj(schema->children[0])) |
6800 | 0 | { |
6801 | 0 | sInfo.eNominalFieldType = OFTString; |
6802 | 0 | if (eOGRType == sInfo.eNominalFieldType) |
6803 | 0 | { |
6804 | 0 | bTypeOK = true; |
6805 | 0 | } |
6806 | 0 | else |
6807 | 0 | { |
6808 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6809 | 0 | "For field %s, OGR field type is %s whereas " |
6810 | 0 | "Arrow type implies %s", |
6811 | 0 | sInfo.osName.c_str(), |
6812 | 0 | OGR_GetFieldTypeName(eOGRType), |
6813 | 0 | OGR_GetFieldTypeName(OFTString)); |
6814 | 0 | return false; |
6815 | 0 | } |
6816 | 0 | } |
6817 | | |
6818 | 0 | if (!bTypeOK) |
6819 | 0 | { |
6820 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", |
6821 | 0 | ("List of type '" + std::string(childFormat) + |
6822 | 0 | "' for field " + osFieldPrefix + fieldName + |
6823 | 0 | " is not supported.") |
6824 | 0 | .c_str()); |
6825 | 0 | return false; |
6826 | 0 | } |
6827 | 0 | } |
6828 | | |
6829 | 0 | if (!bTypeOK && IsDecimal(format)) |
6830 | 0 | { |
6831 | 0 | if (!ParseDecimalFormat(format, sInfo.nPrecision, sInfo.nScale, |
6832 | 0 | sInfo.nWidthInBytes)) |
6833 | 0 | { |
6834 | 0 | CPLError(CE_Failure, CPLE_AppDefined, "%s", |
6835 | 0 | (std::string("Invalid field format ") + format + |
6836 | 0 | " for field " + osFieldPrefix + fieldName) |
6837 | 0 | .c_str()); |
6838 | 0 | return false; |
6839 | 0 | } |
6840 | | |
6841 | 0 | const char *pszError = GetErrorIfUnsupportedDecimal( |
6842 | 0 | sInfo.nWidthInBytes, sInfo.nPrecision); |
6843 | 0 | if (pszError) |
6844 | 0 | { |
6845 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", pszError); |
6846 | 0 | return false; |
6847 | 0 | } |
6848 | | |
6849 | 0 | sInfo.eNominalFieldType = OFTReal; |
6850 | 0 | if (eOGRType == sInfo.eNominalFieldType) |
6851 | 0 | { |
6852 | 0 | bTypeOK = true; |
6853 | 0 | } |
6854 | 0 | else if (eOGRType == OFTString) |
6855 | 0 | { |
6856 | 0 | bFallbackTypesUsed = true; |
6857 | 0 | bTypeOK = true; |
6858 | 0 | } |
6859 | 0 | else |
6860 | 0 | { |
6861 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6862 | 0 | "For field %s, OGR field type is %s whereas " |
6863 | 0 | "Arrow type implies %s", |
6864 | 0 | sInfo.osName.c_str(), |
6865 | 0 | OGR_GetFieldTypeName(eOGRType), |
6866 | 0 | OGR_GetFieldTypeName(OFTReal)); |
6867 | 0 | return false; |
6868 | 0 | } |
6869 | 0 | } |
6870 | | |
6871 | 0 | if (!bTypeOK) |
6872 | 0 | { |
6873 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", |
6874 | 0 | ("Type '" + std::string(format) + "' for field " + |
6875 | 0 | osFieldPrefix + fieldName + " is not supported.") |
6876 | 0 | .c_str()); |
6877 | 0 | return false; |
6878 | 0 | } |
6879 | 0 | } |
6880 | 0 | else |
6881 | 0 | { |
6882 | 0 | sInfo.iOGRFieldIdx = poFeatureDefn->GetGeomFieldIndex( |
6883 | 0 | osExpectedOGRFieldName.c_str()); |
6884 | 0 | if (sInfo.iOGRFieldIdx < 0) |
6885 | 0 | { |
6886 | 0 | if (pszGeomFieldName && pszGeomFieldName == sInfo.osName) |
6887 | 0 | { |
6888 | 0 | if (poFeatureDefn->GetGeomFieldCount() == 0) |
6889 | 0 | { |
6890 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6891 | 0 | "Cannot find OGR geometry field for Arrow " |
6892 | 0 | "array %s", |
6893 | 0 | sInfo.osName.c_str()); |
6894 | 0 | return false; |
6895 | 0 | } |
6896 | 0 | sInfo.iOGRFieldIdx = 0; |
6897 | 0 | } |
6898 | 0 | else |
6899 | 0 | { |
6900 | | // Check if ARROW:extension:name = ogc.wkb or geoarrow.wkb |
6901 | 0 | const char *pabyMetadata = schema->metadata; |
6902 | 0 | if (pabyMetadata) |
6903 | 0 | { |
6904 | 0 | const auto oMetadata = |
6905 | 0 | OGRParseArrowMetadata(pabyMetadata); |
6906 | 0 | auto oIter = oMetadata.find(ARROW_EXTENSION_NAME_KEY); |
6907 | 0 | if (oIter != oMetadata.end() && |
6908 | 0 | (oIter->second == EXTENSION_NAME_OGC_WKB || |
6909 | 0 | oIter->second == EXTENSION_NAME_GEOARROW_WKB)) |
6910 | 0 | { |
6911 | 0 | if (poFeatureDefn->GetGeomFieldCount() == 0) |
6912 | 0 | { |
6913 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6914 | 0 | "Cannot find OGR geometry field " |
6915 | 0 | "for Arrow array %s", |
6916 | 0 | sInfo.osName.c_str()); |
6917 | 0 | return false; |
6918 | 0 | } |
6919 | 0 | sInfo.iOGRFieldIdx = 0; |
6920 | 0 | } |
6921 | 0 | } |
6922 | 0 | } |
6923 | | |
6924 | 0 | if (sInfo.iOGRFieldIdx < 0) |
6925 | 0 | { |
6926 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6927 | 0 | "Cannot find OGR field for Arrow array %s", |
6928 | 0 | sInfo.osName.c_str()); |
6929 | 0 | return false; |
6930 | 0 | } |
6931 | 0 | } |
6932 | | |
6933 | 0 | if (!IsBinary(format) && !IsLargeBinary(format)) |
6934 | 0 | { |
6935 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
6936 | 0 | "Geometry column '%s' should be of Arrow format " |
6937 | 0 | "'z' (binary) or 'Z' (large binary)", |
6938 | 0 | sInfo.osName.c_str()); |
6939 | 0 | return false; |
6940 | 0 | } |
6941 | 0 | sInfo.bIsGeomCol = true; |
6942 | 0 | } |
6943 | 0 | } |
6944 | | |
6945 | 0 | asFieldInfo.emplace_back(std::move(sInfo)); |
6946 | 0 | return true; |
6947 | 0 | } |
6948 | | |
6949 | | /************************************************************************/ |
6950 | | /* GetUInt64Value() */ |
6951 | | /************************************************************************/ |
6952 | | |
6953 | | static inline uint64_t GetUInt64Value(const struct ArrowSchema *schema, |
6954 | | const struct ArrowArray *array, |
6955 | | size_t iFeature) |
6956 | 0 | { |
6957 | 0 | uint64_t nVal = 0; |
6958 | 0 | CPLAssert(schema->format[1] == 0); |
6959 | 0 | switch (schema->format[0]) |
6960 | 0 | { |
6961 | 0 | case ARROW_LETTER_INT8: |
6962 | 0 | nVal = GetValue<int8_t>(array, iFeature); |
6963 | 0 | break; |
6964 | 0 | case ARROW_LETTER_UINT8: |
6965 | 0 | nVal = GetValue<uint8_t>(array, iFeature); |
6966 | 0 | break; |
6967 | 0 | case ARROW_LETTER_INT16: |
6968 | 0 | nVal = GetValue<int16_t>(array, iFeature); |
6969 | 0 | break; |
6970 | 0 | case ARROW_LETTER_UINT16: |
6971 | 0 | nVal = GetValue<uint16_t>(array, iFeature); |
6972 | 0 | break; |
6973 | 0 | case ARROW_LETTER_INT32: |
6974 | 0 | nVal = GetValue<int32_t>(array, iFeature); |
6975 | 0 | break; |
6976 | 0 | case ARROW_LETTER_UINT32: |
6977 | 0 | nVal = GetValue<uint32_t>(array, iFeature); |
6978 | 0 | break; |
6979 | 0 | case ARROW_LETTER_INT64: |
6980 | 0 | nVal = GetValue<int64_t>(array, iFeature); |
6981 | 0 | break; |
6982 | 0 | case ARROW_LETTER_UINT64: |
6983 | 0 | nVal = GetValue<uint64_t>(array, iFeature); |
6984 | 0 | break; |
6985 | 0 | default: |
6986 | | // Shouldn't happen given checks in BuildOGRFieldInfo() |
6987 | 0 | CPLAssert(false); |
6988 | 0 | break; |
6989 | 0 | } |
6990 | 0 | return nVal; |
6991 | 0 | } |
6992 | | |
6993 | | /************************************************************************/ |
6994 | | /* GetWorkingBufferSize() */ |
6995 | | /************************************************************************/ |
6996 | | |
6997 | | static size_t GetWorkingBufferSize(const struct ArrowSchema *schema, |
6998 | | const struct ArrowArray *array, |
6999 | | size_t iFeature, int &iArrowIdxInOut, |
7000 | | const std::vector<FieldInfo> &asFieldInfo) |
7001 | 0 | { |
7002 | 0 | const char *fieldName = schema->name; |
7003 | 0 | const char *format = schema->format; |
7004 | 0 | if (IsStructure(format)) |
