/src/postgres/src/backend/utils/adt/datetime.c
Line | Count | Source (jump to first uncovered line) |
1 | | /*------------------------------------------------------------------------- |
2 | | * |
3 | | * datetime.c |
4 | | * Support functions for date/time types. |
5 | | * |
6 | | * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group |
7 | | * Portions Copyright (c) 1994, Regents of the University of California |
8 | | * |
9 | | * |
10 | | * IDENTIFICATION |
11 | | * src/backend/utils/adt/datetime.c |
12 | | * |
13 | | *------------------------------------------------------------------------- |
14 | | */ |
15 | | #include "postgres.h" |
16 | | |
17 | | #include <ctype.h> |
18 | | #include <limits.h> |
19 | | #include <math.h> |
20 | | |
21 | | #include "access/htup_details.h" |
22 | | #include "access/xact.h" |
23 | | #include "common/int.h" |
24 | | #include "common/string.h" |
25 | | #include "funcapi.h" |
26 | | #include "miscadmin.h" |
27 | | #include "nodes/nodeFuncs.h" |
28 | | #include "parser/scansup.h" |
29 | | #include "utils/builtins.h" |
30 | | #include "utils/date.h" |
31 | | #include "utils/datetime.h" |
32 | | #include "utils/guc.h" |
33 | | #include "utils/tzparser.h" |
34 | | |
35 | | static int DecodeNumber(int flen, char *str, bool haveTextMonth, |
36 | | int fmask, int *tmask, |
37 | | struct pg_tm *tm, fsec_t *fsec, bool *is2digits); |
38 | | static int DecodeNumberField(int len, char *str, |
39 | | int fmask, int *tmask, |
40 | | struct pg_tm *tm, fsec_t *fsec, bool *is2digits); |
41 | | static int DecodeTimeCommon(char *str, int fmask, int range, |
42 | | int *tmask, struct pg_itm *itm); |
43 | | static int DecodeTime(char *str, int fmask, int range, |
44 | | int *tmask, struct pg_tm *tm, fsec_t *fsec); |
45 | | static int DecodeTimeForInterval(char *str, int fmask, int range, |
46 | | int *tmask, struct pg_itm_in *itm_in); |
47 | | static const datetkn *datebsearch(const char *key, const datetkn *base, int nel); |
48 | | static int DecodeDate(char *str, int fmask, int *tmask, bool *is2digits, |
49 | | struct pg_tm *tm); |
50 | | static char *AppendSeconds(char *cp, int sec, fsec_t fsec, |
51 | | int precision, bool fillzeros); |
52 | | static bool int64_multiply_add(int64 val, int64 multiplier, int64 *sum); |
53 | | static bool AdjustFractMicroseconds(double frac, int64 scale, |
54 | | struct pg_itm_in *itm_in); |
55 | | static bool AdjustFractDays(double frac, int scale, |
56 | | struct pg_itm_in *itm_in); |
57 | | static bool AdjustFractYears(double frac, int scale, |
58 | | struct pg_itm_in *itm_in); |
59 | | static bool AdjustMicroseconds(int64 val, double fval, int64 scale, |
60 | | struct pg_itm_in *itm_in); |
61 | | static bool AdjustDays(int64 val, int scale, |
62 | | struct pg_itm_in *itm_in); |
63 | | static bool AdjustMonths(int64 val, struct pg_itm_in *itm_in); |
64 | | static bool AdjustYears(int64 val, int scale, |
65 | | struct pg_itm_in *itm_in); |
66 | | static int DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp, |
67 | | pg_time_t *tp); |
68 | | static bool DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t, |
69 | | const char *abbr, pg_tz *tzp, |
70 | | int *offset, int *isdst); |
71 | | static pg_tz *FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp, |
72 | | DateTimeErrorExtra *extra); |
73 | | |
74 | | |
75 | | const int day_tab[2][13] = |
76 | | { |
77 | | {31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 0}, |
78 | | {31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31, 0} |
79 | | }; |
80 | | |
81 | | const char *const months[] = {"Jan", "Feb", "Mar", "Apr", "May", "Jun", |
82 | | "Jul", "Aug", "Sep", "Oct", "Nov", "Dec", NULL}; |
83 | | |
84 | | const char *const days[] = {"Sunday", "Monday", "Tuesday", "Wednesday", |
85 | | "Thursday", "Friday", "Saturday", NULL}; |
86 | | |
87 | | |
88 | | /***************************************************************************** |
89 | | * PRIVATE ROUTINES * |
90 | | *****************************************************************************/ |
91 | | |
92 | | /* |
93 | | * datetktbl holds date/time keywords. |
94 | | * |
95 | | * Note that this table must be strictly alphabetically ordered to allow an |
96 | | * O(ln(N)) search algorithm to be used. |
97 | | * |
98 | | * The token field must be NUL-terminated; we truncate entries to TOKMAXLEN |
99 | | * characters to fit. |
100 | | * |
101 | | * The static table contains no TZ, DTZ, or DYNTZ entries; rather those |
102 | | * are loaded from configuration files and stored in zoneabbrevtbl, whose |
103 | | * abbrevs[] field has the same format as the static datetktbl. |
104 | | */ |
105 | | static const datetkn datetktbl[] = { |
106 | | /* token, type, value */ |
107 | | {"+infinity", RESERV, DTK_LATE}, /* same as "infinity" */ |
108 | | {EARLY, RESERV, DTK_EARLY}, /* "-infinity" reserved for "early time" */ |
109 | | {DA_D, ADBC, AD}, /* "ad" for years > 0 */ |
110 | | {"allballs", RESERV, DTK_ZULU}, /* 00:00:00 */ |
111 | | {"am", AMPM, AM}, |
112 | | {"apr", MONTH, 4}, |
113 | | {"april", MONTH, 4}, |
114 | | {"at", IGNORE_DTF, 0}, /* "at" (throwaway) */ |
115 | | {"aug", MONTH, 8}, |
116 | | {"august", MONTH, 8}, |
117 | | {DB_C, ADBC, BC}, /* "bc" for years <= 0 */ |
118 | | {"d", UNITS, DTK_DAY}, /* "day of month" for ISO input */ |
119 | | {"dec", MONTH, 12}, |
120 | | {"december", MONTH, 12}, |
121 | | {"dow", UNITS, DTK_DOW}, /* day of week */ |
122 | | {"doy", UNITS, DTK_DOY}, /* day of year */ |
123 | | {"dst", DTZMOD, SECS_PER_HOUR}, |
124 | | {EPOCH, RESERV, DTK_EPOCH}, /* "epoch" reserved for system epoch time */ |
125 | | {"feb", MONTH, 2}, |
126 | | {"february", MONTH, 2}, |
127 | | {"fri", DOW, 5}, |
128 | | {"friday", DOW, 5}, |
129 | | {"h", UNITS, DTK_HOUR}, /* "hour" */ |
130 | | {LATE, RESERV, DTK_LATE}, /* "infinity" reserved for "late time" */ |
131 | | {"isodow", UNITS, DTK_ISODOW}, /* ISO day of week, Sunday == 7 */ |
132 | | {"isoyear", UNITS, DTK_ISOYEAR}, /* year in terms of the ISO week date */ |
133 | | {"j", UNITS, DTK_JULIAN}, |
134 | | {"jan", MONTH, 1}, |
135 | | {"january", MONTH, 1}, |
136 | | {"jd", UNITS, DTK_JULIAN}, |
137 | | {"jul", MONTH, 7}, |
138 | | {"julian", UNITS, DTK_JULIAN}, |
139 | | {"july", MONTH, 7}, |
140 | | {"jun", MONTH, 6}, |
141 | | {"june", MONTH, 6}, |
142 | | {"m", UNITS, DTK_MONTH}, /* "month" for ISO input */ |
143 | | {"mar", MONTH, 3}, |
144 | | {"march", MONTH, 3}, |
145 | | {"may", MONTH, 5}, |
146 | | {"mm", UNITS, DTK_MINUTE}, /* "minute" for ISO input */ |
147 | | {"mon", DOW, 1}, |
148 | | {"monday", DOW, 1}, |
149 | | {"nov", MONTH, 11}, |
150 | | {"november", MONTH, 11}, |
151 | | {NOW, RESERV, DTK_NOW}, /* current transaction time */ |
152 | | {"oct", MONTH, 10}, |
153 | | {"october", MONTH, 10}, |
154 | | {"on", IGNORE_DTF, 0}, /* "on" (throwaway) */ |
155 | | {"pm", AMPM, PM}, |
156 | | {"s", UNITS, DTK_SECOND}, /* "seconds" for ISO input */ |
157 | | {"sat", DOW, 6}, |
158 | | {"saturday", DOW, 6}, |
159 | | {"sep", MONTH, 9}, |
160 | | {"sept", MONTH, 9}, |
161 | | {"september", MONTH, 9}, |
162 | | {"sun", DOW, 0}, |
163 | | {"sunday", DOW, 0}, |
164 | | {"t", ISOTIME, DTK_TIME}, /* Filler for ISO time fields */ |
165 | | {"thu", DOW, 4}, |
166 | | {"thur", DOW, 4}, |
167 | | {"thurs", DOW, 4}, |
168 | | {"thursday", DOW, 4}, |
169 | | {TODAY, RESERV, DTK_TODAY}, /* midnight */ |
170 | | {TOMORROW, RESERV, DTK_TOMORROW}, /* tomorrow midnight */ |
171 | | {"tue", DOW, 2}, |
172 | | {"tues", DOW, 2}, |
173 | | {"tuesday", DOW, 2}, |
174 | | {"wed", DOW, 3}, |
175 | | {"wednesday", DOW, 3}, |
176 | | {"weds", DOW, 3}, |
177 | | {"y", UNITS, DTK_YEAR}, /* "year" for ISO input */ |
178 | | {YESTERDAY, RESERV, DTK_YESTERDAY} /* yesterday midnight */ |
179 | | }; |
180 | | |
181 | | static const int szdatetktbl = sizeof datetktbl / sizeof datetktbl[0]; |
182 | | |
183 | | /* |
184 | | * deltatktbl: same format as datetktbl, but holds keywords used to represent |
185 | | * time units (eg, for intervals, and for EXTRACT). |
186 | | */ |
187 | | static const datetkn deltatktbl[] = { |
188 | | /* token, type, value */ |
189 | | {"@", IGNORE_DTF, 0}, /* postgres relative prefix */ |
190 | | {DAGO, AGO, 0}, /* "ago" indicates negative time offset */ |
191 | | {"c", UNITS, DTK_CENTURY}, /* "century" relative */ |
192 | | {"cent", UNITS, DTK_CENTURY}, /* "century" relative */ |
193 | | {"centuries", UNITS, DTK_CENTURY}, /* "centuries" relative */ |
194 | | {DCENTURY, UNITS, DTK_CENTURY}, /* "century" relative */ |
195 | | {"d", UNITS, DTK_DAY}, /* "day" relative */ |
196 | | {DDAY, UNITS, DTK_DAY}, /* "day" relative */ |
197 | | {"days", UNITS, DTK_DAY}, /* "days" relative */ |
198 | | {"dec", UNITS, DTK_DECADE}, /* "decade" relative */ |
199 | | {DDECADE, UNITS, DTK_DECADE}, /* "decade" relative */ |
200 | | {"decades", UNITS, DTK_DECADE}, /* "decades" relative */ |
201 | | {"decs", UNITS, DTK_DECADE}, /* "decades" relative */ |
202 | | {"h", UNITS, DTK_HOUR}, /* "hour" relative */ |
203 | | {DHOUR, UNITS, DTK_HOUR}, /* "hour" relative */ |
204 | | {"hours", UNITS, DTK_HOUR}, /* "hours" relative */ |
205 | | {"hr", UNITS, DTK_HOUR}, /* "hour" relative */ |
206 | | {"hrs", UNITS, DTK_HOUR}, /* "hours" relative */ |
207 | | {"m", UNITS, DTK_MINUTE}, /* "minute" relative */ |
208 | | {"microsecon", UNITS, DTK_MICROSEC}, /* "microsecond" relative */ |
209 | | {"mil", UNITS, DTK_MILLENNIUM}, /* "millennium" relative */ |
210 | | {"millennia", UNITS, DTK_MILLENNIUM}, /* "millennia" relative */ |
211 | | {DMILLENNIUM, UNITS, DTK_MILLENNIUM}, /* "millennium" relative */ |
212 | | {"millisecon", UNITS, DTK_MILLISEC}, /* relative */ |
213 | | {"mils", UNITS, DTK_MILLENNIUM}, /* "millennia" relative */ |
214 | | {"min", UNITS, DTK_MINUTE}, /* "minute" relative */ |
215 | | {"mins", UNITS, DTK_MINUTE}, /* "minutes" relative */ |
216 | | {DMINUTE, UNITS, DTK_MINUTE}, /* "minute" relative */ |
217 | | {"minutes", UNITS, DTK_MINUTE}, /* "minutes" relative */ |
218 | | {"mon", UNITS, DTK_MONTH}, /* "months" relative */ |
219 | | {"mons", UNITS, DTK_MONTH}, /* "months" relative */ |
220 | | {DMONTH, UNITS, DTK_MONTH}, /* "month" relative */ |
221 | | {"months", UNITS, DTK_MONTH}, |
222 | | {"ms", UNITS, DTK_MILLISEC}, |
223 | | {"msec", UNITS, DTK_MILLISEC}, |
224 | | {DMILLISEC, UNITS, DTK_MILLISEC}, |
225 | | {"mseconds", UNITS, DTK_MILLISEC}, |
226 | | {"msecs", UNITS, DTK_MILLISEC}, |
227 | | {"qtr", UNITS, DTK_QUARTER}, /* "quarter" relative */ |
228 | | {DQUARTER, UNITS, DTK_QUARTER}, /* "quarter" relative */ |
229 | | {"s", UNITS, DTK_SECOND}, |
230 | | {"sec", UNITS, DTK_SECOND}, |
231 | | {DSECOND, UNITS, DTK_SECOND}, |
232 | | {"seconds", UNITS, DTK_SECOND}, |
233 | | {"secs", UNITS, DTK_SECOND}, |
234 | | {DTIMEZONE, UNITS, DTK_TZ}, /* "timezone" time offset */ |
235 | | {"timezone_h", UNITS, DTK_TZ_HOUR}, /* timezone hour units */ |
236 | | {"timezone_m", UNITS, DTK_TZ_MINUTE}, /* timezone minutes units */ |
237 | | {"us", UNITS, DTK_MICROSEC}, /* "microsecond" relative */ |
238 | | {"usec", UNITS, DTK_MICROSEC}, /* "microsecond" relative */ |
239 | | {DMICROSEC, UNITS, DTK_MICROSEC}, /* "microsecond" relative */ |
240 | | {"useconds", UNITS, DTK_MICROSEC}, /* "microseconds" relative */ |
241 | | {"usecs", UNITS, DTK_MICROSEC}, /* "microseconds" relative */ |
242 | | {"w", UNITS, DTK_WEEK}, /* "week" relative */ |
243 | | {DWEEK, UNITS, DTK_WEEK}, /* "week" relative */ |
244 | | {"weeks", UNITS, DTK_WEEK}, /* "weeks" relative */ |
245 | | {"y", UNITS, DTK_YEAR}, /* "year" relative */ |
246 | | {DYEAR, UNITS, DTK_YEAR}, /* "year" relative */ |
247 | | {"years", UNITS, DTK_YEAR}, /* "years" relative */ |
248 | | {"yr", UNITS, DTK_YEAR}, /* "year" relative */ |
249 | | {"yrs", UNITS, DTK_YEAR} /* "years" relative */ |
250 | | }; |
251 | | |
252 | | static const int szdeltatktbl = sizeof deltatktbl / sizeof deltatktbl[0]; |
253 | | |
254 | | static TimeZoneAbbrevTable *zoneabbrevtbl = NULL; |
255 | | |
256 | | /* Caches of recent lookup results in the above tables */ |
257 | | |
258 | | static const datetkn *datecache[MAXDATEFIELDS] = {NULL}; |
259 | | |
260 | | static const datetkn *deltacache[MAXDATEFIELDS] = {NULL}; |
261 | | |
262 | | /* Cache for results of timezone abbreviation lookups */ |
263 | | |
264 | | typedef struct TzAbbrevCache |
265 | | { |
266 | | char abbrev[TOKMAXLEN + 1]; /* always NUL-terminated */ |
267 | | char ftype; /* TZ, DTZ, or DYNTZ */ |
268 | | int offset; /* GMT offset, if fixed-offset */ |
269 | | pg_tz *tz; /* relevant zone, if variable-offset */ |
270 | | } TzAbbrevCache; |
271 | | |
272 | | static TzAbbrevCache tzabbrevcache[MAXDATEFIELDS]; |
273 | | |
274 | | |
275 | | /* |
276 | | * Calendar time to Julian date conversions. |
277 | | * Julian date is commonly used in astronomical applications, |
278 | | * since it is numerically accurate and computationally simple. |
279 | | * The algorithms here will accurately convert between Julian day |
280 | | * and calendar date for all non-negative Julian days |
281 | | * (i.e. from Nov 24, -4713 on). |
282 | | * |
283 | | * Rewritten to eliminate overflow problems. This now allows the |
284 | | * routines to work correctly for all Julian day counts from |
285 | | * 0 to 2147483647 (Nov 24, -4713 to Jun 3, 5874898) assuming |
286 | | * a 32-bit integer. Longer types should also work to the limits |
287 | | * of their precision. |
288 | | * |
289 | | * Actually, date2j() will work sanely, in the sense of producing |
290 | | * valid negative Julian dates, significantly before Nov 24, -4713. |
291 | | * We rely on it to do so back to Nov 1, -4713; see IS_VALID_JULIAN() |
292 | | * and associated commentary in timestamp.h. |
293 | | */ |
294 | | |
295 | | int |
296 | | date2j(int year, int month, int day) |
297 | 0 | { |
298 | 0 | int julian; |
299 | 0 | int century; |
300 | |
|
301 | 0 | if (month > 2) |
302 | 0 | { |
303 | 0 | month += 1; |
304 | 0 | year += 4800; |
305 | 0 | } |
306 | 0 | else |
307 | 0 | { |
308 | 0 | month += 13; |
309 | 0 | year += 4799; |