7005 | 0 | { |
7006 | 0 | size_t nRet = 0; |
7007 | 0 | for (int64_t i = 0; i < array->n_children; ++i) |
7008 | 0 | { |
7009 | 0 | nRet += GetWorkingBufferSize( |
7010 | 0 | schema->children[i], array->children[i], |
7011 | 0 | iFeature + static_cast<size_t>(array->offset), iArrowIdxInOut, |
7012 | 0 | asFieldInfo); |
7013 | 0 | } |
7014 | 0 | return nRet; |
7015 | 0 | } |
7016 | 0 | const int iArrowIdx = iArrowIdxInOut; |
7017 | 0 | ++iArrowIdxInOut; |
7018 | |
|
7019 | 0 | if (!asFieldInfo[iArrowIdx].bUseStringOptim) |
7020 | 0 | return 0; |
7021 | | |
7022 | 0 | const uint8_t *pabyValidity = |
7023 | 0 | static_cast<const uint8_t *>(array->buffers[0]); |
7024 | 0 | if (array->null_count != 0 && pabyValidity && |
7025 | 0 | !TestBit(pabyValidity, static_cast<size_t>(iFeature + array->offset))) |
7026 | 0 | { |
7027 | | // empty string |
7028 | 0 | return 0; |
7029 | 0 | } |
7030 | | |
7031 | 0 | if (asFieldInfo[iArrowIdx].bUseDictionary) |
7032 | 0 | { |
7033 | 0 | const uint64_t nDictIdx = GetUInt64Value(schema, array, iFeature); |
7034 | 0 | const auto dictArray = array->dictionary; |
7035 | 0 | if (nDictIdx >= static_cast<uint64_t>(dictArray->length)) |
7036 | 0 | { |
7037 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
7038 | 0 | "Feature %" PRIu64 |
7039 | 0 | ", field %s: invalid dictionary index: %" PRIu64, |
7040 | 0 | static_cast<uint64_t>(iFeature), fieldName, nDictIdx); |
7041 | 0 | return 0; |
7042 | 0 | } |
7043 | | |
7044 | 0 | array = dictArray; |
7045 | 0 | schema = schema->dictionary; |
7046 | 0 | format = schema->format; |
7047 | 0 | iFeature = static_cast<size_t>(nDictIdx); |
7048 | 0 | } |
7049 | | |
7050 | 0 | if (IsString(format)) |
7051 | 0 | { |
7052 | 0 | const auto *panOffsets = |
7053 | 0 | static_cast<const uint32_t *>(array->buffers[1]) + array->offset; |
7054 | 0 | return 1 + (panOffsets[iFeature + 1] - panOffsets[iFeature]); |
7055 | 0 | } |
7056 | 0 | else if (IsLargeString(format)) |
7057 | 0 | { |
7058 | 0 | const auto *panOffsets = |
7059 | 0 | static_cast<const uint64_t *>(array->buffers[1]) + array->offset; |
7060 | 0 | return 1 + static_cast<size_t>(panOffsets[iFeature + 1] - |
7061 | 0 | panOffsets[iFeature]); |
7062 | 0 | } |
7063 | 0 | return 0; |
7064 | 0 | } |
7065 | | |
7066 | | /************************************************************************/ |
7067 | | /* FillField() */ |
7068 | | /************************************************************************/ |
7069 | | |
7070 | | template <typename ArrowType, typename OGRType = ArrowType> |
7071 | | inline static void FillField(const struct ArrowArray *array, int iOGRFieldIdx, |
7072 | | size_t iFeature, OGRFeature &oFeature) |
7073 | 0 | { |
7074 | 0 | const auto *panValues = static_cast<const ArrowType *>(array->buffers[1]); |
7075 | 0 | oFeature.SetFieldSameTypeUnsafe( |
7076 | 0 | iOGRFieldIdx, |
7077 | 0 | static_cast<OGRType>(panValues[iFeature + array->offset])); |
7078 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<signed char, signed char>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<unsigned char, unsigned char>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<short, short>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<unsigned short, unsigned short>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<int, int>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<unsigned int, long long>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<long, long long>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<unsigned long, double>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<float, float>(ArrowArray const*, int, unsigned long, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillField<double, double>(ArrowArray const*, int, unsigned long, OGRFeature&) |
7079 | | |
7080 | | /************************************************************************/ |
7081 | | /* FillFieldString() */ |
7082 | | /************************************************************************/ |
7083 | | |
7084 | | template <typename OffsetType> |
7085 | | inline static void |
7086 | | FillFieldString(const struct ArrowArray *array, int iOGRFieldIdx, |
7087 | | size_t iFeature, int iArrowIdx, |
7088 | | const std::vector<FieldInfo> &asFieldInfo, |
7089 | | std::string &osWorkingBuffer, OGRFeature &oFeature) |
7090 | 0 | { |
7091 | 0 | const auto *panOffsets = |
7092 | 0 | static_cast<const OffsetType *>(array->buffers[1]) + array->offset; |
7093 | 0 | const char *pszStr = static_cast<const char *>(array->buffers[2]); |
7094 | 0 | const size_t nLen = |
7095 | 0 | static_cast<size_t>(panOffsets[iFeature + 1] - panOffsets[iFeature]); |
7096 | 0 | if (asFieldInfo[iArrowIdx].bUseStringOptim) |
7097 | 0 | { |
7098 | 0 | oFeature.SetFieldSameTypeUnsafe( |
7099 | 0 | iOGRFieldIdx, &osWorkingBuffer[0] + osWorkingBuffer.size()); |
7100 | 0 | osWorkingBuffer.append(pszStr + panOffsets[iFeature], nLen); |
7101 | 0 | osWorkingBuffer.push_back(0); // append null character |
7102 | 0 | } |
7103 | 0 | else |
7104 | 0 | { |
7105 | 0 | const std::string osTmp(pszStr, nLen); |
7106 | 0 | oFeature.SetField(iOGRFieldIdx, osTmp.c_str()); |
7107 | 0 | } |
7108 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldString<unsigned int>(ArrowArray const*, int, unsigned long, int, std::__1::vector<FieldInfo, std::__1::allocator<FieldInfo> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >&, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:void FillFieldString<unsigned long>(ArrowArray const*, int, unsigned long, int, std::__1::vector<FieldInfo, std::__1::allocator<FieldInfo> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >&, OGRFeature&) |
7109 | | |
7110 | | /************************************************************************/ |
7111 | | /* FillFieldBinary() */ |
7112 | | /************************************************************************/ |
7113 | | |
7114 | | template <typename OffsetType> |
7115 | | inline static bool |
7116 | | FillFieldBinary(const struct ArrowArray *array, int iOGRFieldIdx, |
7117 | | size_t iFeature, int iArrowIdx, |
7118 | | const std::vector<FieldInfo> &asFieldInfo, |
7119 | | const std::string &osFieldPrefix, const char *pszFieldName, |
7120 | | OGRFeature &oFeature) |
7121 | 0 | { |
7122 | 0 | const auto *panOffsets = |
7123 | 0 | static_cast<const OffsetType *>(array->buffers[1]) + array->offset; |
7124 | 0 | const GByte *pabyData = static_cast<const GByte *>(array->buffers[2]) + |
7125 | 0 | static_cast<size_t>(panOffsets[iFeature]); |
7126 | 0 | const size_t nLen = |
7127 | 0 | static_cast<size_t>(panOffsets[iFeature + 1] - panOffsets[iFeature]); |
7128 | 0 | if (asFieldInfo[iArrowIdx].bIsGeomCol) |
7129 | 0 | { |
7130 | 0 | size_t nBytesConsumedOut = 0; |
7131 | | |
7132 | | // Check if we can reuse the existing geometry, to save dynamic memory |
7133 | | // allocations. |
7134 | 0 | if (nLen >= 5 && pabyData[0] == wkbNDR && pabyData[1] <= wkbTriangle && |
7135 | 0 | pabyData[2] == 0 && pabyData[3] == 0 && pabyData[4] == 0) |
7136 | 0 | { |
7137 | 0 | const auto poExistingGeom = oFeature.GetGeomFieldRef(iOGRFieldIdx); |
7138 | 0 | if (poExistingGeom && |
7139 | 0 | poExistingGeom->getGeometryType() == pabyData[1]) |
7140 | 0 | { |
7141 | 0 | poExistingGeom->importFromWkb(pabyData, nLen, wkbVariantIso, |
7142 | 0 | nBytesConsumedOut); |
7143 | 0 | return true; |
7144 | 0 | } |
7145 | 0 | } |
7146 | | |
7147 | 0 | OGRGeometry *poGeometry = nullptr; |
7148 | 0 | OGRGeometryFactory::createFromWkb(pabyData, nullptr, &poGeometry, nLen, |
7149 | 0 | wkbVariantIso, nBytesConsumedOut); |
7150 | 0 | oFeature.SetGeomFieldDirectly(iOGRFieldIdx, poGeometry); |
7151 | 0 | } |
7152 | 0 | else |
7153 | 0 | { |
7154 | 0 | if (nLen > static_cast<size_t>(std::numeric_limits<int>::max())) |
7155 | 0 | { |
7156 | 0 | CPLError(CE_Failure, CPLE_NotSupported, |
7157 | 0 | "Content for field %s%s is too large", |
7158 | 0 | osFieldPrefix.c_str(), pszFieldName); |
7159 | 0 | return false; |
7160 | 0 | } |
7161 | 0 | oFeature.SetField(iOGRFieldIdx, static_cast<int>(nLen), pabyData); |
7162 | 0 | } |
7163 | 0 | return true; |
7164 | 0 | } Unexecuted instantiation: ogrlayerarrow.cpp:bool FillFieldBinary<unsigned int>(ArrowArray const*, int, unsigned long, int, std::__1::vector<FieldInfo, std::__1::allocator<FieldInfo> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, char const*, OGRFeature&) Unexecuted instantiation: ogrlayerarrow.cpp:bool FillFieldBinary<unsigned long>(ArrowArray const*, int, unsigned long, int, std::__1::vector<FieldInfo, std::__1::allocator<FieldInfo> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, char const*, OGRFeature&) |
7165 | | |
7166 | | /************************************************************************/ |
7167 | | /* FillFeature() */ |
7168 | | /************************************************************************/ |
7169 | | |
7170 | | static bool FillFeature(OGRLayer *poLayer, const struct ArrowSchema *schema, |
7171 | | const struct ArrowArray *array, |
7172 | | const std::string &osFieldPrefix, size_t iFeature, |
7173 | | int &iArrowIdxInOut, |
7174 | | const std::vector<FieldInfo> &asFieldInfo, |
7175 | | OGRFeature &oFeature, std::string &osWorkingBuffer) |
7176 | | |
7177 | 0 | { |
7178 | 0 | const char *fieldName = schema->name; |
7179 | 0 | const char *format = schema->format; |
7180 | 0 | if (IsStructure(format)) |
7181 | 0 | { |
7182 | 0 | const std::string osNewPrefix(osFieldPrefix + fieldName + "."); |
7183 | 0 | for (int64_t i = 0; i < array->n_children; ++i) |
7184 | 0 | { |
7185 | 0 | if (!FillFeature( |
7186 | 0 | poLayer, schema->children[i], array->children[i], |
7187 | 0 | osNewPrefix, iFeature + static_cast<size_t>(array->offset), |
7188 | 0 | iArrowIdxInOut, asFieldInfo, oFeature, osWorkingBuffer)) |
7189 | 0 | return false; |
7190 | 0 | } |
7191 | 0 | return true; |
7192 | 0 | } |
7193 | 0 | const int iArrowIdx = iArrowIdxInOut; |
7194 | 0 | ++iArrowIdxInOut; |
7195 | 0 | const int iOGRFieldIdx = asFieldInfo[iArrowIdx].iOGRFieldIdx; |
7196 | |
|
7197 | 0 | if (asFieldInfo[iArrowIdx].bUseDictionary) |
7198 | 0 | { |
7199 | 0 | format = schema->dictionary->format; |
7200 | 0 | } |
7201 | |
|
7202 | 0 | if (array->null_count != 0) |
7203 | 0 | { |
7204 | 0 | const uint8_t *pabyValidity = |
7205 | 0 | static_cast<const uint8_t *>(array->buffers[0]); |
7206 | 0 | if (pabyValidity && |
7207 | 0 | !TestBit(pabyValidity, |
7208 | 0 | static_cast<size_t>(iFeature + array->offset))) |
7209 | 0 | { |
7210 | 0 | if (iOGRFieldIdx == FID_COLUMN_SPECIAL_OGR_FIELD_IDX) |
7211 | 0 | oFeature.SetFID(OGRNullFID); |
7212 | 0 | else if (asFieldInfo[iArrowIdx].bIsGeomCol) |
7213 | 0 | oFeature.SetGeomFieldDirectly(iOGRFieldIdx, nullptr); |
7214 | 0 | else if (asFieldInfo[iArrowIdx].eSetFeatureFieldType == OFTString) |
7215 | 0 | { |
7216 | 0 | OGRField *psField = oFeature.GetRawFieldRef(iOGRFieldIdx); |
7217 | 0 | if (!asFieldInfo[iArrowIdx].bUseStringOptim) |
7218 | 0 | { |
7219 | 0 | if (IsValidField(psField)) |
7220 | 0 | { |
7221 | 0 | CPLFree(psField->String); |
7222 | 0 | OGR_RawField_SetNull(psField); |
7223 | 0 | } |
7224 | 0 | } |
7225 | 0 | else |
7226 | 0 | { |
7227 | 0 | OGR_RawField_SetNull(psField); |
7228 | 0 | } |
7229 | 0 | } |
7230 | 0 | else |
7231 | 0 | { |
7232 | 0 | OGRField *psField = oFeature.GetRawFieldRef(iOGRFieldIdx); |
7233 | 0 | switch (asFieldInfo[iArrowIdx].eSetFeatureFieldType) |
7234 | 0 | { |
7235 | 0 | case OFTRealList: |
7236 | 0 | case OFTIntegerList: |
7237 | 0 | case OFTInteger64List: |
7238 | 0 | if (IsValidField(psField)) |
7239 | 0 | CPLFree(psField->IntegerList.paList); |
7240 | 0 | break; |
7241 | | |
7242 | 0 | case OFTStringList: |
7243 | 0 | if (IsValidField(psField)) |
7244 | 0 | CSLDestroy(psField->StringList.paList); |
7245 | 0 | break; |
7246 | | |
7247 | 0 | case OFTBinary: |
7248 | 0 | if (IsValidField(psField)) |
7249 | 0 | CPLFree(psField->Binary.paData); |
7250 | 0 | break; |
7251 | | |
7252 | 0 | default: |
7253 | 0 | break; |
7254 | 0 | } |
7255 | 0 | OGR_RawField_SetNull(psField); |
7256 | 0 | } |
7257 | 0 | return true; |
7258 | 0 | } |
7259 | 0 | } |
7260 | | |
7261 | 0 | if (asFieldInfo[iArrowIdx].bUseDictionary) |
7262 | 0 | { |
7263 | 0 | const uint64_t nDictIdx = GetUInt64Value(schema, array, iFeature); |
7264 | 0 | auto dictArray = array->dictionary; |
7265 | 0 | if (nDictIdx >= static_cast<uint64_t>(dictArray->length)) |
7266 | 0 | { |
7267 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
7268 | 0 | "Feature %" PRIu64 |
7269 | 0 | ", field %s: invalid dictionary index: %" PRIu64, |
7270 | 0 | static_cast<uint64_t>(iFeature), |
7271 | 0 | (osFieldPrefix + fieldName).c_str(), nDictIdx); |
7272 | 0 | return false; |
7273 | 0 | } |
7274 | 0 | array = dictArray; |
7275 | 0 | schema = schema->dictionary; |
7276 | 0 | iFeature = static_cast<size_t>(nDictIdx); |
7277 | 0 | } |
7278 | | |
7279 | 0 | if (IsBoolean(format)) |
7280 | 0 | { |
7281 | 0 | const uint8_t *pabyValues = |
7282 | 0 | static_cast<const uint8_t *>(array->buffers[1]); |
7283 | 0 | oFeature.SetFieldSameTypeUnsafe( |
7284 | 0 | iOGRFieldIdx, |
7285 | 0 | TestBit(pabyValues, static_cast<size_t>(iFeature + array->offset)) |
7286 | 0 | ? 1 |
7287 | 0 | : 0); |
7288 | 0 | return true; |
7289 | 0 | } |
7290 | 0 | else if (IsInt8(format)) |
7291 | 0 | { |
7292 | 0 | FillField<int8_t>(array, iOGRFieldIdx, iFeature, oFeature); |
7293 | 0 | return true; |
7294 | 0 | } |
7295 | 0 | else if (IsUInt8(format)) |
7296 | 0 | { |
7297 | 0 | FillField<uint8_t>(array, iOGRFieldIdx, iFeature, oFeature); |
7298 | 0 | return true; |
7299 | 0 | } |
7300 | 0 | else if (IsInt16(format)) |
7301 | 0 | { |
7302 | 0 | FillField<int16_t>(array, iOGRFieldIdx, iFeature, oFeature); |
7303 | 0 | return true; |
7304 | 0 | } |
7305 | 0 | else if (IsUInt16(format)) |
7306 | 0 | { |
7307 | 0 | FillField<uint16_t>(array, iOGRFieldIdx, iFeature, oFeature); |
7308 | 0 | return true; |
7309 | 0 | } |
7310 | 0 | else if (IsInt32(format)) |
7311 | 0 | { |
7312 | 0 | if (iOGRFieldIdx == FID_COLUMN_SPECIAL_OGR_FIELD_IDX) |
7313 | 0 | { |
7314 | 0 | const auto *panValues = |
7315 | 0 | static_cast<const int32_t *>(array->buffers[1]); |
7316 | 0 | oFeature.SetFID(panValues[iFeature + array->offset]); |
7317 | 0 | } |
7318 | 0 | else |
7319 | 0 | { |
7320 | 0 | FillField<int32_t>(array, iOGRFieldIdx, iFeature, oFeature); |
7321 | 0 | } |
7322 | 0 | return true; |
7323 | 0 | } |
7324 | 0 | else if (IsUInt32(format)) |
7325 | 0 | { |
7326 | 0 | FillField<uint32_t, GIntBig>(array, iOGRFieldIdx, iFeature, oFeature); |