310 | 0 | } |
311 | |
|
312 | 0 | century = year / 100; |
313 | 0 | julian = year * 365 - 32167; |
314 | 0 | julian += year / 4 - century + century / 4; |
315 | 0 | julian += 7834 * month / 256 + day; |
316 | |
|
317 | 0 | return julian; |
318 | 0 | } /* date2j() */ |
319 | | |
320 | | void |
321 | | j2date(int jd, int *year, int *month, int *day) |
322 | 0 | { |
323 | 0 | unsigned int julian; |
324 | 0 | unsigned int quad; |
325 | 0 | unsigned int extra; |
326 | 0 | int y; |
327 | |
|
328 | 0 | julian = jd; |
329 | 0 | julian += 32044; |
330 | 0 | quad = julian / 146097; |
331 | 0 | extra = (julian - quad * 146097) * 4 + 3; |
332 | 0 | julian += 60 + quad * 3 + extra / 146097; |
333 | 0 | quad = julian / 1461; |
334 | 0 | julian -= quad * 1461; |
335 | 0 | y = julian * 4 / 1461; |
336 | 0 | julian = ((y != 0) ? ((julian + 305) % 365) : ((julian + 306) % 366)) |
337 | 0 | + 123; |
338 | 0 | y += quad * 4; |
339 | 0 | *year = y - 4800; |
340 | 0 | quad = julian * 2141 / 65536; |
341 | 0 | *day = julian - 7834 * quad / 256; |
342 | 0 | *month = (quad + 10) % MONTHS_PER_YEAR + 1; |
343 | 0 | } /* j2date() */ |
344 | | |
345 | | |
346 | | /* |
347 | | * j2day - convert Julian date to day-of-week (0..6 == Sun..Sat) |
348 | | * |
349 | | * Note: various places use the locution j2day(date - 1) to produce a |
350 | | * result according to the convention 0..6 = Mon..Sun. This is a bit of |
351 | | * a crock, but will work as long as the computation here is just a modulo. |
352 | | */ |
353 | | int |
354 | | j2day(int date) |
355 | 0 | { |
356 | 0 | date += 1; |
357 | 0 | date %= 7; |
358 | | /* Cope if division truncates towards zero, as it probably does */ |
359 | 0 | if (date < 0) |
360 | 0 | date += 7; |
361 | |
|
362 | 0 | return date; |
363 | 0 | } /* j2day() */ |
364 | | |
365 | | |
366 | | /* |
367 | | * GetCurrentDateTime() |
368 | | * |
369 | | * Get the transaction start time ("now()") broken down as a struct pg_tm, |
370 | | * converted according to the session timezone setting. |
371 | | * |
372 | | * This is just a convenience wrapper for GetCurrentTimeUsec, to cover the |
373 | | * case where caller doesn't need either fractional seconds or tz offset. |
374 | | */ |
375 | | void |
376 | | GetCurrentDateTime(struct pg_tm *tm) |
377 | 0 | { |
378 | 0 | fsec_t fsec; |
379 | |
|
380 | 0 | GetCurrentTimeUsec(tm, &fsec, NULL); |
381 | 0 | } |
382 | | |
383 | | /* |
384 | | * GetCurrentTimeUsec() |
385 | | * |
386 | | * Get the transaction start time ("now()") broken down as a struct pg_tm, |
387 | | * including fractional seconds and timezone offset. The time is converted |
388 | | * according to the session timezone setting. |
389 | | * |
390 | | * Callers may pass tzp = NULL if they don't need the offset, but this does |
391 | | * not affect the conversion behavior (unlike timestamp2tm()). |
392 | | * |
393 | | * Internally, we cache the result, since this could be called many times |
394 | | * in a transaction, within which now() doesn't change. |
395 | | */ |
396 | | void |
397 | | GetCurrentTimeUsec(struct pg_tm *tm, fsec_t *fsec, int *tzp) |
398 | 0 | { |
399 | 0 | TimestampTz cur_ts = GetCurrentTransactionStartTimestamp(); |
400 | | |
401 | | /* |
402 | | * The cache key must include both current time and current timezone. By |
403 | | * representing the timezone by just a pointer, we're assuming that |
404 | | * distinct timezone settings could never have the same pointer value. |
405 | | * This is true by virtue of the hashtable used inside pg_tzset(); |
406 | | * however, it might need another look if we ever allow entries in that |
407 | | * hash to be recycled. |
408 | | */ |
409 | 0 | static TimestampTz cache_ts = 0; |
410 | 0 | static pg_tz *cache_timezone = NULL; |
411 | 0 | static struct pg_tm cache_tm; |
412 | 0 | static fsec_t cache_fsec; |
413 | 0 | static int cache_tz; |
414 | |
|
415 | 0 | if (cur_ts != cache_ts || session_timezone != cache_timezone) |
416 | 0 | { |
417 | | /* |
418 | | * Make sure cache is marked invalid in case of error after partial |
419 | | * update within timestamp2tm. |
420 | | */ |
421 | 0 | cache_timezone = NULL; |
422 | | |
423 | | /* |
424 | | * Perform the computation, storing results into cache. We do not |
425 | | * really expect any error here, since current time surely ought to be |
426 | | * within range, but check just for sanity's sake. |
427 | | */ |
428 | 0 | if (timestamp2tm(cur_ts, &cache_tz, &cache_tm, &cache_fsec, |
429 | 0 | NULL, session_timezone) != 0) |
430 | 0 | ereport(ERROR, |
431 | 0 | (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE), |
432 | 0 | errmsg("timestamp out of range"))); |
433 | | |
434 | | /* OK, so mark the cache valid. */ |
435 | 0 | cache_ts = cur_ts; |
436 | 0 | cache_timezone = session_timezone; |
437 | 0 | } |
438 | | |
439 | 0 | *tm = cache_tm; |
440 | 0 | *fsec = cache_fsec; |
441 | 0 | if (tzp != NULL) |
442 | 0 | *tzp = cache_tz; |
443 | 0 | } |
444 | | |
445 | | |
446 | | /* |
447 | | * Append seconds and fractional seconds (if any) at *cp. |
448 | | * |
449 | | * precision is the max number of fraction digits, fillzeros says to |
450 | | * pad to two integral-seconds digits. |
451 | | * |
452 | | * Returns a pointer to the new end of string. No NUL terminator is put |
453 | | * there; callers are responsible for NUL terminating str themselves. |
454 | | * |
455 | | * Note that any sign is stripped from the input sec and fsec values. |
456 | | */ |
457 | | static char * |
458 | | AppendSeconds(char *cp, int sec, fsec_t fsec, int precision, bool fillzeros) |
459 | 0 | { |
460 | 0 | Assert(precision >= 0); |
461 | |
|
462 | 0 | if (fillzeros) |
463 | 0 | cp = pg_ultostr_zeropad(cp, abs(sec), 2); |
464 | 0 | else |
465 | 0 | cp = pg_ultostr(cp, abs(sec)); |
466 | | |
467 | | /* fsec_t is just an int32 */ |
468 | 0 | if (fsec != 0) |
469 | 0 | { |
470 | 0 | int32 value = abs(fsec); |
471 | 0 | char *end = &cp[precision + 1]; |
472 | 0 | bool gotnonzero = false; |
473 | |
|
474 | 0 | *cp++ = '.'; |
475 | | |
476 | | /* |
477 | | * Append the fractional seconds part. Note that we don't want any |
478 | | * trailing zeros here, so since we're building the number in reverse |
479 | | * we'll skip appending zeros until we've output a non-zero digit. |
480 | | */ |
481 | 0 | while (precision--) |
482 | 0 | { |
483 | 0 | int32 oldval = value; |
484 | 0 | int32 remainder; |
485 | |
|
486 | 0 | value /= 10; |
487 | 0 | remainder = oldval - value * 10; |
488 | | |
489 | | /* check if we got a non-zero */ |
490 | 0 | if (remainder) |
491 | 0 | gotnonzero = true; |
492 | |
|
493 | 0 | if (gotnonzero) |
494 | 0 | cp[precision] = '0' + remainder; |
495 | 0 | else |
496 | 0 | end = &cp[precision]; |
497 | 0 | } |
498 | | |
499 | | /* |
500 | | * If we still have a non-zero value then precision must have not been |
501 | | * enough to print the number. We punt the problem to pg_ultostr(), |
502 | | * which will generate a correct answer in the minimum valid width. |
503 | | */ |
504 | 0 | if (value) |
505 | 0 | return pg_ultostr(cp, abs(fsec)); |
506 | | |
507 | 0 | return end; |
508 | 0 | } |
509 | 0 | else |
510 | 0 | return cp; |
511 | 0 | } |
512 | | |
513 | | |
514 | | /* |
515 | | * Variant of above that's specialized to timestamp case. |
516 | | * |
517 | | * Returns a pointer to the new end of string. No NUL terminator is put |
518 | | * there; callers are responsible for NUL terminating str themselves. |
519 | | */ |
520 | | static char * |
521 | | AppendTimestampSeconds(char *cp, struct pg_tm *tm, fsec_t fsec) |
522 | 0 | { |
523 | 0 | return AppendSeconds(cp, tm->tm_sec, fsec, MAX_TIMESTAMP_PRECISION, true); |
524 | 0 | } |
525 | | |
526 | | |
527 | | /* |
528 | | * Add val * multiplier to *sum. |
529 | | * Returns true if successful, false on overflow. |
530 | | */ |
531 | | static bool |
532 | | int64_multiply_add(int64 val, int64 multiplier, int64 *sum) |
533 | 0 | { |
534 | 0 | int64 product; |
535 | |
|
536 | 0 | if (pg_mul_s64_overflow(val, multiplier, &product) || |
537 | 0 | pg_add_s64_overflow(*sum, product, sum)) |
538 | 0 | return false; |
539 | 0 | return true; |
540 | 0 | } |
541 | | |
542 | | /* |
543 | | * Multiply frac by scale (to produce microseconds) and add to itm_in->tm_usec. |
544 | | * Returns true if successful, false if itm_in overflows. |
545 | | */ |
546 | | static bool |
547 | | AdjustFractMicroseconds(double frac, int64 scale, |
548 | | struct pg_itm_in *itm_in) |
549 | 0 | { |
550 | 0 | int64 usec; |
551 | | |
552 | | /* Fast path for common case */ |
553 | 0 | if (frac == 0) |
554 | 0 | return true; |
555 | | |
556 | | /* |
557 | | * We assume the input frac has abs value less than 1, so overflow of frac |
558 | | * or usec is not an issue for interesting values of scale. |
559 | | */ |
560 | 0 | frac *= scale; |
561 | 0 | usec = (int64) frac; |
562 | | |
563 | | /* Round off any fractional microsecond */ |
564 | 0 | frac -= usec; |
565 | 0 | if (frac > 0.5) |
566 | 0 | usec++; |
567 | 0 | else if (frac < -0.5) |
568 | 0 | usec--; |
569 | |
|
570 | 0 | return !pg_add_s64_overflow(itm_in->tm_usec, usec, &itm_in->tm_usec); |
571 | 0 | } |
572 | | |
573 | | /* |
574 | | * Multiply frac by scale (to produce days). Add the integral part of the |
575 | | * result to itm_in->tm_mday, the fractional part to itm_in->tm_usec. |
576 | | * Returns true if successful, false if itm_in overflows. |
577 | | */ |
578 | | static bool |
579 | | AdjustFractDays(double frac, int scale, |
580 | | struct pg_itm_in *itm_in) |
581 | 0 | { |
582 | 0 | int extra_days; |
583 | | |
584 | | /* Fast path for common case */ |
585 | 0 | if (frac == 0) |
586 | 0 | return true; |
587 | | |
588 | | /* |
589 | | * We assume the input frac has abs value less than 1, so overflow of frac |
590 | | * or extra_days is not an issue. |
591 | | */ |
592 | 0 | frac *= scale; |
593 | 0 | extra_days = (int) frac; |
594 | | |
595 | | /* ... but this could overflow, if tm_mday is already nonzero */ |
596 | 0 | if (pg_add_s32_overflow(itm_in->tm_mday, extra_days, &itm_in->tm_mday)) |
597 | 0 | return false; |
598 | | |
599 | | /* Handle any fractional day */ |
600 | 0 | frac -= extra_days; |
601 | 0 | return AdjustFractMicroseconds(frac, USECS_PER_DAY, itm_in); |
602 | 0 | } |
603 | | |
604 | | /* |
605 | | * Multiply frac by scale (to produce years), then further scale up to months. |
606 | | * Add the integral part of the result to itm_in->tm_mon, discarding any |
607 | | * fractional part. |
608 | | * Returns true if successful, false if itm_in overflows. |
609 | | */ |
610 | | static bool |
611 | | AdjustFractYears(double frac, int scale, |
612 | | struct pg_itm_in *itm_in) |
613 | 0 | { |
614 | | /* |
615 | | * As above, we assume abs(frac) < 1, so this can't overflow for any |
616 | | * interesting value of scale. |
617 | | */ |
618 | 0 | int extra_months = (int) rint(frac * scale * MONTHS_PER_YEAR); |
619 | |
|
620 | 0 | return !pg_add_s32_overflow(itm_in->tm_mon, extra_months, &itm_in->tm_mon); |
621 | 0 | } |
622 | | |
623 | | /* |
624 | | * Add (val + fval) * scale to itm_in->tm_usec. |
625 | | * Returns true if successful, false if itm_in overflows. |
626 | | */ |
627 | | static bool |
628 | | AdjustMicroseconds(int64 val, double fval, int64 scale, |
629 | | struct pg_itm_in *itm_in) |
630 | 0 | { |
631 | | /* Handle the integer part */ |
632 | 0 | if (!int64_multiply_add(val, scale, &itm_in->tm_usec)) |
633 | 0 | return false; |
634 | | /* Handle the float part */ |
635 | 0 | return AdjustFractMicroseconds(fval, scale, itm_in); |
636 | 0 | } |
637 | | |
638 | | /* |
639 | | * Multiply val by scale (to produce days) and add to itm_in->tm_mday. |
640 | | * Returns true if successful, false if itm_in overflows. |
641 | | */ |
642 | | static bool |
643 | | AdjustDays(int64 val, int scale, struct pg_itm_in *itm_in) |
644 | 0 | { |
645 | 0 | int days; |
646 | |
|
647 | 0 | if (val < INT_MIN || val > INT_MAX) |
648 | 0 | return false; |
649 | 0 | return !pg_mul_s32_overflow((int32) val, scale, &days) && |
650 | 0 | !pg_add_s32_overflow(itm_in->tm_mday, days, &itm_in->tm_mday); |
651 | 0 | } |
652 | | |
653 | | /* |
654 | | * Add val to itm_in->tm_mon (no need for scale here, as val is always |
655 | | * in months already). |
656 | | * Returns true if successful, false if itm_in overflows. |
657 | | */ |
658 | | static bool |
659 | | AdjustMonths(int64 val, struct pg_itm_in *itm_in) |
660 | 0 | { |
661 | 0 | if (val < INT_MIN || val > INT_MAX) |
662 | 0 | return false; |
663 | 0 | return !pg_add_s32_overflow(itm_in->tm_mon, (int32) val, &itm_in->tm_mon); |
664 | 0 | } |
665 | | |
666 | | /* |
667 | | * Multiply val by scale (to produce years) and add to itm_in->tm_year. |
668 | | * Returns true if successful, false if itm_in overflows. |
669 | | */ |
670 | | static bool |
671 | | AdjustYears(int64 val, int scale, |
672 | | struct pg_itm_in *itm_in) |
673 | 0 | { |
674 | 0 | int years; |
675 | |
|
676 | 0 | if (val < INT_MIN || val > INT_MAX) |
677 | 0 | return false; |
678 | 0 | return !pg_mul_s32_overflow((int32) val, scale, &years) && |
679 | 0 | !pg_add_s32_overflow(itm_in->tm_year, years, &itm_in->tm_year); |
680 | 0 | } |
681 | | |
682 | | |
683 | | /* |
684 | | * Parse the fractional part of a number (decimal point and optional digits, |
685 | | * followed by end of string). Returns the fractional value into *frac. |
686 | | * |
687 | | * Returns 0 if successful, DTERR code if bogus input detected. |
688 | | */ |
689 | | static int |
690 | | ParseFraction(char *cp, double *frac) |
691 | 0 | { |
692 | | /* Caller should always pass the start of the fraction part */ |
693 | 0 | Assert(*cp == '.'); |
694 | | |
695 | | /* |
696 | | * We want to allow just "." with no digits, but some versions of strtod |
697 | | * will report EINVAL for that, so special-case it. |
698 | | */ |
699 | 0 | if (cp[1] == '\0') |
700 | 0 | { |
701 | 0 | *frac = 0; |
702 | 0 | } |
703 | 0 | else |
704 | 0 | { |
705 | | /* |
706 | | * On the other hand, let's reject anything that's not digits after |