7327 | 0 | return true; |
7328 | 0 | } |
7329 | 0 | else if (IsInt64(format)) |
7330 | 0 | { |
7331 | 0 | if (iOGRFieldIdx == FID_COLUMN_SPECIAL_OGR_FIELD_IDX) |
7332 | 0 | { |
7333 | 0 | const auto *panValues = |
7334 | 0 | static_cast<const int64_t *>(array->buffers[1]); |
7335 | 0 | oFeature.SetFID(panValues[iFeature + array->offset]); |
7336 | 0 | } |
7337 | 0 | else |
7338 | 0 | { |
7339 | 0 | FillField<int64_t, GIntBig>(array, iOGRFieldIdx, iFeature, |
7340 | 0 | oFeature); |
7341 | 0 | } |
7342 | 0 | return true; |
7343 | 0 | } |
7344 | 0 | else if (IsUInt64(format)) |
7345 | 0 | { |
7346 | 0 | FillField<uint64_t, double>(array, iOGRFieldIdx, iFeature, oFeature); |
7347 | 0 | return true; |
7348 | 0 | } |
7349 | 0 | else if (IsFloat32(format)) |
7350 | 0 | { |
7351 | 0 | FillField<float>(array, iOGRFieldIdx, iFeature, oFeature); |
7352 | 0 | return true; |
7353 | 0 | } |
7354 | 0 | else if (IsFloat64(format)) |
7355 | 0 | { |
7356 | 0 | FillField<double>(array, iOGRFieldIdx, iFeature, oFeature); |
7357 | 0 | return true; |
7358 | 0 | } |
7359 | 0 | else if (IsString(format)) |
7360 | 0 | { |
7361 | 0 | FillFieldString<uint32_t>(array, iOGRFieldIdx, iFeature, iArrowIdx, |
7362 | 0 | asFieldInfo, osWorkingBuffer, oFeature); |
7363 | 0 | return true; |
7364 | 0 | } |
7365 | 0 | else if (IsLargeString(format)) |
7366 | 0 | { |
7367 | 0 | FillFieldString<uint64_t>(array, iOGRFieldIdx, iFeature, iArrowIdx, |
7368 | 0 | asFieldInfo, osWorkingBuffer, oFeature); |
7369 | 0 | return true; |
7370 | 0 | } |
7371 | 0 | else if (IsBinary(format)) |
7372 | 0 | { |
7373 | 0 | return FillFieldBinary<uint32_t>(array, iOGRFieldIdx, iFeature, |
7374 | 0 | iArrowIdx, asFieldInfo, osFieldPrefix, |
7375 | 0 | fieldName, oFeature); |
7376 | 0 | } |
7377 | 0 | else if (IsLargeBinary(format)) |
7378 | 0 | { |
7379 | 0 | return FillFieldBinary<uint64_t>(array, iOGRFieldIdx, iFeature, |
7380 | 0 | iArrowIdx, asFieldInfo, osFieldPrefix, |
7381 | 0 | fieldName, oFeature); |
7382 | 0 | } |
7383 | 0 | else if (asFieldInfo[iArrowIdx].nPrecision > 0) |
7384 | 0 | { |
7385 | | // fits on a int64 |
7386 | 0 | CPLAssert(asFieldInfo[iArrowIdx].nPrecision <= 19); |
7387 | | // either 128 or 256 bits |
7388 | 0 | CPLAssert((asFieldInfo[iArrowIdx].nWidthInBytes % 8) == 0); |
7389 | 0 | const int nWidthIn64BitWord = asFieldInfo[iArrowIdx].nWidthInBytes / 8; |
7390 | |
|
7391 | 0 | if (IsList(format)) |
7392 | 0 | { |
7393 | 0 | const auto panOffsets = |
7394 | 0 | static_cast<const uint32_t *>(array->buffers[1]) + |
7395 | 0 | array->offset; |
7396 | 0 | const auto childArray = array->children[0]; |
7397 | 0 | std::vector<double> aValues; |
7398 | 0 | for (auto i = panOffsets[iFeature]; i < panOffsets[iFeature + 1]; |
7399 | 0 | ++i) |
7400 | 0 | { |
7401 | 0 | aValues.push_back(GetValueDecimal(childArray, nWidthIn64BitWord, |
7402 | 0 | asFieldInfo[iArrowIdx].nScale, |
7403 | 0 | i)); |
7404 | 0 | } |
7405 | 0 | oFeature.SetField(iOGRFieldIdx, static_cast<int>(aValues.size()), |
7406 | 0 | aValues.data()); |
7407 | 0 | return true; |
7408 | 0 | } |
7409 | 0 | else if (IsLargeList(format)) |
7410 | 0 | { |
7411 | 0 | const auto panOffsets = |
7412 | 0 | static_cast<const uint64_t *>(array->buffers[1]) + |
7413 | 0 | array->offset; |
7414 | 0 | const auto childArray = array->children[0]; |
7415 | 0 | std::vector<double> aValues; |
7416 | 0 | for (auto i = static_cast<size_t>(panOffsets[iFeature]); |
7417 | 0 | i < static_cast<size_t>(panOffsets[iFeature + 1]); ++i) |
7418 | 0 | { |
7419 | 0 | aValues.push_back(GetValueDecimal(childArray, nWidthIn64BitWord, |
7420 | 0 | asFieldInfo[iArrowIdx].nScale, |
7421 | 0 | i)); |
7422 | 0 | } |
7423 | 0 | oFeature.SetField(iOGRFieldIdx, static_cast<int>(aValues.size()), |
7424 | 0 | aValues.data()); |
7425 | 0 | return true; |
7426 | 0 | } |
7427 | 0 | else if (IsFixedSizeList(format)) |
7428 | 0 | { |
7429 | 0 | const int nVals = GetFixedSizeList(format); |
7430 | 0 | const auto childArray = array->children[0]; |
7431 | 0 | std::vector<double> aValues; |
7432 | 0 | for (int i = 0; i < nVals; ++i) |
7433 | 0 | { |
7434 | 0 | aValues.push_back(GetValueDecimal(childArray, nWidthIn64BitWord, |
7435 | 0 | asFieldInfo[iArrowIdx].nScale, |
7436 | 0 | iFeature * nVals + i)); |
7437 | 0 | } |
7438 | 0 | oFeature.SetField(iOGRFieldIdx, nVals, aValues.data()); |
7439 | 0 | return true; |
7440 | 0 | } |
7441 | | |
7442 | 0 | CPLAssert(format[0] == ARROW_LETTER_DECIMAL); |
7443 | | |
7444 | 0 | oFeature.SetFieldSameTypeUnsafe( |
7445 | 0 | iOGRFieldIdx, |
7446 | 0 | GetValueDecimal(array, nWidthIn64BitWord, |
7447 | 0 | asFieldInfo[iArrowIdx].nScale, iFeature)); |
7448 | 0 | return true; |
7449 | 0 | } |
7450 | 0 | else if (SetFieldForOtherFormats( |
7451 | 0 | oFeature, iOGRFieldIdx, |
7452 | 0 | static_cast<size_t>(iFeature + array->offset), schema, array)) |
7453 | 0 | { |
7454 | 0 | return true; |
7455 | 0 | } |
7456 | | |
7457 | 0 | CPLError(CE_Failure, CPLE_NotSupported, "%s", |
7458 | 0 | ("Type '" + std::string(format) + "' for field " + osFieldPrefix + |
7459 | 0 | fieldName + " is not supported.") |
7460 | 0 | .c_str()); |
7461 | 0 | return false; |
7462 | 0 | } |
7463 | | |
7464 | | /************************************************************************/ |
7465 | | /* OGRLayer::WriteArrowBatch() */ |
7466 | | /************************************************************************/ |
7467 | | |
7468 | | // clang-format off |
7469 | | /** Writes a batch of rows from an ArrowArray. |
7470 | | * |
7471 | | * This is semantically close to calling CreateFeature() with multiple features |
7472 | | * at once. |
7473 | | * |
7474 | | * The ArrowArray must be of type struct (format=+s), and its children generally |
7475 | | * map to a OGR attribute or geometry field (unless they are struct themselves). |
7476 | | * |
7477 | | * Method IsArrowSchemaSupported() can be called to determine if the schema |
7478 | | * will be supported by WriteArrowBatch(). |
7479 | | * |
7480 | | * OGR fields for the corresponding children arrays must exist and be of a |
7481 | | * compatible type. For attribute fields, they should generally be created with |
7482 | | * CreateFieldFromArrowSchema(). This is strictly required for output drivers |
7483 | | * Arrow or Parquet, and strongly recommended otherwise. For geometry fields, |
7484 | | * they should be created either implicitly at CreateLayer() type |
7485 | | * (if geom_type != wkbNone), or explicitly with CreateGeomField(). |
7486 | | * |
7487 | | * Starting with GDAL 3.9, some tolerance has been introduced in the base |
7488 | | * implementation of WriteArrowBatch() for scenarios that involve appending to |
7489 | | * an already existing output layer when the input Arrow field type and the |
7490 | | * OGR layer field type are 32/64-bi integers or real number, but do not match |
7491 | | * exactly, which may cause lossy conversions. The IF_FIELD_NOT_PRESERVED option |
7492 | | * can be used to control the behavior in case of lossy conversion. |
7493 | | * |
7494 | | * Arrays for geometry columns should be of binary or large binary type and |
7495 | | * contain WKB geometry. |
7496 | | * |
7497 | | * Note that the passed array may be set to a released state |
7498 | | * (array->release==NULL) after this call (not by the base implementation, |
7499 | | * but in specialized ones such as Parquet or Arrow for example) |
7500 | | * |
7501 | | * Supported options of the base implementation are: |
7502 | | * <ul> |
7503 | | * <li>FID=name. Name of the FID column in the array. If not provided, |
7504 | | * GetFIDColumn() is used to determine it. The special name |
7505 | | * OGRLayer::DEFAULT_ARROW_FID_NAME is also recognized if neither FID nor |
7506 | | * GetFIDColumn() are set. |
7507 | | * The corresponding ArrowArray must be of type int32 (i) or int64 (l). |
7508 | | * On input, values of the FID column are used to create the feature. |
7509 | | * On output, the values of the FID column may be set with the FID of the |
7510 | | * created feature (if the array is not released). |
7511 | | * </li> |
7512 | | * <li>IF_FID_NOT_PRESERVED=NOTHING/ERROR/WARNING. Action to perform when the |
7513 | | * input FID is not preserved in the output layer. The default is NOTHING. |
7514 | | * Setting it to ERROR will cause the function to error out. Setting it |
7515 | | * to WARNING will cause the function to emit a warning but continue its |
7516 | | * processing. |
7517 | | * </li> |
7518 | | * <li>IF_FIELD_NOT_PRESERVED=ERROR/WARNING. (since GDAL 3.9) |
7519 | | * Action to perform when the input field value is not preserved in the |
7520 | | * output layer. |
7521 | | * The default is WARNING, which will cause the function to emit a warning |
7522 | | * but continue its processing. |
7523 | | * Setting it to ERROR will cause the function to error out if a lossy |
7524 | | * conversion is detected. |
7525 | | * </li> |
7526 | | * <li>GEOMETRY_NAME=name. Name of the geometry column. If not provided, |
7527 | | * GetGeometryColumn() is used. The special name |
7528 | | * OGRLayer::DEFAULT_ARROW_GEOMETRY_NAME is also recognized if neither |
7529 | | * GEOMETRY_NAME nor GetGeometryColumn() are set. |
7530 | | * Geometry columns are also identified if they have |
7531 | | * ARROW:extension:name=ogc.wkb as a field metadata. |
7532 | | * The corresponding ArrowArray must be of type binary (w) or large |
7533 | | * binary (W). |
7534 | | * </li> |
7535 | | * </ul> |
7536 | | * |
7537 | | * The following example demonstrates how to copy a layer from one format to |
7538 | | * another one (assuming it has at most a single geometry column): |
7539 | | \code{.py} |
7540 | | def copy_layer(src_lyr, out_filename, out_format, lcos = {}): |
7541 | | stream = src_lyr.GetArrowStream() |
7542 | | schema = stream.GetSchema() |
7543 | | |
7544 | | # If the source layer has a FID column and the output driver supports |
7545 | | # a FID layer creation option, set it to the source FID column name. |
7546 | | if src_lyr.GetFIDColumn(): |
7547 | | creationOptions = gdal.GetDriverByName(out_format).GetMetadataItem( |
7548 | | "DS_LAYER_CREATIONOPTIONLIST" |
7549 | | ) |
7550 | | if creationOptions and '"FID"' in creationOptions: |
7551 | | lcos["FID"] = src_lyr.GetFIDColumn() |
7552 | | |
7553 | | with ogr.GetDriverByName(out_format).CreateDataSource(out_filename) as out_ds: |
7554 | | if src_lyr.GetLayerDefn().GetGeomFieldCount() > 1: |
7555 | | out_lyr = out_ds.CreateLayer( |
7556 | | src_lyr.GetName(), geom_type=ogr.wkbNone, options=lcos |
7557 | | ) |
7558 | | for i in range(src_lyr.GetLayerDefn().GetGeomFieldCount()): |
7559 | | out_lyr.CreateGeomField(src_lyr.GetLayerDefn().GetGeomFieldDefn(i)) |
7560 | | else: |
7561 | | out_lyr = out_ds.CreateLayer( |
7562 | | src_lyr.GetName(), |
7563 | | geom_type=src_lyr.GetGeomType(), |
7564 | | srs=src_lyr.GetSpatialRef(), |
7565 | | options=lcos, |
7566 | | ) |
7567 | | |
7568 | | success, error_msg = out_lyr.IsArrowSchemaSupported(schema) |
7569 | | assert success, error_msg |
7570 | | |
7571 | | src_geom_field_names = [ |
7572 | | src_lyr.GetLayerDefn().GetGeomFieldDefn(i).GetName() |
7573 | | for i in range(src_lyr.GetLayerDefn().GetGeomFieldCount()) |
7574 | | ] |
7575 | | for i in range(schema.GetChildrenCount()): |
7576 | | # GetArrowStream() may return "OGC_FID" for a unnamed source FID |
7577 | | # column and "wkb_geometry" for a unnamed source geometry column. |
7578 | | # Also test GetFIDColumn() and src_geom_field_names if they are |
7579 | | # named. |
7580 | | if ( |
7581 | | schema.GetChild(i).GetName() |
7582 | | not in ("OGC_FID", "wkb_geometry", src_lyr.GetFIDColumn()) |
7583 | | and schema.GetChild(i).GetName() not in src_geom_field_names |
7584 | | ): |
7585 | | out_lyr.CreateFieldFromArrowSchema(schema.GetChild(i)) |
7586 | | |
7587 | | write_options = [] |
7588 | | if src_lyr.GetFIDColumn(): |
7589 | | write_options.append("FID=" + src_lyr.GetFIDColumn()) |
7590 | | if ( |
7591 | | src_lyr.GetLayerDefn().GetGeomFieldCount() == 1 |
7592 | | and src_lyr.GetGeometryColumn() |
7593 | | ): |
7594 | | write_options.append("GEOMETRY_NAME=" + src_lyr.GetGeometryColumn()) |
7595 | | |
7596 | | while True: |
7597 | | array = stream.GetNextRecordBatch() |
7598 | | if array is None: |
7599 | | break |
7600 | | out_lyr.WriteArrowBatch(schema, array, write_options) |
7601 | | \endcode |
7602 | | * |
7603 | | * This method and CreateFeature() are mutually exclusive in the same session. |
7604 | | * |
7605 | | * This method is the same as the C function OGR_L_WriteArrowBatch(). |
7606 | | * |
7607 | | * @param schema Schema of array |
7608 | | * @param array Array of type struct. It may be released (array->release==NULL) |
7609 | | * after calling this method. |
7610 | | * @param papszOptions Options. Null terminated list, or nullptr. |
7611 | | * @return true in case of success |
7612 | | * @since 3.8 |
7613 | | */ |
7614 | | // clang-format on |
7615 | | |
7616 | | bool OGRLayer::WriteArrowBatch(const struct ArrowSchema *schema, |
7617 | | struct ArrowArray *array, |
7618 | | CSLConstList papszOptions) |
7619 | 0 | { |
7620 | 0 | const char *format = schema->format; |
7621 | 0 | if (!IsStructure(format)) |
7622 | 0 | { |
7623 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
7624 | 0 | "WriteArrowBatch() should be called on a schema that is a " |
7625 | 0 | "struct of fields"); |
7626 | 0 | return false; |
7627 | 0 | } |
7628 | | |
7629 | 0 | if (schema->n_children != array->n_children) |
7630 | 0 | { |
7631 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
7632 | 0 | "WriteArrowBatch(): schema->n_children (%d) != " |
7633 | 0 | "array->n_children (%d)", |
7634 | 0 | int(schema->n_children), int(array->n_children)); |
7635 | 0 | return false; |
7636 | 0 | } |
7637 | | |
7638 | 0 | CPLStringList aosNativeTypes; |
7639 | 0 | auto poDS = const_cast<OGRLayer *>(this)->GetDataset(); |
7640 | 0 | if (poDS) |
7641 | 0 | { |
7642 | 0 | auto poDriver = poDS->GetDriver(); |
7643 | 0 | if (poDriver) |
7644 | 0 | { |
7645 | 0 | const char *pszMetadataItem = |
7646 | 0 | poDriver->GetMetadataItem(GDAL_DMD_CREATIONFIELDDATATYPES); |
7647 | 0 | if (pszMetadataItem) |
7648 | 0 | aosNativeTypes = CSLTokenizeString2(pszMetadataItem, " ", 0); |
7649 | 0 | } |
7650 | 0 | } |
7651 | |
|
7652 | 0 | std::vector<FieldInfo> asFieldInfo; |
7653 | 0 | auto poLayerDefn = GetLayerDefn(); |
7654 | 0 | const char *pszFIDName = |
7655 | 0 | CSLFetchNameValueDef(papszOptions, "FID", GetFIDColumn()); |