707 | | * the ".". strtod is happy with input like ".123e9", but that'd |
708 | | * break callers' expectation that the result is in 0..1. (It's quite |
709 | | * difficult to get here with such input, but not impossible.) |
710 | | */ |
711 | 0 | if (strspn(cp + 1, "0123456789") != strlen(cp + 1)) |
712 | 0 | return DTERR_BAD_FORMAT; |
713 | | |
714 | 0 | errno = 0; |
715 | 0 | *frac = strtod(cp, &cp); |
716 | | /* check for parse failure (probably redundant given prior check) */ |
717 | 0 | if (*cp != '\0' || errno != 0) |
718 | 0 | return DTERR_BAD_FORMAT; |
719 | 0 | } |
720 | 0 | return 0; |
721 | 0 | } |
722 | | |
723 | | /* |
724 | | * Fetch a fractional-second value with suitable error checking. |
725 | | * Same as ParseFraction except we convert the result to integer microseconds. |
726 | | */ |
727 | | static int |
728 | | ParseFractionalSecond(char *cp, fsec_t *fsec) |
729 | 0 | { |
730 | 0 | double frac; |
731 | 0 | int dterr; |
732 | |
|
733 | 0 | dterr = ParseFraction(cp, &frac); |
734 | 0 | if (dterr) |
735 | 0 | return dterr; |
736 | 0 | *fsec = rint(frac * 1000000); |
737 | 0 | return 0; |
738 | 0 | } |
739 | | |
740 | | |
741 | | /* ParseDateTime() |
742 | | * Break string into tokens based on a date/time context. |
743 | | * Returns 0 if successful, DTERR code if bogus input detected. |
744 | | * |
745 | | * timestr - the input string |
746 | | * workbuf - workspace for field string storage. This must be |
747 | | * larger than the largest legal input for this datetime type -- |
748 | | * some additional space will be needed to NUL terminate fields. |
749 | | * buflen - the size of workbuf |
750 | | * field[] - pointers to field strings are returned in this array |
751 | | * ftype[] - field type indicators are returned in this array |
752 | | * maxfields - dimensions of the above two arrays |
753 | | * *numfields - set to the actual number of fields detected |
754 | | * |
755 | | * The fields extracted from the input are stored as separate, |
756 | | * null-terminated strings in the workspace at workbuf. Any text is |
757 | | * converted to lower case. |
758 | | * |
759 | | * Several field types are assigned: |
760 | | * DTK_NUMBER - digits and (possibly) a decimal point |
761 | | * DTK_DATE - digits and two delimiters, or digits and text |
762 | | * DTK_TIME - digits, colon delimiters, and possibly a decimal point |
763 | | * DTK_STRING - text (no digits or punctuation) |
764 | | * DTK_SPECIAL - leading "+" or "-" followed by text |
765 | | * DTK_TZ - leading "+" or "-" followed by digits (also eats ':', '.', '-') |
766 | | * |
767 | | * Note that some field types can hold unexpected items: |
768 | | * DTK_NUMBER can hold date fields (yy.ddd) |
769 | | * DTK_STRING can hold months (January) and time zones (PST) |
770 | | * DTK_DATE can hold time zone names (America/New_York, GMT-8) |
771 | | */ |
772 | | int |
773 | | ParseDateTime(const char *timestr, char *workbuf, size_t buflen, |
774 | | char **field, int *ftype, int maxfields, int *numfields) |
775 | 0 | { |
776 | 0 | int nf = 0; |
777 | 0 | const char *cp = timestr; |
778 | 0 | char *bufp = workbuf; |
779 | 0 | const char *bufend = workbuf + buflen; |
780 | | |
781 | | /* |
782 | | * Set the character pointed-to by "bufptr" to "newchar", and increment |
783 | | * "bufptr". "end" gives the end of the buffer -- we return an error if |
784 | | * there is no space left to append a character to the buffer. Note that |
785 | | * "bufptr" is evaluated twice. |
786 | | */ |
787 | 0 | #define APPEND_CHAR(bufptr, end, newchar) \ |
788 | 0 | do \ |
789 | 0 | { \ |
790 | 0 | if (((bufptr) + 1) >= (end)) \ |
791 | 0 | return DTERR_BAD_FORMAT; \ |
792 | 0 | *(bufptr)++ = newchar; \ |
793 | 0 | } while (0) |
794 | | |
795 | | /* outer loop through fields */ |
796 | 0 | while (*cp != '\0') |
797 | 0 | { |
798 | | /* Ignore spaces between fields */ |
799 | 0 | if (isspace((unsigned char) *cp)) |
800 | 0 | { |
801 | 0 | cp++; |
802 | 0 | continue; |
803 | 0 | } |
804 | | |
805 | | /* Record start of current field */ |
806 | 0 | if (nf >= maxfields) |
807 | 0 | return DTERR_BAD_FORMAT; |
808 | 0 | field[nf] = bufp; |
809 | | |
810 | | /* leading digit? then date or time */ |
811 | 0 | if (isdigit((unsigned char) *cp)) |
812 | 0 | { |
813 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
814 | 0 | while (isdigit((unsigned char) *cp)) |
815 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
816 | | |
817 | | /* time field? */ |
818 | 0 | if (*cp == ':') |
819 | 0 | { |
820 | 0 | ftype[nf] = DTK_TIME; |
821 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
822 | 0 | while (isdigit((unsigned char) *cp) || |
823 | 0 | (*cp == ':') || (*cp == '.')) |
824 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
825 | 0 | } |
826 | | /* date field? allow embedded text month */ |
827 | 0 | else if (*cp == '-' || *cp == '/' || *cp == '.') |
828 | 0 | { |
829 | | /* save delimiting character to use later */ |
830 | 0 | char delim = *cp; |
831 | |
|
832 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
833 | | /* second field is all digits? then no embedded text month */ |
834 | 0 | if (isdigit((unsigned char) *cp)) |
835 | 0 | { |
836 | 0 | ftype[nf] = ((delim == '.') ? DTK_NUMBER : DTK_DATE); |
837 | 0 | while (isdigit((unsigned char) *cp)) |
838 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
839 | | |
840 | | /* |
841 | | * insist that the delimiters match to get a three-field |
842 | | * date. |
843 | | */ |
844 | 0 | if (*cp == delim) |
845 | 0 | { |
846 | 0 | ftype[nf] = DTK_DATE; |
847 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
848 | 0 | while (isdigit((unsigned char) *cp) || *cp == delim) |
849 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
850 | 0 | } |
851 | 0 | } |
852 | 0 | else |
853 | 0 | { |
854 | 0 | ftype[nf] = DTK_DATE; |
855 | 0 | while (isalnum((unsigned char) *cp) || *cp == delim) |
856 | 0 | APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++)); |
857 | 0 | } |
858 | 0 | } |
859 | | |
860 | | /* |
861 | | * otherwise, number only and will determine year, month, day, or |
862 | | * concatenated fields later... |
863 | | */ |
864 | 0 | else |
865 | 0 | ftype[nf] = DTK_NUMBER; |
866 | 0 | } |
867 | | /* Leading decimal point? Then fractional seconds... */ |
868 | 0 | else if (*cp == '.') |
869 | 0 | { |
870 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
871 | 0 | while (isdigit((unsigned char) *cp)) |
872 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
873 | | |
874 | 0 | ftype[nf] = DTK_NUMBER; |
875 | 0 | } |
876 | | |
877 | | /* |
878 | | * text? then date string, month, day of week, special, or timezone |
879 | | */ |
880 | 0 | else if (isalpha((unsigned char) *cp)) |
881 | 0 | { |
882 | 0 | bool is_date; |
883 | |
|
884 | 0 | ftype[nf] = DTK_STRING; |
885 | 0 | APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++)); |
886 | 0 | while (isalpha((unsigned char) *cp)) |
887 | 0 | APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++)); |
888 | | |
889 | | /* |
890 | | * Dates can have embedded '-', '/', or '.' separators. It could |
891 | | * also be a timezone name containing embedded '/', '+', '-', '_', |
892 | | * or ':' (but '_' or ':' can't be the first punctuation). If the |
893 | | * next character is a digit or '+', we need to check whether what |
894 | | * we have so far is a recognized non-timezone keyword --- if so, |
895 | | * don't believe that this is the start of a timezone. |
896 | | */ |
897 | 0 | is_date = false; |
898 | 0 | if (*cp == '-' || *cp == '/' || *cp == '.') |
899 | 0 | is_date = true; |
900 | 0 | else if (*cp == '+' || isdigit((unsigned char) *cp)) |
901 | 0 | { |
902 | 0 | *bufp = '\0'; /* null-terminate current field value */ |
903 | | /* we need search only the core token table, not TZ names */ |
904 | 0 | if (datebsearch(field[nf], datetktbl, szdatetktbl) == NULL) |
905 | 0 | is_date = true; |
906 | 0 | } |
907 | 0 | if (is_date) |
908 | 0 | { |
909 | 0 | ftype[nf] = DTK_DATE; |
910 | 0 | do |
911 | 0 | { |
912 | 0 | APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++)); |
913 | 0 | } while (*cp == '+' || *cp == '-' || |
914 | 0 | *cp == '/' || *cp == '_' || |
915 | 0 | *cp == '.' || *cp == ':' || |
916 | 0 | isalnum((unsigned char) *cp)); |
917 | 0 | } |
918 | 0 | } |
919 | | /* sign? then special or numeric timezone */ |
920 | 0 | else if (*cp == '+' || *cp == '-') |
921 | 0 | { |
922 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
923 | | /* soak up leading whitespace */ |
924 | 0 | while (isspace((unsigned char) *cp)) |
925 | 0 | cp++; |
926 | | /* numeric timezone? */ |
927 | | /* note that "DTK_TZ" could also be a signed float or yyyy-mm */ |
928 | 0 | if (isdigit((unsigned char) *cp)) |
929 | 0 | { |
930 | 0 | ftype[nf] = DTK_TZ; |
931 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
932 | 0 | while (isdigit((unsigned char) *cp) || |
933 | 0 | *cp == ':' || *cp == '.' || *cp == '-') |
934 | 0 | APPEND_CHAR(bufp, bufend, *cp++); |
935 | 0 | } |
936 | | /* special? */ |
937 | 0 | else if (isalpha((unsigned char) *cp)) |
938 | 0 | { |
939 | 0 | ftype[nf] = DTK_SPECIAL; |
940 | 0 | APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++)); |
941 | 0 | while (isalpha((unsigned char) *cp)) |
942 | 0 | APPEND_CHAR(bufp, bufend, pg_tolower((unsigned char) *cp++)); |
943 | 0 | } |
944 | | /* otherwise something wrong... */ |
945 | 0 | else |
946 | 0 | return DTERR_BAD_FORMAT; |
947 | 0 | } |
948 | | /* ignore other punctuation but use as delimiter */ |
949 | 0 | else if (ispunct((unsigned char) *cp)) |
950 | 0 | { |
951 | 0 | cp++; |
952 | 0 | continue; |
953 | 0 | } |
954 | | /* otherwise, something is not right... */ |
955 | 0 | else |
956 | 0 | return DTERR_BAD_FORMAT; |
957 | | |
958 | | /* force in a delimiter after each field */ |
959 | 0 | *bufp++ = '\0'; |
960 | 0 | nf++; |
961 | 0 | } |
962 | | |
963 | 0 | *numfields = nf; |
964 | |
|
965 | 0 | return 0; |
966 | 0 | } |
967 | | |
968 | | |
969 | | /* DecodeDateTime() |
970 | | * Interpret previously parsed fields for general date and time. |
971 | | * Return 0 if full date, 1 if only time, and negative DTERR code if problems. |
972 | | * (Currently, all callers treat 1 as an error return too.) |
973 | | * |
974 | | * Inputs are field[] and ftype[] arrays, of length nf. |
975 | | * Other arguments are outputs. |
976 | | * |
977 | | * External format(s): |
978 | | * "<weekday> <month>-<day>-<year> <hour>:<minute>:<second>" |
979 | | * "Fri Feb-7-1997 15:23:27" |
980 | | * "Feb-7-1997 15:23:27" |
981 | | * "2-7-1997 15:23:27" |
982 | | * "1997-2-7 15:23:27" |
983 | | * "1997.038 15:23:27" (day of year 1-366) |
984 | | * Also supports input in compact time: |
985 | | * "970207 152327" |
986 | | * "97038 152327" |
987 | | * "20011225T040506.789-07" |
988 | | * |
989 | | * Use the system-provided functions to get the current time zone |
990 | | * if not specified in the input string. |
991 | | * |
992 | | * If the date is outside the range of pg_time_t (in practice that could only |
993 | | * happen if pg_time_t is just 32 bits), then assume UTC time zone - thomas |
994 | | * 1997-05-27 |
995 | | */ |
996 | | int |
997 | | DecodeDateTime(char **field, int *ftype, int nf, |
998 | | int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp, |
999 | | DateTimeErrorExtra *extra) |
1000 | 0 | { |
1001 | 0 | int fmask = 0, |
1002 | 0 | tmask, |
1003 | 0 | type; |
1004 | 0 | int ptype = 0; /* "prefix type" for ISO and Julian formats */ |
1005 | 0 | int i; |
1006 | 0 | int val; |
1007 | 0 | int dterr; |
1008 | 0 | int mer = HR24; |
1009 | 0 | bool haveTextMonth = false; |
1010 | 0 | bool isjulian = false; |
1011 | 0 | bool is2digits = false; |
1012 | 0 | bool bc = false; |
1013 | 0 | pg_tz *namedTz = NULL; |
1014 | 0 | pg_tz *abbrevTz = NULL; |
1015 | 0 | pg_tz *valtz; |
1016 | 0 | char *abbrev = NULL; |
1017 | 0 | struct pg_tm cur_tm; |
1018 | | |
1019 | | /* |
1020 | | * We'll insist on at least all of the date fields, but initialize the |
1021 | | * remaining fields in case they are not set later... |
1022 | | */ |
1023 | 0 | *dtype = DTK_DATE; |
1024 | 0 | tm->tm_hour = 0; |
1025 | 0 | tm->tm_min = 0; |
1026 | 0 | tm->tm_sec = 0; |
1027 | 0 | *fsec = 0; |
1028 | | /* don't know daylight savings time status apriori */ |
1029 | 0 | tm->tm_isdst = -1; |
1030 | 0 | if (tzp != NULL) |
1031 | 0 | *tzp = 0; |
1032 | |
|
1033 | 0 | for (i = 0; i < nf; i++) |
1034 | 0 | { |
1035 | 0 | switch (ftype[i]) |
1036 | 0 | { |
1037 | 0 | case DTK_DATE: |
1038 | | |
1039 | | /* |
1040 | | * Integral julian day with attached time zone? All other |
1041 | | * forms with JD will be separated into distinct fields, so we |
1042 | | * handle just this case here. |
1043 | | */ |
1044 | 0 | if (ptype == DTK_JULIAN) |
1045 | 0 | { |
1046 | 0 | char *cp; |
1047 | 0 | int jday; |
1048 | |
|
1049 | 0 | if (tzp == NULL) |
1050 | 0 | return DTERR_BAD_FORMAT; |
1051 | | |
1052 | 0 | errno = 0; |
1053 | 0 | jday = strtoint(field[i], &cp, 10); |
1054 | 0 | if (errno == ERANGE || jday < 0) |
1055 | 0 | return DTERR_FIELD_OVERFLOW; |
1056 | | |
1057 | 0 | j2date(jday, &tm->tm_year, &tm->tm_mon, &tm->tm_mday); |
1058 | 0 | isjulian = true; |
1059 | | |
1060 | | /* Get the time zone from the end of the string */ |
1061 | 0 | dterr = DecodeTimezone(cp, tzp); |
1062 | 0 | if (dterr) |
1063 | 0 | return dterr; |
1064 | | |
1065 | 0 | tmask = DTK_DATE_M | DTK_TIME_M | DTK_M(TZ); |
1066 | 0 | ptype = 0; |
1067 | 0 | break; |
1068 | 0 | } |
1069 | | |
1070 | | /* |
1071 | | * Already have a date? Then this might be a time zone name |
1072 | | * with embedded punctuation (e.g. "America/New_York") or a |
1073 | | * run-together time with trailing time zone (e.g. hhmmss-zz). |
1074 | | * - thomas 2001-12-25 |
1075 | | * |
1076 | | * We consider it a time zone if we already have month & day. |
1077 | | * This is to allow the form "mmm dd hhmmss tz year", which |
1078 | | * we've historically accepted. |
1079 | | */ |
1080 | 0 | else if (ptype != 0 || |
1081 | 0 | ((fmask & (DTK_M(MONTH) | DTK_M(DAY))) == |