7656 | 0 | if (!pszFIDName || pszFIDName[0] == 0) |
7657 | 0 | pszFIDName = DEFAULT_ARROW_FID_NAME; |
7658 | 0 | const bool bErrorIfFIDNotPreserved = |
7659 | 0 | EQUAL(CSLFetchNameValueDef(papszOptions, "IF_FID_NOT_PRESERVED", ""), |
7660 | 0 | "ERROR"); |
7661 | 0 | const bool bWarningIfFIDNotPreserved = |
7662 | 0 | EQUAL(CSLFetchNameValueDef(papszOptions, "IF_FID_NOT_PRESERVED", ""), |
7663 | 0 | "WARNING"); |
7664 | 0 | const bool bErrorIfFieldNotPreserved = |
7665 | 0 | EQUAL(CSLFetchNameValueDef(papszOptions, "IF_FIELD_NOT_PRESERVED", ""), |
7666 | 0 | "ERROR"); |
7667 | 0 | const char *pszGeomFieldName = CSLFetchNameValueDef( |
7668 | 0 | papszOptions, "GEOMETRY_NAME", GetGeometryColumn()); |
7669 | 0 | if (!pszGeomFieldName || pszGeomFieldName[0] == 0) |
7670 | 0 | pszGeomFieldName = DEFAULT_ARROW_GEOMETRY_NAME; |
7671 | 0 | const struct ArrowSchema *schemaFIDColumn = nullptr; |
7672 | 0 | struct ArrowArray *arrayFIDColumn = nullptr; |
7673 | 0 | bool bFallbackTypesUsed = false; |
7674 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
7675 | 0 | { |
7676 | 0 | if (!BuildOGRFieldInfo(schema->children[i], array->children[i], |
7677 | 0 | poLayerDefn, std::string(), aosNativeTypes, |
7678 | 0 | bFallbackTypesUsed, asFieldInfo, pszFIDName, |
7679 | 0 | pszGeomFieldName, this, |
7680 | 0 | m_poPrivate->m_oMapArrowFieldNameToOGRFieldName, |
7681 | 0 | schemaFIDColumn, arrayFIDColumn)) |
7682 | 0 | { |
7683 | 0 | return false; |
7684 | 0 | } |
7685 | 0 | } |
7686 | | |
7687 | 0 | std::map<int, int> oMapOGRFieldIndexToFieldInfoIndex; |
7688 | 0 | std::vector<bool> abUseStringOptim(poLayerDefn->GetFieldCount(), false); |
7689 | 0 | for (int i = 0; i < static_cast<int>(asFieldInfo.size()); ++i) |
7690 | 0 | { |
7691 | 0 | if (asFieldInfo[i].iOGRFieldIdx >= 0 && !asFieldInfo[i].bIsGeomCol) |
7692 | 0 | { |
7693 | 0 | CPLAssert(oMapOGRFieldIndexToFieldInfoIndex.find( |
7694 | 0 | asFieldInfo[i].iOGRFieldIdx) == |
7695 | 0 | oMapOGRFieldIndexToFieldInfoIndex.end()); |
7696 | 0 | oMapOGRFieldIndexToFieldInfoIndex[asFieldInfo[i].iOGRFieldIdx] = i; |
7697 | 0 | abUseStringOptim[asFieldInfo[i].iOGRFieldIdx] = |
7698 | 0 | asFieldInfo[i].bUseStringOptim; |
7699 | 0 | } |
7700 | 0 | } |
7701 | | |
7702 | 0 | OGRFeatureDefn oLayerDefnTmp(poLayerDefn->GetName()); |
7703 | |
|
7704 | 0 | struct LayerDefnTmpRefReleaser |
7705 | 0 | { |
7706 | 0 | OGRFeatureDefn &m_oDefn; |
7707 | |
|
7708 | 0 | explicit LayerDefnTmpRefReleaser(OGRFeatureDefn &oDefn) : m_oDefn(oDefn) |
7709 | 0 | { |
7710 | 0 | m_oDefn.Reference(); |
7711 | 0 | } |
7712 | |
|
7713 | 0 | ~LayerDefnTmpRefReleaser() |
7714 | 0 | { |
7715 | 0 | m_oDefn.Dereference(); |
7716 | 0 | } |
7717 | 0 | }; |
7718 | |
|
7719 | 0 | LayerDefnTmpRefReleaser oLayerDefnTmpRefReleaser(oLayerDefnTmp); |
7720 | |
|
7721 | 0 | std::vector<int> anIdentityFieldMap; |
7722 | 0 | if (bFallbackTypesUsed) |
7723 | 0 | { |
7724 | 0 | oLayerDefnTmp.SetGeomType(wkbNone); |
7725 | 0 | for (int i = 0; i < poLayerDefn->GetFieldCount(); ++i) |
7726 | 0 | { |
7727 | 0 | anIdentityFieldMap.push_back(i); |
7728 | 0 | const auto poSrcFieldDefn = poLayerDefn->GetFieldDefn(i); |
7729 | 0 | const auto oIter = oMapOGRFieldIndexToFieldInfoIndex.find(i); |
7730 | 0 | OGRFieldDefn oFieldDefn( |
7731 | 0 | poSrcFieldDefn->GetNameRef(), |
7732 | 0 | oIter == oMapOGRFieldIndexToFieldInfoIndex.end() |
7733 | 0 | ? poSrcFieldDefn->GetType() |
7734 | 0 | : asFieldInfo[oIter->second].eNominalFieldType); |
7735 | 0 | if (oIter != oMapOGRFieldIndexToFieldInfoIndex.end()) |
7736 | 0 | asFieldInfo[oIter->second].eSetFeatureFieldType = |
7737 | 0 | asFieldInfo[oIter->second].eNominalFieldType; |
7738 | 0 | oLayerDefnTmp.AddFieldDefn(&oFieldDefn); |
7739 | 0 | } |
7740 | 0 | for (int i = 0; i < poLayerDefn->GetGeomFieldCount(); ++i) |
7741 | 0 | { |
7742 | 0 | oLayerDefnTmp.AddGeomFieldDefn(poLayerDefn->GetGeomFieldDefn(i)); |
7743 | 0 | } |
7744 | 0 | } |
7745 | 0 | else |
7746 | 0 | { |
7747 | 0 | for (auto &sFieldInfo : asFieldInfo) |
7748 | 0 | sFieldInfo.eSetFeatureFieldType = sFieldInfo.eTargetFieldType; |
7749 | 0 | } |
7750 | |
|
7751 | 0 | struct FeatureCleaner |
7752 | 0 | { |
7753 | 0 | OGRFeature &m_oFeature; |
7754 | 0 | const std::vector<bool> &m_abUseStringOptim; |
7755 | |
|
7756 | 0 | explicit FeatureCleaner(OGRFeature &oFeature, |
7757 | 0 | const std::vector<bool> &abUseStringOptim) |
7758 | 0 | : m_oFeature(oFeature), m_abUseStringOptim(abUseStringOptim) |
7759 | 0 | { |
7760 | 0 | } |
7761 | | |
7762 | | // As we set a value that can't be CPLFree()'d in the .String member |
7763 | | // of string fields, we must take care of manually unsetting it before |
7764 | | // the destructor of OGRFeature gets called. |
7765 | 0 | ~FeatureCleaner() |
7766 | 0 | { |
7767 | 0 | const auto poLayerDefn = m_oFeature.GetDefnRef(); |
7768 | 0 | const int nFieldCount = poLayerDefn->GetFieldCount(); |
7769 | 0 | for (int i = 0; i < nFieldCount; ++i) |
7770 | 0 | { |
7771 | 0 | if (m_abUseStringOptim[i]) |
7772 | 0 | { |
7773 | 0 | if (m_oFeature.IsFieldSetAndNotNullUnsafe(i)) |
7774 | 0 | m_oFeature.SetFieldSameTypeUnsafe( |
7775 | 0 | i, static_cast<char *>(nullptr)); |
7776 | 0 | } |
7777 | 0 | } |
7778 | 0 | } |
7779 | 0 | }; |
7780 | |
|
7781 | 0 | OGRFeature oFeature(bFallbackTypesUsed ? &oLayerDefnTmp : poLayerDefn); |
7782 | 0 | FeatureCleaner oCleaner(oFeature, abUseStringOptim); |
7783 | 0 | OGRFeature oFeatureTarget(poLayerDefn); |
7784 | 0 | OGRFeature *const poFeatureTarget = |
7785 | 0 | bFallbackTypesUsed ? &oFeatureTarget : &oFeature; |
7786 | | |
7787 | | // We accumulate the content of all strings in osWorkingBuffer to avoid |
7788 | | // a few dynamic memory allocations |
7789 | 0 | std::string osWorkingBuffer; |
7790 | |
|
7791 | 0 | bool bTransactionOK; |
7792 | 0 | { |
7793 | 0 | CPLErrorStateBackuper oBackuper(CPLQuietErrorHandler); |
7794 | 0 | bTransactionOK = StartTransaction() == OGRERR_NONE; |
7795 | 0 | } |
7796 | |
|
7797 | 0 | const std::string emptyString; |
7798 | 0 | int64_t fidNullCount = 0; |
7799 | 0 | for (size_t iFeature = 0; iFeature < static_cast<size_t>(array->length); |
7800 | 0 | ++iFeature) |
7801 | 0 | { |
7802 | 0 | oFeature.SetFID(OGRNullFID); |
7803 | |
|
7804 | 0 | int iArrowIdx = 0; |
7805 | 0 | const size_t nWorkingBufferSize = GetWorkingBufferSize( |
7806 | 0 | schema, array, iFeature, iArrowIdx, asFieldInfo); |
7807 | 0 | osWorkingBuffer.clear(); |
7808 | 0 | osWorkingBuffer.reserve(nWorkingBufferSize); |
7809 | 0 | #ifdef DEBUG |
7810 | 0 | const char *pszWorkingBuffer = osWorkingBuffer.c_str(); |
7811 | 0 | CPL_IGNORE_RET_VAL(pszWorkingBuffer); |
7812 | 0 | #endif |
7813 | 0 | iArrowIdx = 0; |
7814 | 0 | for (int64_t i = 0; i < schema->n_children; ++i) |
7815 | 0 | { |
7816 | 0 | if (!FillFeature(this, schema->children[i], array->children[i], |
7817 | 0 | emptyString, iFeature, iArrowIdx, asFieldInfo, |
7818 | 0 | oFeature, osWorkingBuffer)) |
7819 | 0 | { |
7820 | 0 | if (bTransactionOK) |
7821 | 0 | RollbackTransaction(); |
7822 | 0 | return false; |
7823 | 0 | } |
7824 | 0 | } |
7825 | 0 | #ifdef DEBUG |
7826 | | // Check that the buffer didn't get reallocated |
7827 | 0 | CPLAssert(pszWorkingBuffer == osWorkingBuffer.c_str()); |
7828 | 0 | CPLAssert(osWorkingBuffer.size() == nWorkingBufferSize); |