1082 | 0 | (DTK_M(MONTH) | DTK_M(DAY)))) |
1083 | 0 | { |
1084 | | /* No time zone accepted? Then quit... */ |
1085 | 0 | if (tzp == NULL) |
1086 | 0 | return DTERR_BAD_FORMAT; |
1087 | | |
1088 | 0 | if (isdigit((unsigned char) *field[i]) || ptype != 0) |
1089 | 0 | { |
1090 | 0 | char *cp; |
1091 | | |
1092 | | /* |
1093 | | * Allow a preceding "t" field, but no other units. |
1094 | | */ |
1095 | 0 | if (ptype != 0) |
1096 | 0 | { |
1097 | | /* Sanity check; should not fail this test */ |
1098 | 0 | if (ptype != DTK_TIME) |
1099 | 0 | return DTERR_BAD_FORMAT; |
1100 | 0 | ptype = 0; |
1101 | 0 | } |
1102 | | |
1103 | | /* |
1104 | | * Starts with a digit but we already have a time |
1105 | | * field? Then we are in trouble with a date and time |
1106 | | * already... |
1107 | | */ |
1108 | 0 | if ((fmask & DTK_TIME_M) == DTK_TIME_M) |
1109 | 0 | return DTERR_BAD_FORMAT; |
1110 | | |
1111 | 0 | if ((cp = strchr(field[i], '-')) == NULL) |
1112 | 0 | return DTERR_BAD_FORMAT; |
1113 | | |
1114 | | /* Get the time zone from the end of the string */ |
1115 | 0 | dterr = DecodeTimezone(cp, tzp); |
1116 | 0 | if (dterr) |
1117 | 0 | return dterr; |
1118 | 0 | *cp = '\0'; |
1119 | | |
1120 | | /* |
1121 | | * Then read the rest of the field as a concatenated |
1122 | | * time |
1123 | | */ |
1124 | 0 | dterr = DecodeNumberField(strlen(field[i]), field[i], |
1125 | 0 | fmask, |
1126 | 0 | &tmask, tm, |
1127 | 0 | fsec, &is2digits); |
1128 | 0 | if (dterr < 0) |
1129 | 0 | return dterr; |
1130 | | |
1131 | | /* |
1132 | | * modify tmask after returning from |
1133 | | * DecodeNumberField() |
1134 | | */ |
1135 | 0 | tmask |= DTK_M(TZ); |
1136 | 0 | } |
1137 | 0 | else |
1138 | 0 | { |
1139 | 0 | namedTz = pg_tzset(field[i]); |
1140 | 0 | if (!namedTz) |
1141 | 0 | { |
1142 | 0 | extra->dtee_timezone = field[i]; |
1143 | 0 | return DTERR_BAD_TIMEZONE; |
1144 | 0 | } |
1145 | | /* we'll apply the zone setting below */ |
1146 | 0 | tmask = DTK_M(TZ); |
1147 | 0 | } |
1148 | 0 | } |
1149 | 0 | else |
1150 | 0 | { |
1151 | 0 | dterr = DecodeDate(field[i], fmask, |
1152 | 0 | &tmask, &is2digits, tm); |
1153 | 0 | if (dterr) |
1154 | 0 | return dterr; |
1155 | 0 | } |
1156 | 0 | break; |
1157 | | |
1158 | 0 | case DTK_TIME: |
1159 | | |
1160 | | /* |
1161 | | * This might be an ISO time following a "t" field. |
1162 | | */ |
1163 | 0 | if (ptype != 0) |
1164 | 0 | { |
1165 | | /* Sanity check; should not fail this test */ |
1166 | 0 | if (ptype != DTK_TIME) |
1167 | 0 | return DTERR_BAD_FORMAT; |
1168 | 0 | ptype = 0; |
1169 | 0 | } |
1170 | 0 | dterr = DecodeTime(field[i], fmask, INTERVAL_FULL_RANGE, |
1171 | 0 | &tmask, tm, fsec); |
1172 | 0 | if (dterr) |
1173 | 0 | return dterr; |
1174 | | |
1175 | | /* check for time overflow */ |
1176 | 0 | if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec, |
1177 | 0 | *fsec)) |
1178 | 0 | return DTERR_FIELD_OVERFLOW; |
1179 | 0 | break; |
1180 | | |
1181 | 0 | case DTK_TZ: |
1182 | 0 | { |
1183 | 0 | int tz; |
1184 | |
|
1185 | 0 | if (tzp == NULL) |
1186 | 0 | return DTERR_BAD_FORMAT; |
1187 | | |
1188 | 0 | dterr = DecodeTimezone(field[i], &tz); |
1189 | 0 | if (dterr) |
1190 | 0 | return dterr; |
1191 | 0 | *tzp = tz; |
1192 | 0 | tmask = DTK_M(TZ); |
1193 | 0 | } |
1194 | 0 | break; |
1195 | | |
1196 | 0 | case DTK_NUMBER: |
1197 | | |
1198 | | /* |
1199 | | * Deal with cases where previous field labeled this one |
1200 | | */ |
1201 | 0 | if (ptype != 0) |
1202 | 0 | { |
1203 | 0 | char *cp; |
1204 | 0 | int value; |
1205 | |
|
1206 | 0 | errno = 0; |
1207 | 0 | value = strtoint(field[i], &cp, 10); |
1208 | 0 | if (errno == ERANGE) |
1209 | 0 | return DTERR_FIELD_OVERFLOW; |
1210 | 0 | if (*cp != '.' && *cp != '\0') |
1211 | 0 | return DTERR_BAD_FORMAT; |
1212 | | |
1213 | 0 | switch (ptype) |
1214 | 0 | { |
1215 | 0 | case DTK_JULIAN: |
1216 | | /* previous field was a label for "julian date" */ |
1217 | 0 | if (value < 0) |
1218 | 0 | return DTERR_FIELD_OVERFLOW; |
1219 | 0 | tmask = DTK_DATE_M; |
1220 | 0 | j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday); |
1221 | 0 | isjulian = true; |
1222 | | |
1223 | | /* fractional Julian Day? */ |
1224 | 0 | if (*cp == '.') |
1225 | 0 | { |
1226 | 0 | double time; |
1227 | |
|
1228 | 0 | dterr = ParseFraction(cp, &time); |
1229 | 0 | if (dterr) |
1230 | 0 | return dterr; |
1231 | 0 | time *= USECS_PER_DAY; |
1232 | 0 | dt2time(time, |
1233 | 0 | &tm->tm_hour, &tm->tm_min, |
1234 | 0 | &tm->tm_sec, fsec); |
1235 | 0 | tmask |= DTK_TIME_M; |
1236 | 0 | } |
1237 | 0 | break; |
1238 | | |
1239 | 0 | case DTK_TIME: |
1240 | | /* previous field was "t" for ISO time */ |
1241 | 0 | dterr = DecodeNumberField(strlen(field[i]), field[i], |
1242 | 0 | (fmask | DTK_DATE_M), |
1243 | 0 | &tmask, tm, |
1244 | 0 | fsec, &is2digits); |
1245 | 0 | if (dterr < 0) |
1246 | 0 | return dterr; |
1247 | 0 | if (tmask != DTK_TIME_M) |
1248 | 0 | return DTERR_BAD_FORMAT; |
1249 | 0 | break; |
1250 | | |
1251 | 0 | default: |
1252 | 0 | return DTERR_BAD_FORMAT; |
1253 | 0 | break; |
1254 | 0 | } |
1255 | | |
1256 | 0 | ptype = 0; |
1257 | 0 | *dtype = DTK_DATE; |
1258 | 0 | } |
1259 | 0 | else |
1260 | 0 | { |
1261 | 0 | char *cp; |
1262 | 0 | int flen; |
1263 | |
|
1264 | 0 | flen = strlen(field[i]); |
1265 | 0 | cp = strchr(field[i], '.'); |
1266 | | |
1267 | | /* Embedded decimal and no date yet? */ |
1268 | 0 | if (cp != NULL && !(fmask & DTK_DATE_M)) |
1269 | 0 | { |
1270 | 0 | dterr = DecodeDate(field[i], fmask, |
1271 | 0 | &tmask, &is2digits, tm); |
1272 | 0 | if (dterr) |
1273 | 0 | return dterr; |
1274 | 0 | } |
1275 | | /* embedded decimal and several digits before? */ |
1276 | 0 | else if (cp != NULL && flen - strlen(cp) > 2) |
1277 | 0 | { |
1278 | | /* |
1279 | | * Interpret as a concatenated date or time Set the |
1280 | | * type field to allow decoding other fields later. |
1281 | | * Example: 20011223 or 040506 |
1282 | | */ |
1283 | 0 | dterr = DecodeNumberField(flen, field[i], fmask, |
1284 | 0 | &tmask, tm, |
1285 | 0 | fsec, &is2digits); |
1286 | 0 | if (dterr < 0) |
1287 | 0 | return dterr; |
1288 | 0 | } |
1289 | | |
1290 | | /* |
1291 | | * Is this a YMD or HMS specification, or a year number? |
1292 | | * YMD and HMS are required to be six digits or more, so |
1293 | | * if it is 5 digits, it is a year. If it is six or more |
1294 | | * digits, we assume it is YMD or HMS unless no date and |
1295 | | * no time values have been specified. This forces 6+ |
1296 | | * digit years to be at the end of the string, or to use |
1297 | | * the ISO date specification. |
1298 | | */ |
1299 | 0 | else if (flen >= 6 && (!(fmask & DTK_DATE_M) || |
1300 | 0 | !(fmask & DTK_TIME_M))) |
1301 | 0 | { |
1302 | 0 | dterr = DecodeNumberField(flen, field[i], fmask, |
1303 | 0 | &tmask, tm, |
1304 | 0 | fsec, &is2digits); |
1305 | 0 | if (dterr < 0) |
1306 | 0 | return dterr; |
1307 | 0 | } |
1308 | | /* otherwise it is a single date/time field... */ |
1309 | 0 | else |
1310 | 0 | { |
1311 | 0 | dterr = DecodeNumber(flen, field[i], |
1312 | 0 | haveTextMonth, fmask, |
1313 | 0 | &tmask, tm, |
1314 | 0 | fsec, &is2digits); |
1315 | 0 | if (dterr) |
1316 | 0 | return dterr; |
1317 | 0 | } |
1318 | 0 | } |
1319 | 0 | break; |
1320 | | |
1321 | 0 | case DTK_STRING: |
1322 | 0 | case DTK_SPECIAL: |
1323 | | /* timezone abbrevs take precedence over built-in tokens */ |
1324 | 0 | dterr = DecodeTimezoneAbbrev(i, field[i], |
1325 | 0 | &type, &val, &valtz, extra); |
1326 | 0 | if (dterr) |
1327 | 0 | return dterr; |
1328 | 0 | if (type == UNKNOWN_FIELD) |
1329 | 0 | type = DecodeSpecial(i, field[i], &val); |
1330 | 0 | if (type == IGNORE_DTF) |
1331 | 0 | continue; |
1332 | | |
1333 | 0 | tmask = DTK_M(type); |
1334 | 0 | switch (type) |
1335 | 0 | { |
1336 | 0 | case RESERV: |
1337 | 0 | switch (val) |
1338 | 0 | { |
1339 | 0 | case DTK_NOW: |
1340 | 0 | tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ)); |
1341 | 0 | *dtype = DTK_DATE; |
1342 | 0 | GetCurrentTimeUsec(tm, fsec, tzp); |
1343 | 0 | break; |
1344 | | |
1345 | 0 | case DTK_YESTERDAY: |
1346 | 0 | tmask = DTK_DATE_M; |
1347 | 0 | *dtype = DTK_DATE; |
1348 | 0 | GetCurrentDateTime(&cur_tm); |
1349 | 0 | j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) - 1, |
1350 | 0 | &tm->tm_year, &tm->tm_mon, &tm->tm_mday); |
1351 | 0 | break; |
1352 | | |
1353 | 0 | case DTK_TODAY: |
1354 | 0 | tmask = DTK_DATE_M; |
1355 | 0 | *dtype = DTK_DATE; |
1356 | 0 | GetCurrentDateTime(&cur_tm); |
1357 | 0 | tm->tm_year = cur_tm.tm_year; |
1358 | 0 | tm->tm_mon = cur_tm.tm_mon; |
1359 | 0 | tm->tm_mday = cur_tm.tm_mday; |
1360 | 0 | break; |
1361 | | |
1362 | 0 | case DTK_TOMORROW: |
1363 | 0 | tmask = DTK_DATE_M; |
1364 | 0 | *dtype = DTK_DATE; |
1365 | 0 | GetCurrentDateTime(&cur_tm); |
1366 | 0 | j2date(date2j(cur_tm.tm_year, cur_tm.tm_mon, cur_tm.tm_mday) + 1, |
1367 | 0 | &tm->tm_year, &tm->tm_mon, &tm->tm_mday); |
1368 | 0 | break; |
1369 | | |
1370 | 0 | case DTK_ZULU: |
1371 | 0 | tmask = (DTK_TIME_M | DTK_M(TZ)); |
1372 | 0 | *dtype = DTK_DATE; |
1373 | 0 | tm->tm_hour = 0; |
1374 | 0 | tm->tm_min = 0; |
1375 | 0 | tm->tm_sec = 0; |
1376 | 0 | if (tzp != NULL) |
1377 | 0 | *tzp = 0; |
1378 | 0 | break; |
1379 | | |
1380 | 0 | case DTK_EPOCH: |
1381 | 0 | case DTK_LATE: |
1382 | 0 | case DTK_EARLY: |
1383 | 0 | tmask = (DTK_DATE_M | DTK_TIME_M | DTK_M(TZ)); |
1384 | 0 | *dtype = val; |
1385 | | /* caller ignores tm for these dtype codes */ |
1386 | 0 | break; |
1387 | | |
1388 | 0 | default: |
1389 | 0 | elog(ERROR, "unrecognized RESERV datetime token: %d", |
1390 | 0 | val); |
1391 | 0 | } |
1392 | | |
1393 | 0 | break; |
1394 | | |
1395 | 0 | case MONTH: |
1396 | | |
1397 | | /* |
1398 | | * already have a (numeric) month? then see if we can |
1399 | | * substitute... |
1400 | | */ |
1401 | 0 | if ((fmask & DTK_M(MONTH)) && !haveTextMonth && |
1402 | 0 | !(fmask & DTK_M(DAY)) && tm->tm_mon >= 1 && |
1403 | 0 | tm->tm_mon <= 31) |
1404 | 0 | { |
1405 | 0 | tm->tm_mday = tm->tm_mon; |
1406 | 0 | tmask = DTK_M(DAY); |
1407 | 0 | } |
1408 | 0 | haveTextMonth = true; |
1409 | 0 | tm->tm_mon = val; |
1410 | 0 | break; |
1411 | | |
1412 | 0 | case DTZMOD: |
1413 | | |
1414 | | /* |
1415 | | * daylight savings time modifier (solves "MET DST" |
1416 | | * syntax) |
1417 | | */ |
1418 | 0 | tmask |= DTK_M(DTZ); |
1419 | 0 | tm->tm_isdst = 1; |
1420 | 0 | if (tzp == NULL) |
1421 | 0 | return DTERR_BAD_FORMAT; |
1422 | 0 | *tzp -= val; |
1423 | 0 | break; |
1424 | | |
1425 | 0 | case DTZ: |
1426 | | |
1427 | | /* |
1428 | | * set mask for TZ here _or_ check for DTZ later when |
1429 | | * getting default timezone |
1430 | | */ |
1431 | 0 | tmask |= DTK_M(TZ); |
1432 | 0 | tm->tm_isdst = 1; |
1433 | 0 | if (tzp == NULL) |
1434 | 0 | return DTERR_BAD_FORMAT; |
1435 | 0 | *tzp = -val; |
1436 | 0 | break; |
1437 | | |
1438 | 0 | case TZ: |
1439 | 0 | tm->tm_isdst = 0; |
1440 | 0 | if (tzp == NULL) |
1441 | 0 | return DTERR_BAD_FORMAT; |
1442 | 0 | *tzp = -val; |
1443 | 0 | break; |
1444 | | |
1445 | 0 | case DYNTZ: |
1446 | 0 | tmask |= DTK_M(TZ); |
1447 | 0 | if (tzp == NULL) |
1448 | 0 | return DTERR_BAD_FORMAT; |
1449 | | /* we'll determine the actual offset later */ |
1450 | 0 | abbrevTz = valtz; |
1451 | 0 | abbrev = field[i]; |
1452 | 0 | break; |
1453 | | |
1454 | 0 | case AMPM: |
1455 | 0 | mer = val; |
1456 | 0 | break; |
1457 | | |
1458 | 0 | case ADBC: |
1459 | 0 | bc = (val == BC); |
1460 | 0 | break; |
1461 | | |
1462 | 0 | case DOW: |
1463 | 0 | tm->tm_wday = val; |
1464 | 0 | break; |
1465 | | |
1466 | 0 | case UNITS: |
1467 | 0 | tmask = 0; |
1468 | | /* reject consecutive unhandled units */ |
1469 | 0 | if (ptype != 0) |
1470 | 0 | return DTERR_BAD_FORMAT; |
1471 | 0 | ptype = val; |
1472 | 0 | break; |
1473 | | |
1474 | 0 | case ISOTIME: |
1475 | | |
1476 | | /* |
1477 | | * This is a filler field "t" indicating that the next |
1478 | | * field is time. Try to verify that this is sensible. |
1479 | | */ |
1480 | 0 | tmask = 0; |
1481 | | |
1482 | | /* No preceding date? Then quit... */ |
1483 | 0 | if ((fmask & DTK_DATE_M) != DTK_DATE_M) |
1484 | 0 | return DTERR_BAD_FORMAT; |
1485 | | |
1486 | | /* reject consecutive unhandled units */ |
1487 | 0 | if (ptype != 0) |
1488 | 0 | return DTERR_BAD_FORMAT; |
1489 | 0 | ptype = val; |
1490 | 0 | break; |
1491 | | |
1492 | 0 | case UNKNOWN_FIELD: |
1493 | | |
1494 | | /* |
1495 | | * Before giving up and declaring error, check to see |
1496 | | * if it is an all-alpha timezone name. |
1497 | | */ |
1498 | 0 | namedTz = pg_tzset(field[i]); |
1499 | 0 | if (!namedTz) |
1500 | 0 | return DTERR_BAD_FORMAT; |
1501 | | /* we'll apply the zone setting below */ |
1502 | 0 | tmask = DTK_M(TZ); |
1503 | 0 | break; |
1504 | | |
1505 | 0 | default: |
1506 | 0 | return DTERR_BAD_FORMAT; |
1507 | 0 | } |
1508 | 0 | break; |
1509 | | |
1510 | 0 | default: |
1511 | 0 | return DTERR_BAD_FORMAT; |
1512 | 0 | } |
1513 | | |
1514 | 0 | if (tmask & fmask) |
1515 | 0 | return DTERR_BAD_FORMAT; |
1516 | 0 | fmask |= tmask; |
1517 | 0 | } /* end loop over fields */ |
1518 | | |
1519 | | /* reject if prefix type appeared and was never handled */ |
1520 | 0 | if (ptype != 0) |
1521 | 0 | return DTERR_BAD_FORMAT; |
1522 | | |
1523 | | /* do additional checking for normal date specs (but not "infinity" etc) */ |
1524 | 0 | if (*dtype == DTK_DATE) |
1525 | 0 | { |
1526 | | /* do final checking/adjustment of Y/M/D fields */ |
1527 | 0 | dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm); |
1528 | 0 | if (dterr) |
1529 | 0 | return dterr; |
1530 | | |
1531 | | /* handle AM/PM */ |
1532 | 0 | if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2) |
1533 | 0 | return DTERR_FIELD_OVERFLOW; |
1534 | 0 | if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2) |
1535 | 0 | tm->tm_hour = 0; |
1536 | 0 | else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2) |
1537 | 0 | tm->tm_hour += HOURS_PER_DAY / 2; |
1538 | | |
1539 | | /* check for incomplete input */ |
1540 | 0 | if ((fmask & DTK_DATE_M) != DTK_DATE_M) |
1541 | 0 | { |