7829 | 0 | #endif |
7830 | | |
7831 | 0 | if (bFallbackTypesUsed) |
7832 | 0 | { |
7833 | 0 | oFeatureTarget.SetFrom(&oFeature, anIdentityFieldMap.data(), |
7834 | 0 | /*bForgiving=*/true, |
7835 | 0 | /*bUseISO8601ForDateTimeAsString=*/true); |
7836 | 0 | oFeatureTarget.SetFID(oFeature.GetFID()); |
7837 | |
|
7838 | 0 | if (bErrorIfFieldNotPreserved) |
7839 | 0 | { |
7840 | 0 | for (int i = 0; i < poLayerDefn->GetFieldCount(); ++i) |
7841 | 0 | { |
7842 | 0 | if (!oFeature.IsFieldSetAndNotNullUnsafe(i)) |
7843 | 0 | { |
7844 | 0 | continue; |
7845 | 0 | } |
7846 | 0 | bool bLossyConversion = false; |
7847 | 0 | const auto eSrcType = |
7848 | 0 | oLayerDefnTmp.GetFieldDefnUnsafe(i)->GetType(); |
7849 | 0 | const auto eDstType = |
7850 | 0 | poLayerDefn->GetFieldDefnUnsafe(i)->GetType(); |
7851 | |
|
7852 | 0 | const auto IsDoubleCastToInt64EqualTInt64 = |
7853 | 0 | [](double dfVal, int64_t nOtherVal) |
7854 | 0 | { |
7855 | | // Values in the range [INT64_MAX - 1023, INT64_MAX - 1] |
7856 | | // get converted to a double that once cast to int64_t |
7857 | | // is INT64_MAX + 1, hence the strict < comparison |
7858 | 0 | return dfVal >= |
7859 | 0 | static_cast<double>( |
7860 | 0 | std::numeric_limits<int64_t>::min()) && |
7861 | 0 | dfVal < |
7862 | 0 | static_cast<double>( |
7863 | 0 | std::numeric_limits<int64_t>::max()) && |
7864 | 0 | static_cast<int64_t>(dfVal) == nOtherVal; |
7865 | 0 | }; |
7866 | |
|
7867 | 0 | if (eSrcType == OFTInteger64 && eDstType == OFTInteger && |
7868 | 0 | oFeatureTarget.GetFieldAsIntegerUnsafe(i) != |
7869 | 0 | oFeature.GetFieldAsInteger64Unsafe(i)) |
7870 | 0 | { |
7871 | 0 | bLossyConversion = true; |
7872 | 0 | } |
7873 | 0 | else if (eSrcType == OFTReal && eDstType == OFTInteger && |
7874 | 0 | oFeatureTarget.GetFieldAsIntegerUnsafe(i) != |
7875 | 0 | oFeature.GetFieldAsDoubleUnsafe(i)) |
7876 | 0 | { |
7877 | 0 | bLossyConversion = true; |
7878 | 0 | } |
7879 | 0 | else if (eSrcType == OFTReal && eDstType == OFTInteger64 && |
7880 | 0 | static_cast<double>( |
7881 | 0 | oFeatureTarget.GetFieldAsInteger64Unsafe(i)) != |
7882 | 0 | oFeature.GetFieldAsDoubleUnsafe(i)) |
7883 | 0 | { |
7884 | 0 | bLossyConversion = true; |
7885 | 0 | } |
7886 | 0 | else if (eSrcType == OFTInteger64 && eDstType == OFTReal && |
7887 | 0 | !IsDoubleCastToInt64EqualTInt64( |
7888 | 0 | oFeatureTarget.GetFieldAsDoubleUnsafe(i), |
7889 | 0 | oFeature.GetFieldAsInteger64Unsafe(i))) |
7890 | 0 | { |
7891 | 0 | bLossyConversion = true; |
7892 | 0 | } |
7893 | 0 | if (bLossyConversion) |
7894 | 0 | { |
7895 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
7896 | 0 | "For feature " CPL_FRMT_GIB |
7897 | 0 | ", value of field %s cannot not preserved", |
7898 | 0 | oFeatureTarget.GetFID(), |
7899 | 0 | oLayerDefnTmp.GetFieldDefn(i)->GetNameRef()); |
7900 | 0 | if (bTransactionOK) |
7901 | 0 | RollbackTransaction(); |
7902 | 0 | return false; |
7903 | 0 | } |
7904 | 0 | } |
7905 | 0 | } |
7906 | 0 | } |
7907 | | |
7908 | 0 | const auto nInputFID = poFeatureTarget->GetFID(); |
7909 | 0 | if (CreateFeature(poFeatureTarget) != OGRERR_NONE) |
7910 | 0 | { |
7911 | 0 | if (bTransactionOK) |
7912 | 0 | RollbackTransaction(); |
7913 | 0 | return false; |
7914 | 0 | } |
7915 | 0 | if (nInputFID != OGRNullFID) |
7916 | 0 | { |
7917 | 0 | if (bWarningIfFIDNotPreserved && |
7918 | | // cppcheck-suppress knownConditionTrueFalse |
7919 | 0 | poFeatureTarget->GetFID() != nInputFID) |
7920 | 0 | { |
7921 | 0 | CPLError(CE_Warning, CPLE_AppDefined, |
7922 | 0 | "Feature id " CPL_FRMT_GIB " not preserved", |
7923 | 0 | nInputFID); |
7924 | 0 | } |
7925 | 0 | else if (bErrorIfFIDNotPreserved && |
7926 | | // cppcheck-suppress knownConditionTrueFalse |
7927 | 0 | poFeatureTarget->GetFID() != nInputFID) |
7928 | 0 | { |
7929 | 0 | CPLError(CE_Failure, CPLE_AppDefined, |
7930 | 0 | "Feature id " CPL_FRMT_GIB " not preserved", |
7931 | 0 | nInputFID); |
7932 | 0 | if (bTransactionOK) |
7933 | 0 | RollbackTransaction(); |
7934 | 0 | return false; |
7935 | 0 | } |
7936 | 0 | } |
7937 | | |
7938 | 0 | if (arrayFIDColumn) |
7939 | 0 | { |
7940 | 0 | uint8_t *pabyValidity = static_cast<uint8_t *>( |
7941 | 0 | const_cast<void *>(arrayFIDColumn->buffers[0])); |
7942 | 0 | if (IsInt32(schemaFIDColumn->format)) |
7943 | 0 | { |
7944 | 0 | auto *panValues = static_cast<int32_t *>( |
7945 | 0 | const_cast<void *>(arrayFIDColumn->buffers[1])); |
7946 | 0 | if (poFeatureTarget->GetFID() > |
7947 | 0 | std::numeric_limits<int32_t>::max()) |
7948 | 0 | { |
7949 | 0 | if (pabyValidity) |
7950 | 0 | { |
7951 | 0 | ++fidNullCount; |
7952 | 0 | UnsetBit(pabyValidity, |
7953 | 0 | static_cast<size_t>(iFeature + |
7954 | 0 | arrayFIDColumn->offset)); |
7955 | 0 | } |
7956 | 0 | CPLError(CE_Warning, CPLE_AppDefined, |
7957 | 0 | "FID " CPL_FRMT_GIB |
7958 | 0 | " cannot be stored in FID array of type int32", |
7959 | 0 | poFeatureTarget->GetFID()); |
7960 | 0 | } |
7961 | 0 | else |
7962 | 0 | { |
7963 | 0 | if (pabyValidity) |
7964 | 0 | { |
7965 | 0 | SetBit(pabyValidity, |
7966 | 0 | static_cast<size_t>(iFeature + |
7967 | 0 | arrayFIDColumn->offset)); |
7968 | 0 | } |
7969 | 0 | panValues[iFeature + arrayFIDColumn->offset] = |
7970 | 0 | static_cast<int32_t>(poFeatureTarget->GetFID()); |
7971 | 0 | } |
7972 | 0 | } |
7973 | 0 | else if (IsInt64(schemaFIDColumn->format)) |
7974 | 0 | { |
7975 | 0 | if (pabyValidity) |
7976 | 0 | { |
7977 | 0 | SetBit( |
7978 | 0 | pabyValidity, |
7979 | 0 | static_cast<size_t>(iFeature + arrayFIDColumn->offset)); |
7980 | 0 | } |
7981 | 0 | auto *panValues = static_cast<int64_t *>( |
7982 | 0 | const_cast<void *>(arrayFIDColumn->buffers[1])); |
7983 | 0 | panValues[iFeature + arrayFIDColumn->offset] = |
7984 | 0 | poFeatureTarget->GetFID(); |
7985 | 0 | } |
7986 | 0 | else |
7987 | 0 | { |
7988 | 0 | CPLAssert(false); |
7989 | 0 | } |
7990 | 0 | } |
7991 | 0 | } |
7992 | 0 | if (arrayFIDColumn && arrayFIDColumn->buffers[0]) |
7993 | 0 | { |
7994 | 0 | arrayFIDColumn->null_count = fidNullCount; |
7995 | 0 | } |
7996 | |
|
7997 | 0 | bool bRet = true; |
7998 | 0 | if (bTransactionOK) |
7999 | 0 | bRet = CommitTransaction() == OGRERR_NONE; |
8000 | |
|
8001 | 0 | return bRet; |
8002 | 0 | } |
8003 | | |
8004 | | /************************************************************************/ |
8005 | | /* OGR_L_WriteArrowBatch() */ |
8006 | | /************************************************************************/ |
8007 | | |
8008 | | // clang-format off |
8009 | | /** Writes a batch of rows from an ArrowArray. |
8010 | | * |
8011 | | * This is semantically close to calling CreateFeature() with multiple features |
8012 | | * at once. |
8013 | | * |
8014 | | * The ArrowArray must be of type struct (format=+s), and its children generally |
8015 | | * map to a OGR attribute or geometry field (unless they are struct themselves). |
8016 | | * |
8017 | | * Method IsArrowSchemaSupported() can be called to determine if the schema |
8018 | | * will be supported by WriteArrowBatch(). |
8019 | | * |
8020 | | * OGR fields for the corresponding children arrays must exist and be of a |
8021 | | * compatible type. For attribute fields, they should generally be created with |