1542 | 0 | if ((fmask & DTK_TIME_M) == DTK_TIME_M) |
1543 | 0 | return 1; |
1544 | 0 | return DTERR_BAD_FORMAT; |
1545 | 0 | } |
1546 | | |
1547 | | /* |
1548 | | * If we had a full timezone spec, compute the offset (we could not do |
1549 | | * it before, because we need the date to resolve DST status). |
1550 | | */ |
1551 | 0 | if (namedTz != NULL) |
1552 | 0 | { |
1553 | | /* daylight savings time modifier disallowed with full TZ */ |
1554 | 0 | if (fmask & DTK_M(DTZMOD)) |
1555 | 0 | return DTERR_BAD_FORMAT; |
1556 | | |
1557 | 0 | *tzp = DetermineTimeZoneOffset(tm, namedTz); |
1558 | 0 | } |
1559 | | |
1560 | | /* |
1561 | | * Likewise, if we had a dynamic timezone abbreviation, resolve it |
1562 | | * now. |
1563 | | */ |
1564 | 0 | if (abbrevTz != NULL) |
1565 | 0 | { |
1566 | | /* daylight savings time modifier disallowed with dynamic TZ */ |
1567 | 0 | if (fmask & DTK_M(DTZMOD)) |
1568 | 0 | return DTERR_BAD_FORMAT; |
1569 | | |
1570 | 0 | *tzp = DetermineTimeZoneAbbrevOffset(tm, abbrev, abbrevTz); |
1571 | 0 | } |
1572 | | |
1573 | | /* timezone not specified? then use session timezone */ |
1574 | 0 | if (tzp != NULL && !(fmask & DTK_M(TZ))) |
1575 | 0 | { |
1576 | | /* |
1577 | | * daylight savings time modifier but no standard timezone? then |
1578 | | * error |
1579 | | */ |
1580 | 0 | if (fmask & DTK_M(DTZMOD)) |
1581 | 0 | return DTERR_BAD_FORMAT; |
1582 | | |
1583 | 0 | *tzp = DetermineTimeZoneOffset(tm, session_timezone); |
1584 | 0 | } |
1585 | 0 | } |
1586 | | |
1587 | 0 | return 0; |
1588 | 0 | } |
1589 | | |
1590 | | |
1591 | | /* DetermineTimeZoneOffset() |
1592 | | * |
1593 | | * Given a struct pg_tm in which tm_year, tm_mon, tm_mday, tm_hour, tm_min, |
1594 | | * and tm_sec fields are set, and a zic-style time zone definition, determine |
1595 | | * the applicable GMT offset and daylight-savings status at that time. |
1596 | | * Set the struct pg_tm's tm_isdst field accordingly, and return the GMT |
1597 | | * offset as the function result. |
1598 | | * |
1599 | | * Note: if the date is out of the range we can deal with, we return zero |
1600 | | * as the GMT offset and set tm_isdst = 0. We don't throw an error here, |
1601 | | * though probably some higher-level code will. |
1602 | | */ |
1603 | | int |
1604 | | DetermineTimeZoneOffset(struct pg_tm *tm, pg_tz *tzp) |
1605 | 0 | { |
1606 | 0 | pg_time_t t; |
1607 | |
|
1608 | 0 | return DetermineTimeZoneOffsetInternal(tm, tzp, &t); |
1609 | 0 | } |
1610 | | |
1611 | | |
1612 | | /* DetermineTimeZoneOffsetInternal() |
1613 | | * |
1614 | | * As above, but also return the actual UTC time imputed to the date/time |
1615 | | * into *tp. |
1616 | | * |
1617 | | * In event of an out-of-range date, we punt by returning zero into *tp. |
1618 | | * This is okay for the immediate callers but is a good reason for not |
1619 | | * exposing this worker function globally. |
1620 | | * |
1621 | | * Note: it might seem that we should use mktime() for this, but bitter |
1622 | | * experience teaches otherwise. This code is much faster than most versions |
1623 | | * of mktime(), anyway. |
1624 | | */ |
1625 | | static int |
1626 | | DetermineTimeZoneOffsetInternal(struct pg_tm *tm, pg_tz *tzp, pg_time_t *tp) |
1627 | 0 | { |
1628 | 0 | int date, |
1629 | 0 | sec; |
1630 | 0 | pg_time_t day, |
1631 | 0 | mytime, |
1632 | 0 | prevtime, |
1633 | 0 | boundary, |
1634 | 0 | beforetime, |
1635 | 0 | aftertime; |
1636 | 0 | long int before_gmtoff, |
1637 | 0 | after_gmtoff; |
1638 | 0 | int before_isdst, |
1639 | 0 | after_isdst; |
1640 | 0 | int res; |
1641 | | |
1642 | | /* |
1643 | | * First, generate the pg_time_t value corresponding to the given |
1644 | | * y/m/d/h/m/s taken as GMT time. If this overflows, punt and decide the |
1645 | | * timezone is GMT. (For a valid Julian date, integer overflow should be |
1646 | | * impossible with 64-bit pg_time_t, but let's check for safety.) |
1647 | | */ |
1648 | 0 | if (!IS_VALID_JULIAN(tm->tm_year, tm->tm_mon, tm->tm_mday)) |
1649 | 0 | goto overflow; |
1650 | 0 | date = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday) - UNIX_EPOCH_JDATE; |
1651 | |
|
1652 | 0 | day = ((pg_time_t) date) * SECS_PER_DAY; |
1653 | 0 | if (day / SECS_PER_DAY != date) |
1654 | 0 | goto overflow; |
1655 | 0 | sec = tm->tm_sec + (tm->tm_min + tm->tm_hour * MINS_PER_HOUR) * SECS_PER_MINUTE; |
1656 | 0 | mytime = day + sec; |
1657 | | /* since sec >= 0, overflow could only be from +day to -mytime */ |
1658 | 0 | if (mytime < 0 && day > 0) |
1659 | 0 | goto overflow; |
1660 | | |
1661 | | /* |
1662 | | * Find the DST time boundary just before or following the target time. We |
1663 | | * assume that all zones have GMT offsets less than 24 hours, and that DST |
1664 | | * boundaries can't be closer together than 48 hours, so backing up 24 |
1665 | | * hours and finding the "next" boundary will work. |
1666 | | */ |
1667 | 0 | prevtime = mytime - SECS_PER_DAY; |
1668 | 0 | if (mytime < 0 && prevtime > 0) |
1669 | 0 | goto overflow; |
1670 | | |
1671 | 0 | res = pg_next_dst_boundary(&prevtime, |
1672 | 0 | &before_gmtoff, &before_isdst, |
1673 | 0 | &boundary, |
1674 | 0 | &after_gmtoff, &after_isdst, |
1675 | 0 | tzp); |
1676 | 0 | if (res < 0) |
1677 | 0 | goto overflow; /* failure? */ |
1678 | | |
1679 | 0 | if (res == 0) |
1680 | 0 | { |
1681 | | /* Non-DST zone, life is simple */ |
1682 | 0 | tm->tm_isdst = before_isdst; |
1683 | 0 | *tp = mytime - before_gmtoff; |
1684 | 0 | return -(int) before_gmtoff; |
1685 | 0 | } |
1686 | | |
1687 | | /* |
1688 | | * Form the candidate pg_time_t values with local-time adjustment |
1689 | | */ |
1690 | 0 | beforetime = mytime - before_gmtoff; |
1691 | 0 | if ((before_gmtoff > 0 && |
1692 | 0 | mytime < 0 && beforetime > 0) || |
1693 | 0 | (before_gmtoff <= 0 && |
1694 | 0 | mytime > 0 && beforetime < 0)) |
1695 | 0 | goto overflow; |
1696 | 0 | aftertime = mytime - after_gmtoff; |
1697 | 0 | if ((after_gmtoff > 0 && |
1698 | 0 | mytime < 0 && aftertime > 0) || |
1699 | 0 | (after_gmtoff <= 0 && |
1700 | 0 | mytime > 0 && aftertime < 0)) |
1701 | 0 | goto overflow; |
1702 | | |
1703 | | /* |
1704 | | * If both before or both after the boundary time, we know what to do. The |
1705 | | * boundary time itself is considered to be after the transition, which |
1706 | | * means we can accept aftertime == boundary in the second case. |
1707 | | */ |
1708 | 0 | if (beforetime < boundary && aftertime < boundary) |
1709 | 0 | { |
1710 | 0 | tm->tm_isdst = before_isdst; |
1711 | 0 | *tp = beforetime; |
1712 | 0 | return -(int) before_gmtoff; |
1713 | 0 | } |
1714 | 0 | if (beforetime > boundary && aftertime >= boundary) |
1715 | 0 | { |
1716 | 0 | tm->tm_isdst = after_isdst; |
1717 | 0 | *tp = aftertime; |
1718 | 0 | return -(int) after_gmtoff; |
1719 | 0 | } |
1720 | | |
1721 | | /* |
1722 | | * It's an invalid or ambiguous time due to timezone transition. In a |
1723 | | * spring-forward transition, prefer the "before" interpretation; in a |
1724 | | * fall-back transition, prefer "after". (We used to define and implement |
1725 | | * this test as "prefer the standard-time interpretation", but that rule |
1726 | | * does not help to resolve the behavior when both times are reported as |
1727 | | * standard time; which does happen, eg Europe/Moscow in Oct 2014. Also, |
1728 | | * in some zones such as Europe/Dublin, there is widespread confusion |
1729 | | * about which time offset is "standard" time, so it's fortunate that our |
1730 | | * behavior doesn't depend on that.) |
1731 | | */ |
1732 | 0 | if (beforetime > aftertime) |
1733 | 0 | { |
1734 | 0 | tm->tm_isdst = before_isdst; |
1735 | 0 | *tp = beforetime; |
1736 | 0 | return -(int) before_gmtoff; |
1737 | 0 | } |
1738 | 0 | tm->tm_isdst = after_isdst; |
1739 | 0 | *tp = aftertime; |
1740 | 0 | return -(int) after_gmtoff; |
1741 | | |
1742 | 0 | overflow: |
1743 | | /* Given date is out of range, so assume UTC */ |
1744 | 0 | tm->tm_isdst = 0; |
1745 | 0 | *tp = 0; |
1746 | 0 | return 0; |
1747 | 0 | } |
1748 | | |
1749 | | |
1750 | | /* DetermineTimeZoneAbbrevOffset() |
1751 | | * |
1752 | | * Determine the GMT offset and DST flag to be attributed to a dynamic |
1753 | | * time zone abbreviation, that is one whose meaning has changed over time. |
1754 | | * *tm contains the local time at which the meaning should be determined, |
1755 | | * and tm->tm_isdst receives the DST flag. |
1756 | | * |
1757 | | * This differs from the behavior of DetermineTimeZoneOffset() in that a |
1758 | | * standard-time or daylight-time abbreviation forces use of the corresponding |
1759 | | * GMT offset even when the zone was then in DS or standard time respectively. |
1760 | | * (However, that happens only if we can match the given abbreviation to some |
1761 | | * abbreviation that appears in the IANA timezone data. Otherwise, we fall |
1762 | | * back to doing DetermineTimeZoneOffset().) |
1763 | | */ |
1764 | | int |
1765 | | DetermineTimeZoneAbbrevOffset(struct pg_tm *tm, const char *abbr, pg_tz *tzp) |
1766 | 0 | { |
1767 | 0 | pg_time_t t; |
1768 | 0 | int zone_offset; |
1769 | 0 | int abbr_offset; |
1770 | 0 | int abbr_isdst; |
1771 | | |
1772 | | /* |
1773 | | * Compute the UTC time we want to probe at. (In event of overflow, we'll |
1774 | | * probe at the epoch, which is a bit random but probably doesn't matter.) |
1775 | | */ |
1776 | 0 | zone_offset = DetermineTimeZoneOffsetInternal(tm, tzp, &t); |
1777 | | |
1778 | | /* |
1779 | | * Try to match the abbreviation to something in the zone definition. |
1780 | | */ |
1781 | 0 | if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp, |
1782 | 0 | &abbr_offset, &abbr_isdst)) |
1783 | 0 | { |
1784 | | /* Success, so use the abbrev-specific answers. */ |
1785 | 0 | tm->tm_isdst = abbr_isdst; |
1786 | 0 | return abbr_offset; |
1787 | 0 | } |
1788 | | |
1789 | | /* |
1790 | | * No match, so use the answers we already got from |
1791 | | * DetermineTimeZoneOffsetInternal. |
1792 | | */ |
1793 | 0 | return zone_offset; |
1794 | 0 | } |
1795 | | |
1796 | | |
1797 | | /* DetermineTimeZoneAbbrevOffsetTS() |
1798 | | * |
1799 | | * As above but the probe time is specified as a TimestampTz (hence, UTC time), |
1800 | | * and DST status is returned into *isdst rather than into tm->tm_isdst. |
1801 | | */ |
1802 | | int |
1803 | | DetermineTimeZoneAbbrevOffsetTS(TimestampTz ts, const char *abbr, |
1804 | | pg_tz *tzp, int *isdst) |
1805 | 0 | { |
1806 | 0 | pg_time_t t = timestamptz_to_time_t(ts); |
1807 | 0 | int zone_offset; |
1808 | 0 | int abbr_offset; |
1809 | 0 | int tz; |
1810 | 0 | struct pg_tm tm; |
1811 | 0 | fsec_t fsec; |
1812 | | |
1813 | | /* |
1814 | | * If the abbrev matches anything in the zone data, this is pretty easy. |
1815 | | */ |
1816 | 0 | if (DetermineTimeZoneAbbrevOffsetInternal(t, abbr, tzp, |
1817 | 0 | &abbr_offset, isdst)) |
1818 | 0 | return abbr_offset; |
1819 | | |
1820 | | /* |
1821 | | * Else, break down the timestamp so we can use DetermineTimeZoneOffset. |
1822 | | */ |
1823 | 0 | if (timestamp2tm(ts, &tz, &tm, &fsec, NULL, tzp) != 0) |
1824 | 0 | ereport(ERROR, |
1825 | 0 | (errcode(ERRCODE_DATETIME_VALUE_OUT_OF_RANGE), |
1826 | 0 | errmsg("timestamp out of range"))); |
1827 | | |
1828 | 0 | zone_offset = DetermineTimeZoneOffset(&tm, tzp); |
1829 | 0 | *isdst = tm.tm_isdst; |
1830 | 0 | return zone_offset; |
1831 | 0 | } |
1832 | | |
1833 | | |
1834 | | /* DetermineTimeZoneAbbrevOffsetInternal() |
1835 | | * |
1836 | | * Workhorse for above two functions: work from a pg_time_t probe instant. |
1837 | | * On success, return GMT offset and DST status into *offset and *isdst. |
1838 | | */ |
1839 | | static bool |
1840 | | DetermineTimeZoneAbbrevOffsetInternal(pg_time_t t, const char *abbr, pg_tz *tzp, |
1841 | | int *offset, int *isdst) |
1842 | 0 | { |
1843 | 0 | char upabbr[TZ_STRLEN_MAX + 1]; |
1844 | 0 | unsigned char *p; |
1845 | 0 | long int gmtoff; |
1846 | | |
1847 | | /* We need to force the abbrev to upper case */ |
1848 | 0 | strlcpy(upabbr, abbr, sizeof(upabbr)); |
1849 | 0 | for (p = (unsigned char *) upabbr; *p; p++) |
1850 | 0 | *p = pg_toupper(*p); |
1851 | | |
1852 | | /* Look up the abbrev's meaning at this time in this zone */ |
1853 | 0 | if (pg_interpret_timezone_abbrev(upabbr, |
1854 | 0 | &t, |
1855 | 0 | &gmtoff, |
1856 | 0 | isdst, |
1857 | 0 | tzp)) |
1858 | 0 | { |
1859 | | /* Change sign to agree with DetermineTimeZoneOffset() */ |
1860 | 0 | *offset = (int) -gmtoff; |
1861 | 0 | return true; |
1862 | 0 | } |
1863 | 0 | return false; |
1864 | 0 | } |
1865 | | |
1866 | | |
1867 | | /* TimeZoneAbbrevIsKnown() |
1868 | | * |
1869 | | * Detect whether the given string is a time zone abbreviation that's known |
1870 | | * in the specified TZDB timezone, and if so whether it's fixed or varying |
1871 | | * meaning. The match is not case-sensitive. |
1872 | | */ |
1873 | | static bool |
1874 | | TimeZoneAbbrevIsKnown(const char *abbr, pg_tz *tzp, |
1875 | | bool *isfixed, int *offset, int *isdst) |
1876 | 0 | { |
1877 | 0 | char upabbr[TZ_STRLEN_MAX + 1]; |
1878 | 0 | unsigned char *p; |
1879 | 0 | long int gmtoff; |
1880 | | |
1881 | | /* We need to force the abbrev to upper case */ |
1882 | 0 | strlcpy(upabbr, abbr, sizeof(upabbr)); |
1883 | 0 | for (p = (unsigned char *) upabbr; *p; p++) |
1884 | 0 | *p = pg_toupper(*p); |
1885 | | |
1886 | | /* Look up the abbrev's meaning in this zone */ |
1887 | 0 | if (pg_timezone_abbrev_is_known(upabbr, |
1888 | 0 | isfixed, |
1889 | 0 | &gmtoff, |
1890 | 0 | isdst, |
1891 | 0 | tzp)) |
1892 | 0 | { |
1893 | | /* Change sign to agree with DetermineTimeZoneOffset() */ |
1894 | 0 | *offset = (int) -gmtoff; |
1895 | 0 | return true; |
1896 | 0 | } |
1897 | 0 | return false; |
1898 | 0 | } |
1899 | | |
1900 | | |
1901 | | /* DecodeTimeOnly() |
1902 | | * Interpret parsed string as time fields only. |
1903 | | * Returns 0 if successful, DTERR code if bogus input detected. |
1904 | | * |
1905 | | * Inputs are field[] and ftype[] arrays, of length nf. |
1906 | | * Other arguments are outputs. |
1907 | | * |