8022 | | * CreateFieldFromArrowSchema(). This is strictly required for output drivers |
8023 | | * Arrow or Parquet, and strongly recommended otherwise. For geometry fields, |
8024 | | * they should be created either implicitly at CreateLayer() type |
8025 | | * (if geom_type != wkbNone), or explicitly with CreateGeomField(). |
8026 | | * |
8027 | | * Starting with GDAL 3.9, some tolerance has been introduced in the base |
8028 | | * implementation of WriteArrowBatch() for scenarios that involve appending to |
8029 | | * an already existing output layer when the input Arrow field type and the |
8030 | | * OGR layer field type are 32/64-bi integers or real number, but do not match |
8031 | | * exactly, which may cause lossy conversions. The IF_FIELD_NOT_PRESERVED option |
8032 | | * can be used to control the behavior in case of lossy conversion. |
8033 | | * |
8034 | | * Arrays for geometry columns should be of binary or large binary type and |
8035 | | * contain WKB geometry. |
8036 | | * |
8037 | | * Note that the passed array may be set to a released state |
8038 | | * (array->release==NULL) after this call (not by the base implementation, |
8039 | | * but in specialized ones such as Parquet or Arrow for example) |
8040 | | * |
8041 | | * Supported options of the base implementation are: |
8042 | | * <ul> |
8043 | | * <li>FID=name. Name of the FID column in the array. If not provided, |
8044 | | * GetFIDColumn() is used to determine it. The special name |
8045 | | * OGRLayer::DEFAULT_ARROW_FID_NAME is also recognized if neither FID nor |
8046 | | * GetFIDColumn() are set. |
8047 | | * The corresponding ArrowArray must be of type int32 (i) or int64 (l). |
8048 | | * On input, values of the FID column are used to create the feature. |
8049 | | * On output, the values of the FID column may be set with the FID of the |
8050 | | * created feature (if the array is not released). |
8051 | | * </li> |
8052 | | * <li>IF_FID_NOT_PRESERVED=NOTHING/ERROR/WARNING. Action to perform when the |
8053 | | * input FID is not preserved in the output layer. The default is NOTHING. |
8054 | | * Setting it to ERROR will cause the function to error out. Setting it |
8055 | | * to WARNING will cause the function to emit a warning but continue its |
8056 | | * processing. |
8057 | | * </li> |
8058 | | * <li>IF_FIELD_NOT_PRESERVED=ERROR/WARNING. (since GDAL 3.9) |
8059 | | * Action to perform when the input field value is not preserved in the |
8060 | | * output layer. |
8061 | | * The default is WARNING, which will cause the function to emit a warning |
8062 | | * but continue its processing. |
8063 | | * Setting it to ERROR will cause the function to error out if a lossy |
8064 | | * conversion is detected. |
8065 | | * </li> |
8066 | | * <li>GEOMETRY_NAME=name. Name of the geometry column. If not provided, |
8067 | | * GetGeometryColumn() is used. The special name |
8068 | | * OGRLayer::DEFAULT_ARROW_GEOMETRY_NAME is also recognized if neither |
8069 | | * GEOMETRY_NAME nor GetGeometryColumn() are set. |
8070 | | * Geometry columns are also identified if they have |
8071 | | * ARROW:extension:name=ogc.wkb as a field metadata. |
8072 | | * The corresponding ArrowArray must be of type binary (w) or large |
8073 | | * binary (W). |
8074 | | * </li> |
8075 | | * </ul> |
8076 | | * |
8077 | | * The following example demonstrates how to copy a layer from one format to |
8078 | | * another one (assuming it has at most a single geometry column): |
8079 | | \code{.py} |
8080 | | def copy_layer(src_lyr, out_filename, out_format, lcos = {}): |
8081 | | stream = src_lyr.GetArrowStream() |
8082 | | schema = stream.GetSchema() |
8083 | | |
8084 | | # If the source layer has a FID column and the output driver supports |
8085 | | # a FID layer creation option, set it to the source FID column name. |
8086 | | if src_lyr.GetFIDColumn(): |
8087 | | creationOptions = gdal.GetDriverByName(out_format).GetMetadataItem( |
8088 | | "DS_LAYER_CREATIONOPTIONLIST" |
8089 | | ) |
8090 | | if creationOptions and '"FID"' in creationOptions: |
8091 | | lcos["FID"] = src_lyr.GetFIDColumn() |
8092 | | |
8093 | | with ogr.GetDriverByName(out_format).CreateDataSource(out_filename) as out_ds: |
8094 | | if src_lyr.GetLayerDefn().GetGeomFieldCount() > 1: |
8095 | | out_lyr = out_ds.CreateLayer( |
8096 | | src_lyr.GetName(), geom_type=ogr.wkbNone, options=lcos |
8097 | | ) |
8098 | | for i in range(src_lyr.GetLayerDefn().GetGeomFieldCount()): |
8099 | | out_lyr.CreateGeomField(src_lyr.GetLayerDefn().GetGeomFieldDefn(i)) |
8100 | | else: |
8101 | | out_lyr = out_ds.CreateLayer( |
8102 | | src_lyr.GetName(), |
8103 | | geom_type=src_lyr.GetGeomType(), |
8104 | | srs=src_lyr.GetSpatialRef(), |
8105 | | options=lcos, |
8106 | | ) |
8107 | | |
8108 | | success, error_msg = out_lyr.IsArrowSchemaSupported(schema) |
8109 | | assert success, error_msg |
8110 | | |
8111 | | src_geom_field_names = [ |
8112 | | src_lyr.GetLayerDefn().GetGeomFieldDefn(i).GetName() |
8113 | | for i in range(src_lyr.GetLayerDefn().GetGeomFieldCount()) |
8114 | | ] |
8115 | | for i in range(schema.GetChildrenCount()): |
8116 | | # GetArrowStream() may return "OGC_FID" for a unnamed source FID |
8117 | | # column and "wkb_geometry" for a unnamed source geometry column. |
8118 | | # Also test GetFIDColumn() and src_geom_field_names if they are |
8119 | | # named. |
8120 | | if ( |
8121 | | schema.GetChild(i).GetName() |
8122 | | not in ("OGC_FID", "wkb_geometry", src_lyr.GetFIDColumn()) |
8123 | | and schema.GetChild(i).GetName() not in src_geom_field_names |
8124 | | ): |
8125 | | out_lyr.CreateFieldFromArrowSchema(schema.GetChild(i)) |
8126 | | |
8127 | | write_options = [] |
8128 | | if src_lyr.GetFIDColumn(): |
8129 | | write_options.append("FID=" + src_lyr.GetFIDColumn()) |
8130 | | if ( |
8131 | | src_lyr.GetLayerDefn().GetGeomFieldCount() == 1 |
8132 | | and src_lyr.GetGeometryColumn() |
8133 | | ): |
8134 | | write_options.append("GEOMETRY_NAME=" + src_lyr.GetGeometryColumn()) |
8135 | | |
8136 | | while True: |
8137 | | array = stream.GetNextRecordBatch() |
8138 | | if array is None: |
8139 | | break |
8140 | | out_lyr.WriteArrowBatch(schema, array, write_options) |
8141 | | \endcode |
8142 | | * |
8143 | | * This method and CreateFeature() are mutually exclusive in the same session. |
8144 | | * |
8145 | | * This method is the same as the C++ method OGRLayer::WriteArrowBatch(). |
8146 | | * |
8147 | | * @param hLayer Layer. |
8148 | | * @param schema Schema of array. |
8149 | | * @param array Array of type struct. It may be released (array->release==NULL) |
8150 | | * after calling this method. |
8151 | | * @param papszOptions Options. Null terminated list, or nullptr. |
8152 | | * @return true in case of success |
8153 | | * @since 3.8 |
8154 | | */ |
8155 | | // clang-format on |
8156 | | |
8157 | | bool OGR_L_WriteArrowBatch(OGRLayerH hLayer, const struct ArrowSchema *schema, |
8158 | | struct ArrowArray *array, char **papszOptions) |
8159 | 0 | { |
8160 | 0 | VALIDATE_POINTER1(hLayer, __func__, false); |
8161 | 0 | VALIDATE_POINTER1(schema, __func__, false); |
8162 | 0 | VALIDATE_POINTER1(array, __func__, false); |
8163 | | |
8164 | 0 | return OGRLayer::FromHandle(hLayer)->WriteArrowBatch(schema, array, |
8165 | 0 | papszOptions); |
8166 | 0 | } |