1908 | | * Note that support for time zone is here for |
1909 | | * SQL TIME WITH TIME ZONE, but it reveals |
1910 | | * bogosity with SQL date/time standards, since |
1911 | | * we must infer a time zone from current time. |
1912 | | * - thomas 2000-03-10 |
1913 | | * Allow specifying date to get a better time zone, |
1914 | | * if time zones are allowed. - thomas 2001-12-26 |
1915 | | */ |
1916 | | int |
1917 | | DecodeTimeOnly(char **field, int *ftype, int nf, |
1918 | | int *dtype, struct pg_tm *tm, fsec_t *fsec, int *tzp, |
1919 | | DateTimeErrorExtra *extra) |
1920 | 0 | { |
1921 | 0 | int fmask = 0, |
1922 | 0 | tmask, |
1923 | 0 | type; |
1924 | 0 | int ptype = 0; /* "prefix type" for ISO and Julian formats */ |
1925 | 0 | int i; |
1926 | 0 | int val; |
1927 | 0 | int dterr; |
1928 | 0 | bool isjulian = false; |
1929 | 0 | bool is2digits = false; |
1930 | 0 | bool bc = false; |
1931 | 0 | int mer = HR24; |
1932 | 0 | pg_tz *namedTz = NULL; |
1933 | 0 | pg_tz *abbrevTz = NULL; |
1934 | 0 | char *abbrev = NULL; |
1935 | 0 | pg_tz *valtz; |
1936 | |
|
1937 | 0 | *dtype = DTK_TIME; |
1938 | 0 | tm->tm_hour = 0; |
1939 | 0 | tm->tm_min = 0; |
1940 | 0 | tm->tm_sec = 0; |
1941 | 0 | *fsec = 0; |
1942 | | /* don't know daylight savings time status apriori */ |
1943 | 0 | tm->tm_isdst = -1; |
1944 | |
|
1945 | 0 | if (tzp != NULL) |
1946 | 0 | *tzp = 0; |
1947 | |
|
1948 | 0 | for (i = 0; i < nf; i++) |
1949 | 0 | { |
1950 | 0 | switch (ftype[i]) |
1951 | 0 | { |
1952 | 0 | case DTK_DATE: |
1953 | | |
1954 | | /* |
1955 | | * Time zone not allowed? Then should not accept dates or time |
1956 | | * zones no matter what else! |
1957 | | */ |
1958 | 0 | if (tzp == NULL) |
1959 | 0 | return DTERR_BAD_FORMAT; |
1960 | | |
1961 | | /* Under limited circumstances, we will accept a date... */ |
1962 | 0 | if (i == 0 && nf >= 2 && |
1963 | 0 | (ftype[nf - 1] == DTK_DATE || ftype[1] == DTK_TIME)) |
1964 | 0 | { |
1965 | 0 | dterr = DecodeDate(field[i], fmask, |
1966 | 0 | &tmask, &is2digits, tm); |
1967 | 0 | if (dterr) |
1968 | 0 | return dterr; |
1969 | 0 | } |
1970 | | /* otherwise, this is a time and/or time zone */ |
1971 | 0 | else |
1972 | 0 | { |
1973 | 0 | if (isdigit((unsigned char) *field[i])) |
1974 | 0 | { |
1975 | 0 | char *cp; |
1976 | | |
1977 | | /* |
1978 | | * Starts with a digit but we already have a time |
1979 | | * field? Then we are in trouble with time already... |
1980 | | */ |
1981 | 0 | if ((fmask & DTK_TIME_M) == DTK_TIME_M) |
1982 | 0 | return DTERR_BAD_FORMAT; |
1983 | | |
1984 | | /* |
1985 | | * Should not get here and fail. Sanity check only... |
1986 | | */ |
1987 | 0 | if ((cp = strchr(field[i], '-')) == NULL) |
1988 | 0 | return DTERR_BAD_FORMAT; |
1989 | | |
1990 | | /* Get the time zone from the end of the string */ |
1991 | 0 | dterr = DecodeTimezone(cp, tzp); |
1992 | 0 | if (dterr) |
1993 | 0 | return dterr; |
1994 | 0 | *cp = '\0'; |
1995 | | |
1996 | | /* |
1997 | | * Then read the rest of the field as a concatenated |
1998 | | * time |
1999 | | */ |
2000 | 0 | dterr = DecodeNumberField(strlen(field[i]), field[i], |
2001 | 0 | (fmask | DTK_DATE_M), |
2002 | 0 | &tmask, tm, |
2003 | 0 | fsec, &is2digits); |
2004 | 0 | if (dterr < 0) |
2005 | 0 | return dterr; |
2006 | 0 | ftype[i] = dterr; |
2007 | |
|
2008 | 0 | tmask |= DTK_M(TZ); |
2009 | 0 | } |
2010 | 0 | else |
2011 | 0 | { |
2012 | 0 | namedTz = pg_tzset(field[i]); |
2013 | 0 | if (!namedTz) |
2014 | 0 | { |
2015 | 0 | extra->dtee_timezone = field[i]; |
2016 | 0 | return DTERR_BAD_TIMEZONE; |
2017 | 0 | } |
2018 | | /* we'll apply the zone setting below */ |
2019 | 0 | ftype[i] = DTK_TZ; |
2020 | 0 | tmask = DTK_M(TZ); |
2021 | 0 | } |
2022 | 0 | } |
2023 | 0 | break; |
2024 | | |
2025 | 0 | case DTK_TIME: |
2026 | | |
2027 | | /* |
2028 | | * This might be an ISO time following a "t" field. |
2029 | | */ |
2030 | 0 | if (ptype != 0) |
2031 | 0 | { |
2032 | 0 | if (ptype != DTK_TIME) |
2033 | 0 | return DTERR_BAD_FORMAT; |
2034 | 0 | ptype = 0; |
2035 | 0 | } |
2036 | | |
2037 | 0 | dterr = DecodeTime(field[i], (fmask | DTK_DATE_M), |
2038 | 0 | INTERVAL_FULL_RANGE, |
2039 | 0 | &tmask, tm, fsec); |
2040 | 0 | if (dterr) |
2041 | 0 | return dterr; |
2042 | 0 | break; |
2043 | | |
2044 | 0 | case DTK_TZ: |
2045 | 0 | { |
2046 | 0 | int tz; |
2047 | |
|
2048 | 0 | if (tzp == NULL) |
2049 | 0 | return DTERR_BAD_FORMAT; |
2050 | | |
2051 | 0 | dterr = DecodeTimezone(field[i], &tz); |
2052 | 0 | if (dterr) |
2053 | 0 | return dterr; |
2054 | 0 | *tzp = tz; |
2055 | 0 | tmask = DTK_M(TZ); |
2056 | 0 | } |
2057 | 0 | break; |
2058 | | |
2059 | 0 | case DTK_NUMBER: |
2060 | | |
2061 | | /* |
2062 | | * Deal with cases where previous field labeled this one |
2063 | | */ |
2064 | 0 | if (ptype != 0) |
2065 | 0 | { |
2066 | 0 | char *cp; |
2067 | 0 | int value; |
2068 | |
|
2069 | 0 | errno = 0; |
2070 | 0 | value = strtoint(field[i], &cp, 10); |
2071 | 0 | if (errno == ERANGE) |
2072 | 0 | return DTERR_FIELD_OVERFLOW; |
2073 | 0 | if (*cp != '.' && *cp != '\0') |
2074 | 0 | return DTERR_BAD_FORMAT; |
2075 | | |
2076 | 0 | switch (ptype) |
2077 | 0 | { |
2078 | 0 | case DTK_JULIAN: |
2079 | | /* previous field was a label for "julian date" */ |
2080 | 0 | if (tzp == NULL) |
2081 | 0 | return DTERR_BAD_FORMAT; |
2082 | 0 | if (value < 0) |
2083 | 0 | return DTERR_FIELD_OVERFLOW; |
2084 | 0 | tmask = DTK_DATE_M; |
2085 | 0 | j2date(value, &tm->tm_year, &tm->tm_mon, &tm->tm_mday); |
2086 | 0 | isjulian = true; |
2087 | |
|
2088 | 0 | if (*cp == '.') |
2089 | 0 | { |
2090 | 0 | double time; |
2091 | |
|
2092 | 0 | dterr = ParseFraction(cp, &time); |
2093 | 0 | if (dterr) |
2094 | 0 | return dterr; |
2095 | 0 | time *= USECS_PER_DAY; |
2096 | 0 | dt2time(time, |
2097 | 0 | &tm->tm_hour, &tm->tm_min, |
2098 | 0 | &tm->tm_sec, fsec); |
2099 | 0 | tmask |= DTK_TIME_M; |
2100 | 0 | } |
2101 | 0 | break; |
2102 | | |
2103 | 0 | case DTK_TIME: |
2104 | | /* previous field was "t" for ISO time */ |
2105 | 0 | dterr = DecodeNumberField(strlen(field[i]), field[i], |
2106 | 0 | (fmask | DTK_DATE_M), |
2107 | 0 | &tmask, tm, |
2108 | 0 | fsec, &is2digits); |
2109 | 0 | if (dterr < 0) |
2110 | 0 | return dterr; |
2111 | 0 | ftype[i] = dterr; |
2112 | |
|
2113 | 0 | if (tmask != DTK_TIME_M) |
2114 | 0 | return DTERR_BAD_FORMAT; |
2115 | 0 | break; |
2116 | | |
2117 | 0 | default: |
2118 | 0 | return DTERR_BAD_FORMAT; |
2119 | 0 | break; |
2120 | 0 | } |
2121 | | |
2122 | 0 | ptype = 0; |
2123 | 0 | *dtype = DTK_DATE; |
2124 | 0 | } |
2125 | 0 | else |
2126 | 0 | { |
2127 | 0 | char *cp; |
2128 | 0 | int flen; |
2129 | |
|
2130 | 0 | flen = strlen(field[i]); |
2131 | 0 | cp = strchr(field[i], '.'); |
2132 | | |
2133 | | /* Embedded decimal? */ |
2134 | 0 | if (cp != NULL) |
2135 | 0 | { |
2136 | | /* |
2137 | | * Under limited circumstances, we will accept a |
2138 | | * date... |
2139 | | */ |
2140 | 0 | if (i == 0 && nf >= 2 && ftype[nf - 1] == DTK_DATE) |
2141 | 0 | { |
2142 | 0 | dterr = DecodeDate(field[i], fmask, |
2143 | 0 | &tmask, &is2digits, tm); |
2144 | 0 | if (dterr) |
2145 | 0 | return dterr; |
2146 | 0 | } |
2147 | | /* embedded decimal and several digits before? */ |
2148 | 0 | else if (flen - strlen(cp) > 2) |
2149 | 0 | { |
2150 | | /* |
2151 | | * Interpret as a concatenated date or time Set |
2152 | | * the type field to allow decoding other fields |
2153 | | * later. Example: 20011223 or 040506 |
2154 | | */ |
2155 | 0 | dterr = DecodeNumberField(flen, field[i], |
2156 | 0 | (fmask | DTK_DATE_M), |
2157 | 0 | &tmask, tm, |
2158 | 0 | fsec, &is2digits); |
2159 | 0 | if (dterr < 0) |
2160 | 0 | return dterr; |
2161 | 0 | ftype[i] = dterr; |
2162 | 0 | } |
2163 | 0 | else |
2164 | 0 | return DTERR_BAD_FORMAT; |
2165 | 0 | } |
2166 | 0 | else if (flen > 4) |
2167 | 0 | { |
2168 | 0 | dterr = DecodeNumberField(flen, field[i], |
2169 | 0 | (fmask | DTK_DATE_M), |
2170 | 0 | &tmask, tm, |
2171 | 0 | fsec, &is2digits); |
2172 | 0 | if (dterr < 0) |
2173 | 0 | return dterr; |
2174 | 0 | ftype[i] = dterr; |
2175 | 0 | } |
2176 | | /* otherwise it is a single date/time field... */ |
2177 | 0 | else |
2178 | 0 | { |
2179 | 0 | dterr = DecodeNumber(flen, field[i], |
2180 | 0 | false, |
2181 | 0 | (fmask | DTK_DATE_M), |
2182 | 0 | &tmask, tm, |
2183 | 0 | fsec, &is2digits); |
2184 | 0 | if (dterr) |
2185 | 0 | return dterr; |
2186 | 0 | } |
2187 | 0 | } |
2188 | 0 | break; |
2189 | | |
2190 | 0 | case DTK_STRING: |
2191 | 0 | case DTK_SPECIAL: |
2192 | | /* timezone abbrevs take precedence over built-in tokens */ |
2193 | 0 | dterr = DecodeTimezoneAbbrev(i, field[i], |
2194 | 0 | &type, &val, &valtz, extra); |
2195 | 0 | if (dterr) |
2196 | 0 | return dterr; |
2197 | 0 | if (type == UNKNOWN_FIELD) |
2198 | 0 | type = DecodeSpecial(i, field[i], &val); |
2199 | 0 | if (type == IGNORE_DTF) |
2200 | 0 | continue; |
2201 | | |
2202 | 0 | tmask = DTK_M(type); |
2203 | 0 | switch (type) |
2204 | 0 | { |
2205 | 0 | case RESERV: |
2206 | 0 | switch (val) |
2207 | 0 | { |
2208 | 0 | case DTK_NOW: |
2209 | 0 | tmask = DTK_TIME_M; |
2210 | 0 | *dtype = DTK_TIME; |
2211 | 0 | GetCurrentTimeUsec(tm, fsec, NULL); |
2212 | 0 | break; |
2213 | | |
2214 | 0 | case DTK_ZULU: |
2215 | 0 | tmask = (DTK_TIME_M | DTK_M(TZ)); |
2216 | 0 | *dtype = DTK_TIME; |
2217 | 0 | tm->tm_hour = 0; |
2218 | 0 | tm->tm_min = 0; |
2219 | 0 | tm->tm_sec = 0; |
2220 | 0 | tm->tm_isdst = 0; |
2221 | 0 | break; |
2222 | | |
2223 | 0 | default: |
2224 | 0 | return DTERR_BAD_FORMAT; |
2225 | 0 | } |
2226 | | |
2227 | 0 | break; |
2228 | | |
2229 | 0 | case DTZMOD: |
2230 | | |
2231 | | /* |
2232 | | * daylight savings time modifier (solves "MET DST" |
2233 | | * syntax) |
2234 | | */ |
2235 | 0 | tmask |= DTK_M(DTZ); |
2236 | 0 | tm->tm_isdst = 1; |
2237 | 0 | if (tzp == NULL) |
2238 | 0 | return DTERR_BAD_FORMAT; |
2239 | 0 | *tzp -= val; |
2240 | 0 | break; |
2241 | | |
2242 | 0 | case DTZ: |
2243 | | |
2244 | | /* |
2245 | | * set mask for TZ here _or_ check for DTZ later when |
2246 | | * getting default timezone |
2247 | | */ |
2248 | 0 | tmask |= DTK_M(TZ); |
2249 | 0 | tm->tm_isdst = 1; |
2250 | 0 | if (tzp == NULL) |
2251 | 0 | return DTERR_BAD_FORMAT; |
2252 | 0 | *tzp = -val; |
2253 | 0 | ftype[i] = DTK_TZ; |
2254 | 0 | break; |
2255 | | |
2256 | 0 | case TZ: |
2257 | 0 | tm->tm_isdst = 0; |
2258 | 0 | if (tzp == NULL) |
2259 | 0 | return DTERR_BAD_FORMAT; |
2260 | 0 | *tzp = -val; |
2261 | 0 | ftype[i] = DTK_TZ; |
2262 | 0 | break; |
2263 | | |
2264 | 0 | case DYNTZ: |
2265 | 0 | tmask |= DTK_M(TZ); |
2266 | 0 | if (tzp == NULL) |
2267 | 0 | return DTERR_BAD_FORMAT; |
2268 | | /* we'll determine the actual offset later */ |
2269 | 0 | abbrevTz = valtz; |
2270 | 0 | abbrev = field[i]; |
2271 | 0 | ftype[i] = DTK_TZ; |
2272 | 0 | break; |
2273 | | |
2274 | 0 | case AMPM: |
2275 | 0 | mer = val; |
2276 | 0 | break; |
2277 | | |
2278 | 0 | case ADBC: |
2279 | 0 | bc = (val == BC); |
2280 | 0 | break; |
2281 | | |
2282 | 0 | case UNITS: |
2283 | 0 | tmask = 0; |
2284 | | /* reject consecutive unhandled units */ |
2285 | 0 | if (ptype != 0) |
2286 | 0 | return DTERR_BAD_FORMAT; |
2287 | 0 | ptype = val; |
2288 | 0 | break; |
2289 | | |
2290 | 0 | case ISOTIME: |
2291 | 0 | tmask = 0; |
2292 | | /* reject consecutive unhandled units */ |
2293 | 0 | if (ptype != 0) |
2294 | 0 | return DTERR_BAD_FORMAT; |
2295 | 0 | ptype = val; |
2296 | 0 | break; |
2297 | | |
2298 | 0 | case UNKNOWN_FIELD: |
2299 | | |
2300 | | /* |
2301 | | * Before giving up and declaring error, check to see |
2302 | | * if it is an all-alpha timezone name. |
2303 | | */ |
2304 | 0 | namedTz = pg_tzset(field[i]); |
2305 | 0 | if (!namedTz) |
2306 | 0 | return DTERR_BAD_FORMAT; |
2307 | | /* we'll apply the zone setting below */ |
2308 | 0 | tmask = DTK_M(TZ); |
2309 | 0 | break; |
2310 | | |
2311 | 0 | default: |
2312 | 0 | return DTERR_BAD_FORMAT; |
2313 | 0 | } |
2314 | 0 | break; |
2315 | | |
2316 | 0 | default: |
2317 | 0 | return DTERR_BAD_FORMAT; |
2318 | 0 | } |
2319 | | |
2320 | 0 | if (tmask & fmask) |
2321 | 0 | return DTERR_BAD_FORMAT; |
2322 | 0 | fmask |= tmask; |
2323 | 0 | } /* end loop over fields */ |
2324 | | |
2325 | | /* reject if prefix type appeared and was never handled */ |
2326 | 0 | if (ptype != 0) |
2327 | 0 | return DTERR_BAD_FORMAT; |
2328 | | |
2329 | | /* do final checking/adjustment of Y/M/D fields */ |
2330 | 0 | dterr = ValidateDate(fmask, isjulian, is2digits, bc, tm); |
2331 | 0 | if (dterr) |
2332 | 0 | return dterr; |
2333 | | |
2334 | | /* handle AM/PM */ |
2335 | 0 | if (mer != HR24 && tm->tm_hour > HOURS_PER_DAY / 2) |
2336 | 0 | return DTERR_FIELD_OVERFLOW; |
2337 | 0 | if (mer == AM && tm->tm_hour == HOURS_PER_DAY / 2) |
2338 | 0 | tm->tm_hour = 0; |
2339 | 0 | else if (mer == PM && tm->tm_hour != HOURS_PER_DAY / 2) |
2340 | 0 | tm->tm_hour += HOURS_PER_DAY / 2; |
2341 | | |
2342 | | /* check for time overflow */ |
2343 | 0 | if (time_overflows(tm->tm_hour, tm->tm_min, tm->tm_sec, *fsec)) |
2344 | 0 | return DTERR_FIELD_OVERFLOW; |
2345 | | |
2346 | 0 | if ((fmask & DTK_TIME_M) != DTK_TIME_M) |
2347 | 0 | return DTERR_BAD_FORMAT; |
2348 | | |
2349 | | /* |
2350 | | * If we had a full timezone spec, compute the offset (we could not do it |
2351 | | * before, because we may need the date to resolve DST status). |
2352 | | */ |
2353 | 0 | if (namedTz != NULL) |
2354 | 0 | { |
2355 | 0 | long int gmtoff; |
2356 | | |
2357 | | /* daylight savings time modifier disallowed with full TZ */ |
2358 | 0 | if (fmask & DTK_M(DTZMOD)) |
2359 | 0 | return DTERR_BAD_FORMAT; |
2360 | | |
2361 | | /* if non-DST zone, we do not need to know the date */ |
2362 | 0 | if (pg_get_timezone_offset(namedTz, &gmtoff)) |
2363 | 0 | { |
2364 | 0 | *tzp = -(int) gmtoff; |
2365 | 0 | } |
2366 | 0 | else |
2367 | 0 | { |
2368 | | /* a date has to be specified */ |
2369 | 0 | if ((fmask & DTK_DATE_M) != DTK_DATE_M) |
2370 | 0 | return DTERR_BAD_FORMAT; |
2371 | 0 | *tzp = DetermineTimeZoneOffset(tm, namedTz); |
2372 | 0 | } |
2373 | 0 | } |
2374 | | |
2375 | | /* |
2376 | | * Likewise, if we had a dynamic timezone abbreviation, resolve it now. |
2377 | | */ |
2378 | 0 | if (abbrevTz != NULL) |
2379 | 0 | { |
2380 | 0 | struct pg_tm tt, |
2381 | 0 | *tmp = &tt; |
2382 | | |
2383 | | /* |
2384 | | * daylight savings time modifier but no standard timezone? then error |
2385 | | */ |
2386 | 0 | if (fmask & DTK_M(DTZMOD)) |
2387 | 0 | return DTERR_BAD_FORMAT; |
2388 | | |
2389 | 0 | if ((fmask & DTK_DATE_M) == 0) |
2390 | 0 | GetCurrentDateTime(tmp); |
2391 | 0 | else |
2392 | 0 | { |
2393 | | /* a date has to be specified */ |
2394 | 0 | if ((fmask & DTK_DATE_M) != DTK_DATE_M) |
2395 | 0 | return DTERR_BAD_FORMAT; |
2396 | 0 | tmp->tm_year = tm->tm_year; |
2397 | 0 | tmp->tm_mon = tm->tm_mon; |
2398 | 0 | tmp->tm_mday = tm->tm_mday; |
2399 | 0 | } |
2400 | 0 | tmp->tm_hour = tm->tm_hour; |
2401 | 0 | tmp->tm_min = tm->tm_min; |
2402 | 0 | tmp->tm_sec = tm->tm_sec; |
2403 | 0 | *tzp = DetermineTimeZoneAbbrevOffset(tmp, abbrev, abbrevTz); |
2404 | 0 | tm->tm_isdst = tmp->tm_isdst; |
2405 | 0 | } |
2406 | | |
2407 | | /* timezone not specified? then use session timezone */ |
2408 | 0 | if (tzp != NULL && !(fmask & DTK_M(TZ))) |
2409 | 0 | { |
2410 | 0 | struct pg_tm tt, |
2411 | 0 | *tmp = &tt; |
2412 | | |
2413 | | /* |
2414 | | * daylight savings time modifier but no standard timezone? then error |
2415 | | */ |
2416 | 0 | if (fmask & DTK_M(DTZMOD)) |
2417 | 0 | return DTERR_BAD_FORMAT; |
2418 | | |
2419 | 0 | if ((fmask & DTK_DATE_M) == 0) |
2420 | 0 | GetCurrentDateTime(tmp); |
2421 | 0 | else |
2422 | 0 | { |
2423 | | /* a date has to be specified */ |
2424 | 0 | if ((fmask & DTK_DATE_M) != DTK_DATE_M) |
2425 | 0 | return DTERR_BAD_FORMAT; |
2426 | 0 | tmp->tm_year = tm->tm_year; |
2427 | 0 | tmp->tm_mon = tm->tm_mon; |
2428 | 0 | tmp->tm_mday = tm->tm_mday; |
2429 | 0 | } |
2430 | 0 | tmp->tm_hour = tm->tm_hour; |
2431 | 0 | tmp->tm_min = tm->tm_min; |
2432 | 0 | tmp->tm_sec = tm->tm_sec; |
2433 | 0 | *tzp = DetermineTimeZoneOffset(tmp, session_timezone); |
2434 | 0 | tm->tm_isdst = tmp->tm_isdst; |
2435 | 0 | } |
2436 | | |
2437 | 0 | return 0; |
2438 | 0 | } |
2439 | | |
2440 | | /* DecodeDate() |
2441 | | * Decode date string which includes delimiters. |
2442 | | * Return 0 if okay, a DTERR code if not. |
2443 | | * |
2444 | | * str: field to be parsed |
2445 | | * fmask: bitmask for field types already seen |
2446 | | * *tmask: receives bitmask for fields found here |
2447 | | * *is2digits: set to true if we find 2-digit year |
2448 | | * *tm: field values are stored into appropriate members of this struct |
2449 | | */ |
2450 | | static int |
2451 | | DecodeDate(char *str, int fmask, int *tmask, bool *is2digits, |
2452 | | struct pg_tm *tm) |
2453 | 0 | { |
2454 | 0 | fsec_t fsec; |
2455 | 0 | int nf = 0; |
2456 | 0 | int i, |
2457 | 0 | len; |
2458 | 0 | int dterr; |
2459 | 0 | bool haveTextMonth = false; |
2460 | 0 | int type, |
2461 | 0 | val, |
2462 | 0 | dmask = 0; |
2463 | 0 | char *field[MAXDATEFIELDS]; |
2464 | |
|
2465 | 0 | *tmask = 0; |
2466 | | |
2467 | | /* parse this string... */ |
2468 | 0 | while (*str != '\0' && nf < MAXDATEFIELDS) |
2469 | 0 | { |
2470 | | /* skip field separators */ |
2471 | 0 | while (*str != '\0' && !isalnum((unsigned char) *str)) |
2472 | 0 | str++; |
2473 | |
|
2474 | 0 | if (*str == '\0') |
2475 | 0 | return DTERR_BAD_FORMAT; /* end of string after separator */ |
2476 | | |
2477 | 0 | field[nf] = str; |
2478 | 0 | if (isdigit((unsigned char) *str)) |
2479 | 0 | { |
2480 | 0 | while (isdigit((unsigned char) *str)) |
2481 | 0 | str++; |
2482 | 0 | } |
2483 | 0 | else if (isalpha((unsigned char) *str)) |
2484 | 0 | { |
2485 | 0 | while (isalpha((unsigned char) *str)) |
2486 | 0 | str++; |
2487 | 0 | } |
2488 | | |
2489 | | /* Just get rid of any non-digit, non-alpha characters... */ |
2490 | 0 | if (*str != '\0') |
2491 | 0 | *str++ = '\0'; |
2492 | 0 | nf++; |
2493 | 0 | } |
2494 | | |
2495 | | /* look first for text fields, since that will be unambiguous month */ |
2496 | 0 | for (i = 0; i < nf; i++) |
2497 | 0 | { |
2498 | 0 | if (isalpha((unsigned char) *field[i])) |
2499 | 0 | { |
2500 | 0 | type = DecodeSpecial(i, field[i], &val); |
2501 | 0 | if (type == IGNORE_DTF) |
2502 | 0 | continue; |
2503 | | |
2504 | 0 | dmask = DTK_M(type); |
2505 | 0 | switch (type) |
2506 | 0 | { |
2507 | 0 | case MONTH: |
2508 | 0 | tm->tm_mon = val; |
2509 | 0 | haveTextMonth = true; |
2510 | 0 | break; |
2511 | | |
2512 | 0 | default: |
2513 | 0 | return DTERR_BAD_FORMAT; |
2514 | 0 | } |
2515 | 0 | if (fmask & dmask) |
2516 | 0 | return DTERR_BAD_FORMAT; |
2517 | | |
2518 | 0 | fmask |= dmask; |
2519 | 0 | *tmask |= dmask; |
2520 | | |
2521 | | /* mark this field as being completed */ |
2522 | 0 | field[i] = NULL; |
2523 | 0 | } |
2524 | 0 | } |
2525 | | |
2526 | | /* now pick up remaining numeric fields */ |
2527 | 0 | for (i = 0; i < nf; i++) |
2528 | 0 | { |
2529 | 0 | if (field[i] == NULL) |
2530 | 0 | continue; |
2531 | | |
2532 | 0 | if ((len = strlen(field[i])) <= 0) |
2533 | 0 | return DTERR_BAD_FORMAT; |
2534 | | |
2535 | 0 | dterr = DecodeNumber(len, field[i], haveTextMonth, fmask, |
2536 | 0 | &dmask, tm, |
2537 | 0 | &fsec, is2digits); |
2538 | 0 | if (dterr) |
2539 | 0 | return dterr; |
2540 | | |
2541 | 0 | if (fmask & dmask) |
2542 | 0 | return DTERR_BAD_FORMAT; |
2543 | | |
2544 | 0 | fmask |= dmask; |
2545 | 0 | *tmask |= dmask; |
2546 | 0 | } |
2547 | | |
2548 | 0 | if ((fmask & ~(DTK_M(DOY) | DTK_M(TZ))) != DTK_DATE_M) |
2549 | 0 | return DTERR_BAD_FORMAT; |
2550 | | |
2551 | | /* validation of the field values must wait until ValidateDate() */ |
2552 | | |
2553 | 0 | return 0; |
2554 | 0 | } |
2555 | | |
2556 | | /* ValidateDate() |
2557 | | * Check valid year/month/day values, handle BC and DOY cases |
2558 | | * Return 0 if okay, a DTERR code if not. |
2559 | | */ |
2560 | | int |
2561 | | ValidateDate(int fmask, bool isjulian, bool is2digits, bool bc, |
2562 | | struct pg_tm *tm) |
2563 | 0 | { |
2564 | 0 | if (fmask & DTK_M(YEAR)) |
2565 | 0 | { |
2566 | 0 | if (isjulian) |
2567 | 0 | { |
2568 | | /* tm_year is correct and should not be touched */ |
2569 | 0 | } |
2570 | 0 | else if (bc) |
2571 | 0 | { |
2572 | | /* there is no year zero in AD/BC notation */ |
2573 | 0 | if (tm->tm_year <= 0) |
2574 | 0 | return DTERR_FIELD_OVERFLOW; |
2575 | | /* internally, we represent 1 BC as year zero, 2 BC as -1, etc */ |
2576 | 0 | tm->tm_year = -(tm->tm_year - 1); |
2577 | 0 | } |
2578 | 0 | else if (is2digits) |
2579 | 0 | { |
2580 | | /* process 1 or 2-digit input as 1970-2069 AD, allow '0' and '00' */ |
2581 | 0 | if (tm->tm_year < 0) /* just paranoia */ |
2582 | 0 | return DTERR_FIELD_OVERFLOW; |
2583 | 0 | if (tm->tm_year < 70) |
2584 | 0 | tm->tm_year += 2000; |
2585 | 0 | else if (tm->tm_year < 100) |
2586 | 0 | tm->tm_year += 1900; |
2587 | 0 | } |
2588 | 0 | else |
2589 | 0 | { |
2590 | | /* there is no year zero in AD/BC notation */ |
2591 | 0 | if (tm->tm_year <= 0) |
2592 | 0 | return DTERR_FIELD_OVERFLOW; |
2593 | 0 | } |
2594 | 0 | } |
2595 | | |
2596 | | /* now that we have correct year, decode DOY */ |
2597 | 0 | if (fmask & DTK_M(DOY)) |
2598 | 0 | { |
2599 | 0 | j2date(date2j(tm->tm_year, 1, 1) + tm->tm_yday - 1, |
2600 | 0 | &tm->tm_year, &tm->tm_mon, &tm->tm_mday); |
2601 | 0 | } |
2602 | | |
2603 | | /* check for valid month */ |
2604 | 0 | if (fmask & DTK_M(MONTH)) |
2605 | 0 | { |
2606 | 0 | if (tm->tm_mon < 1 || tm->tm_mon > MONTHS_PER_YEAR) |
2607 | 0 | return DTERR_MD_FIELD_OVERFLOW; |
2608 | 0 | } |
2609 | | |
2610 | | /* minimal check for valid day */ |
2611 | 0 | if (fmask & DTK_M(DAY)) |
2612 | 0 | { |
2613 | 0 | if (tm->tm_mday < 1 || tm->tm_mday > 31) |
2614 | 0 | return DTERR_MD_FIELD_OVERFLOW; |
2615 | 0 | } |
2616 | | |
2617 | 0 | if ((fmask & DTK_DATE_M) == DTK_DATE_M) |
2618 | 0 | { |
2619 | | /* |
2620 | | * Check for valid day of month, now that we know for sure the month |
2621 | | * and year. Note we don't use MD_FIELD_OVERFLOW here, since it seems |
2622 | | * unlikely that "Feb 29" is a YMD-order error. |
2623 | | */ |
2624 | 0 | if (tm->tm_mday > day_tab[isleap(tm->tm_year)][tm->tm_mon - 1]) |
2625 | 0 | return DTERR_FIELD_OVERFLOW; |
2626 | 0 | } |
2627 | | |
2628 | 0 | return 0; |
2629 | 0 | } |
2630 | | |
2631 | | |
2632 | | /* DecodeTimeCommon() |
2633 | | * Decode time string which includes delimiters. |
2634 | | * Return 0 if okay, a DTERR code if not. |
2635 | | * tmask and itm are output parameters. |
2636 | | * |
2637 | | * This code is shared between the timestamp and interval cases. |
2638 | | * We return a struct pg_itm (of which only the tm_usec, tm_sec, tm_min, |
2639 | | * and tm_hour fields are used) and let the wrapper functions below |
2640 | | * convert and range-check as necessary. |
2641 | | */ |
2642 | | static int |
2643 | | DecodeTimeCommon(char *str, int fmask, int range, |
2644 | | int *tmask, struct pg_itm *itm) |
2645 | 0 | { |
2646 | 0 | char *cp; |
2647 | 0 | int dterr; |
2648 | 0 | fsec_t fsec = 0; |
2649 | |
|
2650 | 0 | *tmask = DTK_TIME_M; |
2651 | |
|
2652 | 0 | errno = 0; |
2653 | 0 | itm->tm_hour = strtoi64(str, &cp, 10); |
2654 | 0 | if (errno == ERANGE) |
2655 | 0 | return DTERR_FIELD_OVERFLOW; |
2656 | 0 | if (*cp != ':') |
2657 | 0 | return DTERR_BAD_FORMAT; |
2658 | 0 | errno = 0; |
2659 | 0 | itm->tm_min = strtoint(cp + 1, &cp, 10); |
2660 | 0 | if (errno == ERANGE) |
2661 | 0 | return DTERR_FIELD_OVERFLOW; |
2662 | 0 | if (*cp == '\0') |
2663 | 0 | { |
2664 | 0 | itm->tm_sec = 0; |
2665 | | /* If it's a MINUTE TO SECOND interval, take 2 fields as being mm:ss */ |
2666 | 0 | if (range == (INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND))) |
2667 | 0 | { |
2668 | 0 | if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN) |
2669 | 0 | return DTERR_FIELD_OVERFLOW; |
2670 | 0 | itm->tm_sec = itm->tm_min; |
2671 | 0 | itm->tm_min = (int) itm->tm_hour; |
2672 | 0 | itm->tm_hour = 0; |
2673 | 0 | } |
2674 | 0 | } |
2675 | 0 | else if (*cp == '.') |
2676 | 0 | { |
2677 | | /* always assume mm:ss.sss is MINUTE TO SECOND */ |
2678 | 0 | dterr = ParseFractionalSecond(cp, &fsec); |
2679 | 0 | if (dterr) |
2680 | 0 | return dterr; |
2681 | 0 | if (itm->tm_hour > INT_MAX || itm->tm_hour < INT_MIN) |
2682 | 0 | return DTERR_FIELD_OVERFLOW; |
2683 | 0 | itm->tm_sec = itm->tm_min; |
2684 | 0 | itm->tm_min = (int) itm->tm_hour; |
2685 | 0 | itm->tm_hour = 0; |
2686 | 0 | } |
2687 | 0 | else if (*cp == ':') |
2688 | 0 | { |
2689 | 0 | errno = 0; |
2690 | 0 | itm->tm_sec = strtoint(cp + 1, &cp, 10); |
2691 | 0 | if (errno == ERANGE) |
2692 | 0 | return DTERR_FIELD_OVERFLOW; |
2693 | 0 | if (*cp == '.') |
2694 | 0 | { |
2695 | 0 | dterr = ParseFractionalSecond(cp, &fsec); |
2696 | 0 | if (dterr) |
2697 | 0 | return dterr; |
2698 | 0 | } |
2699 | 0 | else if (*cp != '\0') |
2700 | 0 | return DTERR_BAD_FORMAT; |
2701 | 0 | } |
2702 | 0 | else |
2703 | 0 | return DTERR_BAD_FORMAT; |
2704 | | |
2705 | | /* do a sanity check; but caller must check the range of tm_hour */ |
2706 | 0 | if (itm->tm_hour < 0 || |
2707 | 0 | itm->tm_min < 0 || itm->tm_min > MINS_PER_HOUR - 1 || |
2708 | 0 | itm->tm_sec < 0 || itm->tm_sec > SECS_PER_MINUTE || |
2709 | 0 | fsec < 0 || fsec > USECS_PER_SEC) |
2710 | 0 | return DTERR_FIELD_OVERFLOW; |
2711 | | |
2712 | 0 | itm->tm_usec = (int) fsec; |
2713 | |
|
2714 | 0 | return 0; |
2715 | 0 | } |
2716 | | |
2717 | | /* DecodeTime() |
2718 | | * Decode time string which includes delimiters. |
2719 | | * Return 0 if okay, a DTERR code if not. |
2720 | | * |
2721 | | * This version is used for timestamps. The results are returned into |
2722 | | * the tm_hour/tm_min/tm_sec fields of *tm, and microseconds into *fsec. |
2723 | | */ |
2724 | | static int |
2725 | | DecodeTime(char *str, int fmask, int range, |
2726 | | int *tmask, struct pg_tm *tm, fsec_t *fsec) |
2727 | 0 | { |
2728 | 0 | struct pg_itm itm; |
2729 | 0 | int dterr; |
2730 | |
|
2731 | 0 | dterr = DecodeTimeCommon(str, fmask, range, |
2732 | 0 | tmask, &itm); |
2733 | 0 | if (dterr) |
2734 | 0 | return dterr; |
2735 | | |
2736 | 0 | if (itm.tm_hour > INT_MAX) |
2737 | 0 | return DTERR_FIELD_OVERFLOW; |
2738 | 0 | tm->tm_hour = (int) itm.tm_hour; |
2739 | 0 | tm->tm_min = itm.tm_min; |
2740 | 0 | tm->tm_sec = itm.tm_sec; |
2741 | 0 | *fsec = itm.tm_usec; |
2742 | |
|
2743 | 0 | return 0; |
2744 | 0 | } |
2745 | | |
2746 | | /* DecodeTimeForInterval() |
2747 | | * Decode time string which includes delimiters. |
2748 | | * Return 0 if okay, a DTERR code if not. |
2749 | | * |
2750 | | * This version is used for intervals. The results are returned into |
2751 | | * itm_in->tm_usec. |
2752 | | */ |
2753 | | static int |
2754 | | DecodeTimeForInterval(char *str, int fmask, int range, |
2755 | | int *tmask, struct pg_itm_in *itm_in) |
2756 | 0 | { |
2757 | 0 | struct pg_itm itm; |
2758 | 0 | int dterr; |
2759 | |
|
2760 | 0 | dterr = DecodeTimeCommon(str, fmask, range, |
2761 | 0 | tmask, &itm); |
2762 | 0 | if (dterr) |
2763 | 0 | return dterr; |
2764 | | |
2765 | 0 | itm_in->tm_usec = itm.tm_usec; |
2766 | 0 | if (!int64_multiply_add(itm.tm_hour, USECS_PER_HOUR, &itm_in->tm_usec) || |
2767 | 0 | !int64_multiply_add(itm.tm_min, USECS_PER_MINUTE, &itm_in->tm_usec) || |
2768 | 0 | !int64_multiply_add(itm.tm_sec, USECS_PER_SEC, &itm_in->tm_usec)) |
2769 | 0 | return DTERR_FIELD_OVERFLOW; |
2770 | | |
2771 | 0 | return 0; |
2772 | 0 | } |
2773 | | |
2774 | | |
2775 | | /* DecodeNumber() |
2776 | | * Interpret plain numeric field as a date value in context. |
2777 | | * Return 0 if okay, a DTERR code if not. |
2778 | | */ |
2779 | | static int |
2780 | | DecodeNumber(int flen, char *str, bool haveTextMonth, int fmask, |
2781 | | int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits) |
2782 | 0 | { |
2783 | 0 | int val; |
2784 | 0 | char *cp; |
2785 | 0 | int dterr; |
2786 | |
|
2787 | 0 | *tmask = 0; |
2788 | |
|
2789 | 0 | errno = 0; |
2790 | 0 | val = strtoint(str, &cp, 10); |
2791 | 0 | if (errno == ERANGE) |
2792 | 0 | return DTERR_FIELD_OVERFLOW; |
2793 | 0 | if (cp == str) |
2794 | 0 | return DTERR_BAD_FORMAT; |
2795 | | |
2796 | 0 | if (*cp == '.') |
2797 | 0 | { |
2798 | | /* |
2799 | | * More than two digits before decimal point? Then could be a date or |
2800 | | * a run-together time: 2001.360 20011225 040506.789 |
2801 | | */ |
2802 | 0 | if (cp - str > 2) |
2803 | 0 | { |
2804 | 0 | dterr = DecodeNumberField(flen, str, |
2805 | 0 | (fmask | DTK_DATE_M), |
2806 | 0 | tmask, tm, |
2807 | 0 | fsec, is2digits); |
2808 | 0 | if (dterr < 0) |
2809 | 0 | return dterr; |
2810 | 0 | return 0; |
2811 | 0 | } |
2812 | | |
2813 | 0 | dterr = ParseFractionalSecond(cp, fsec); |
2814 | 0 | if (dterr) |
2815 | 0 | return dterr; |
2816 | 0 | } |
2817 | 0 | else if (*cp != '\0') |
2818 | 0 | return DTERR_BAD_FORMAT; |
2819 | | |
2820 | | /* Special case for day of year */ |
2821 | 0 | if (flen == 3 && (fmask & DTK_DATE_M) == DTK_M(YEAR) && val >= 1 && |
2822 | 0 | val <= 366) |
2823 | 0 | { |
2824 | 0 | *tmask = (DTK_M(DOY) | DTK_M(MONTH) | DTK_M(DAY)); |
2825 | 0 | tm->tm_yday = val; |
2826 | | /* tm_mon and tm_mday can't actually be set yet ... */ |
2827 | 0 | return 0; |
2828 | 0 | } |
2829 | | |
2830 | | /* Switch based on what we have so far */ |
2831 | 0 | switch (fmask & DTK_DATE_M) |
2832 | 0 | { |
2833 | 0 | case 0: |
2834 | | |
2835 | | /* |
2836 | | * Nothing so far; make a decision about what we think the input |
2837 | | * is. There used to be lots of heuristics here, but the |
2838 | | * consensus now is to be paranoid. It *must* be either |
2839 | | * YYYY-MM-DD (with a more-than-two-digit year field), or the |
2840 | | * field order defined by DateOrder. |
2841 | | */ |
2842 | 0 | if (flen >= 3 || DateOrder == DATEORDER_YMD) |
2843 | 0 | { |
2844 | 0 | *tmask = DTK_M(YEAR); |
2845 | 0 | tm->tm_year = val; |
2846 | 0 | } |
2847 | 0 | else if (DateOrder == DATEORDER_DMY) |
2848 | 0 | { |
2849 | 0 | *tmask = DTK_M(DAY); |
2850 | 0 | tm->tm_mday = val; |
2851 | 0 | } |
2852 | 0 | else |
2853 | 0 | { |
2854 | 0 | *tmask = DTK_M(MONTH); |
2855 | 0 | tm->tm_mon = val; |
2856 | 0 | } |
2857 | 0 | break; |
2858 | | |
2859 | 0 | case (DTK_M(YEAR)): |
2860 | | /* Must be at second field of YY-MM-DD */ |
2861 | 0 | *tmask = DTK_M(MONTH); |
2862 | 0 | tm->tm_mon = val; |
2863 | 0 | break; |
2864 | | |
2865 | 0 | case (DTK_M(MONTH)): |
2866 | 0 | if (haveTextMonth) |
2867 | 0 | { |
2868 | | /* |
2869 | | * We are at the first numeric field of a date that included a |
2870 | | * textual month name. We want to support the variants |
2871 | | * MON-DD-YYYY, DD-MON-YYYY, and YYYY-MON-DD as unambiguous |
2872 | | * inputs. We will also accept MON-DD-YY or DD-MON-YY in |
2873 | | * either DMY or MDY modes, as well as YY-MON-DD in YMD mode. |
2874 | | */ |
2875 | 0 | if (flen >= 3 || DateOrder == DATEORDER_YMD) |
2876 | 0 | { |
2877 | 0 | *tmask = DTK_M(YEAR); |
2878 | 0 | tm->tm_year = val; |
2879 | 0 | } |
2880 | 0 | else |
2881 | 0 | { |
2882 | 0 | *tmask = DTK_M(DAY); |
2883 | 0 | tm->tm_mday = val; |
2884 | 0 | } |
2885 | 0 | } |
2886 | 0 | else |
2887 | 0 | { |
2888 | | /* Must be at second field of MM-DD-YY */ |
2889 | 0 | *tmask = DTK_M(DAY); |
2890 | 0 | tm->tm_mday = val; |
2891 | 0 | } |
2892 | 0 | break; |
2893 | | |
2894 | 0 | case (DTK_M(YEAR) | DTK_M(MONTH)): |
2895 | 0 | if (haveTextMonth) |
2896 | 0 | { |
2897 | | /* Need to accept DD-MON-YYYY even in YMD mode */ |
2898 | 0 | if (flen >= 3 && *is2digits) |
2899 | 0 | { |
2900 | | /* Guess that first numeric field is day was wrong */ |
2901 | 0 | *tmask = DTK_M(DAY); /* YEAR is already set */ |
2902 | 0 | tm->tm_mday = tm->tm_year; |
2903 | 0 | tm->tm_year = val; |
2904 | 0 | *is2digits = false; |
2905 | 0 | } |
2906 | 0 | else |
2907 | 0 | { |
2908 | 0 | *tmask = DTK_M(DAY); |
2909 | 0 | tm->tm_mday = val; |
2910 | 0 | } |
2911 | 0 | } |
2912 | 0 | else |
2913 | 0 | { |
2914 | | /* Must be at third field of YY-MM-DD */ |
2915 | 0 | *tmask = DTK_M(DAY); |
2916 | 0 | tm->tm_mday = val; |
2917 | 0 | } |
2918 | 0 | break; |
2919 | | |
2920 | 0 | case (DTK_M(DAY)): |
2921 | | /* Must be at second field of DD-MM-YY */ |
2922 | 0 | *tmask = DTK_M(MONTH); |
2923 | 0 | tm->tm_mon = val; |
2924 | 0 | break; |
2925 | | |
2926 | 0 | case (DTK_M(MONTH) | DTK_M(DAY)): |
2927 | | /* Must be at third field of DD-MM-YY or MM-DD-YY */ |
2928 | 0 | *tmask = DTK_M(YEAR); |
2929 | 0 | tm->tm_year = val; |
2930 | 0 | break; |
2931 | | |
2932 | 0 | case (DTK_M(YEAR) | DTK_M(MONTH) | DTK_M(DAY)): |
2933 | | /* we have all the date, so it must be a time field */ |
2934 | 0 | dterr = DecodeNumberField(flen, str, fmask, |
2935 | 0 | tmask, tm, |
2936 | 0 | fsec, is2digits); |
2937 | 0 | if (dterr < 0) |
2938 | 0 | return dterr; |
2939 | 0 | return 0; |
2940 | | |
2941 | 0 | default: |
2942 | | /* Anything else is bogus input */ |
2943 | 0 | return DTERR_BAD_FORMAT; |
2944 | 0 | } |
2945 | | |
2946 | | /* |
2947 | | * When processing a year field, mark it for adjustment if it's only one |
2948 | | * or two digits. |
2949 | | */ |
2950 | 0 | if (*tmask == DTK_M(YEAR)) |
2951 | 0 | *is2digits = (flen <= 2); |
2952 | |
|
2953 | 0 | return 0; |
2954 | 0 | } |
2955 | | |
2956 | | |
2957 | | /* DecodeNumberField() |
2958 | | * Interpret numeric string as a concatenated date or time field. |
2959 | | * Return a DTK token (>= 0) if successful, a DTERR code (< 0) if not. |
2960 | | * |
2961 | | * Use the context of previously decoded fields to help with |
2962 | | * the interpretation. |
2963 | | */ |
2964 | | static int |
2965 | | DecodeNumberField(int len, char *str, int fmask, |
2966 | | int *tmask, struct pg_tm *tm, fsec_t *fsec, bool *is2digits) |
2967 | 0 | { |
2968 | 0 | char *cp; |
2969 | | |
2970 | | /* |
2971 | | * This function was originally meant to cope only with DTK_NUMBER fields, |
2972 | | * but we now sometimes abuse it to parse (parts of) DTK_DATE fields, |
2973 | | * which can contain letters and other punctuation. Reject if it's not a |
2974 | | * valid DTK_NUMBER, that is digits and decimal point(s). (ParseFraction |
2975 | | * will reject if there's more than one decimal point.) |
2976 | | */ |
2977 | 0 | if (strspn(str, "0123456789.") != len) |
2978 | 0 | return DTERR_BAD_FORMAT; |
2979 | | |
2980 | | /* |
2981 | | * Have a decimal point? Then this is a date or something with a seconds |
2982 | | * field... |
2983 | | */ |
2984 | 0 | if ((cp = strchr(str, '.')) != NULL) |
2985 | 0 | { |
2986 | 0 | int dterr; |
2987 | | |
2988 | | /* Convert the fraction and store at *fsec */ |
2989 | 0 | dterr = ParseFractionalSecond(cp, fsec); |
2990 | 0 | if (dterr) |
2991 | 0 | return dterr; |
2992 | | /* Now truncate off the fraction for further processing */ |
2993 | 0 | *cp = '\0'; |
2994 | 0 | len = strlen(str); |
2995 | 0 | } |
2996 | | /* No decimal point and no complete date yet? */ |
2997 | 0 | else if ((fmask & DTK_DATE_M) != DTK_DATE_M) |
2998 | 0 | { |
2999 | 0 | if (len >= 6) |
3000 | 0 | { |
3001 | 0 | *tmask = DTK_DATE_M; |
3002 | | |
3003 | | /* |
3004 | | * Start from end and consider first 2 as Day, next 2 as Month, |
3005 | | * and the rest as Year. |
3006 | | */ |
3007 | 0 | tm->tm_mday = atoi(str + (len - 2)); |
3008 | 0 | *(str + (len - 2)) = '\0'; |
3009 | 0 | tm->tm_mon = atoi(str + (len - 4)); |
3010 | 0 | *(str + (len - 4)) = '\0'; |
3011 | 0 | tm->tm_year = atoi(str); |
3012 | 0 | if ((len - 4) == 2) |
3013 | 0 | *is2digits = true; |
3014 | |
|
3015 | 0 | return DTK_DATE; |
3016 | 0 | } |
3017 | 0 | } |
3018 | | |
3019 | | /* not all time fields are specified? */ |
3020 | 0 | if ((fmask & DTK_TIME_M) != DTK_TIME_M) |
3021 | 0 | { |
3022 | | /* hhmmss */ |
3023 | 0 | if (len == 6) |
3024 | 0 | { |
3025 | 0 | *tmask = DTK_TIME_M; |
3026 | 0 | tm->tm_sec = atoi(str + 4); |
3027 | 0 | *(str + 4) = '\0'; |
3028 | 0 | tm->tm_min = atoi(str + 2); |
3029 | 0 | *(str + 2) = '\0'; |
3030 | 0 | tm->tm_hour = atoi(str); |
3031 | |
|
3032 | 0 | return DTK_TIME; |
3033 | 0 | } |
3034 | | /* hhmm? */ |
3035 | 0 | else if (len == 4) |
3036 | 0 | { |
3037 | 0 | *tmask = DTK_TIME_M; |
3038 | 0 | tm->tm_sec = 0; |
3039 | 0 | tm->tm_min = atoi(str + 2); |
3040 | 0 | *(str + 2) = '\0'; |
3041 | 0 | tm->tm_hour = atoi(str); |
3042 | |
|
3043 | 0 | return DTK_TIME; |
3044 | 0 | } |
3045 | 0 | } |
3046 | | |
3047 | 0 | return DTERR_BAD_FORMAT; |
3048 | 0 | } |
3049 | | |
3050 | | |
3051 | | /* DecodeTimezone() |
3052 | | * Interpret string as a numeric timezone. |
3053 | | * |
3054 | | * Return 0 if okay (and set *tzp), a DTERR code if not okay. |
3055 | | */ |
3056 | | int |
3057 | | DecodeTimezone(const char *str, int *tzp) |
3058 | 0 | { |
3059 | 0 | int tz; |
3060 | 0 | int hr, |
3061 | 0 | min, |
3062 | 0 | sec = 0; |
3063 | 0 | char *cp; |
3064 | | |
3065 | | /* leading character must be "+" or "-" */ |
3066 | 0 | if (*str != '+' && *str != '-') |
3067 | 0 | return DTERR_BAD_FORMAT; |
3068 | | |
3069 | 0 | errno = 0; |
3070 | 0 | hr = strtoint(str + 1, &cp, 10); |
3071 | 0 | if (errno == ERANGE) |
3072 | 0 | return DTERR_TZDISP_OVERFLOW; |
3073 | | |
3074 | | /* explicit delimiter? */ |
3075 | 0 | if (*cp == ':') |
3076 | 0 | { |
3077 | 0 | errno = 0; |
3078 | 0 | min = strtoint(cp + 1, &cp, 10); |
3079 | 0 | if (errno == ERANGE) |
3080 | 0 | return DTERR_TZDISP_OVERFLOW; |
3081 | 0 | if (*cp == ':') |
3082 | 0 | { |
3083 | 0 | errno = 0; |
3084 | 0 | sec = strtoint(cp + 1, &cp, 10); |
3085 | 0 | if (errno == ERANGE) |
3086 | 0 | return DTERR_TZDISP_OVERFLOW; |
3087 | 0 | } |
3088 | 0 | } |
3089 | | /* otherwise, might have run things together... */ |
3090 | 0 | else if (*cp == '\0' && strlen(str) > 3) |
3091 | 0 | { |
3092 | 0 | min = hr % 100; |
3093 | 0 | hr = hr / 100; |
3094 | | /* we could, but don't, support a run-together hhmmss format */ |
3095 | 0 | } |
3096 | 0 | else |
3097 | 0 | min = 0; |
3098 | | |
3099 | | /* Range-check the values; see notes in datatype/timestamp.h */ |
3100 | 0 | if (hr < 0 || hr > MAX_TZDISP_HOUR) |
3101 | 0 | return DTERR_TZDISP_OVERFLOW; |
3102 | 0 | if (min < 0 || min >= MINS_PER_HOUR) |
3103 | 0 | return DTERR_TZDISP_OVERFLOW; |
3104 | 0 | if (sec < 0 || sec >= SECS_PER_MINUTE) |
3105 | 0 | return DTERR_TZDISP_OVERFLOW; |
3106 | | |
3107 | 0 | tz = (hr * MINS_PER_HOUR + min) * SECS_PER_MINUTE + sec; |
3108 | 0 | if (*str == '-') |
3109 | 0 | tz = -tz; |
3110 | |
|
3111 | 0 | *tzp = -tz; |
3112 | |
|
3113 | 0 | if (*cp != '\0') |
3114 | 0 | return DTERR_BAD_FORMAT; |
3115 | | |
3116 | 0 | return 0; |
3117 | 0 | } |
3118 | | |
3119 | | |
3120 | | /* DecodeTimezoneAbbrev() |
3121 | | * Interpret string as a timezone abbreviation, if possible. |
3122 | | * |
3123 | | * Sets *ftype to an abbreviation type (TZ, DTZ, or DYNTZ), or UNKNOWN_FIELD if |
3124 | | * string is not any known abbreviation. On success, set *offset and *tz to |
3125 | | * represent the UTC offset (for TZ or DTZ) or underlying zone (for DYNTZ). |
3126 | | * Note that full timezone names (such as America/New_York) are not handled |
3127 | | * here, mostly for historical reasons. |
3128 | | * |
3129 | | * The function result is 0 or a DTERR code; in the latter case, *extra |
3130 | | * is filled as needed. Note that unknown-abbreviation is not considered |
3131 | | * an error case. Also note that many callers assume that the DTERR code |
3132 | | * is one that DateTimeParseError does not require "str" or "datatype" |
3133 | | * strings for. |
3134 | | * |
3135 | | * Given string must be lowercased already. |
3136 | | * |
3137 | | * Implement a cache lookup since it is likely that dates |
3138 | | * will be related in format. |
3139 | | */ |
3140 | | int |
3141 | | DecodeTimezoneAbbrev(int field, const char *lowtoken, |
3142 | | int *ftype, int *offset, pg_tz **tz, |
3143 | | DateTimeErrorExtra *extra) |
3144 | 0 | { |
3145 | 0 | TzAbbrevCache *tzc = &tzabbrevcache[field]; |
3146 | 0 | bool isfixed; |
3147 | 0 | int isdst; |
3148 | 0 | const datetkn *tp; |
3149 | | |
3150 | | /* |
3151 | | * Do we have a cached result? Use strncmp so that we match truncated |
3152 | | * names, although we shouldn't really see that happen with normal |
3153 | | * abbreviations. |
3154 | | */ |
3155 | 0 | if (strncmp(lowtoken, tzc->abbrev, TOKMAXLEN) == 0) |
3156 | 0 | { |
3157 | 0 | *ftype = tzc->ftype; |
3158 | 0 | *offset = tzc->offset; |
3159 | 0 | *tz = tzc->tz; |
3160 | 0 | return 0; |
3161 | 0 | } |
3162 | | |
3163 | | /* |
3164 | | * See if the current session_timezone recognizes it. Checking this |
3165 | | * before zoneabbrevtbl allows us to correctly handle abbreviations whose |
3166 | | * meaning varies across zones, such as "LMT". |
3167 | | */ |
3168 | 0 | if (session_timezone && |
3169 | 0 | TimeZoneAbbrevIsKnown(lowtoken, session_timezone, |
3170 | 0 | &isfixed, offset, &isdst)) |
3171 | 0 | { |
3172 | 0 | *ftype = (isfixed ? (isdst ? DTZ : TZ) : DYNTZ); |
3173 | 0 | *tz = (isfixed ? NULL : session_timezone); |
3174 | | /* flip sign to agree with the convention used in zoneabbrevtbl */ |
3175 | 0 | *offset = -(*offset); |
3176 | | /* cache result; use strlcpy to truncate name if necessary */ |
3177 | 0 | strlcpy(tzc->abbrev, lowtoken, TOKMAXLEN + 1); |
3178 | 0 | tzc->ftype = *ftype; |
3179 | 0 | tzc->offset = *offset; |
3180 | 0 | tzc->tz = *tz; |
3181 | 0 | return 0; |
3182 | 0 | } |
3183 | | |
3184 | | /* Nope, so look in zoneabbrevtbl */ |
3185 | 0 | if (zoneabbrevtbl) |
3186 | 0 | tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs, |
3187 | 0 | zoneabbrevtbl->numabbrevs); |
3188 | 0 | else |
3189 | 0 | tp = NULL; |
3190 | 0 | if (tp == NULL) |
3191 | 0 | { |
3192 | 0 | *ftype = UNKNOWN_FIELD; |
3193 | 0 | *offset = 0; |
3194 | 0 | *tz = NULL; |
3195 | | /* failure results are not cached */ |
3196 | 0 | } |
3197 | 0 | else |
3198 | 0 | { |
3199 | 0 | *ftype = tp->type; |
3200 | 0 | if (tp->type == DYNTZ) |
3201 | 0 | { |
3202 | 0 | *offset = 0; |
3203 | 0 | *tz = FetchDynamicTimeZone(zoneabbrevtbl, tp, extra); |
3204 | 0 | if (*tz == NULL) |
3205 | 0 | return DTERR_BAD_ZONE_ABBREV; |
3206 | 0 | } |
3207 | 0 | else |
3208 | 0 | { |
3209 | 0 | *offset = tp->value; |
3210 | 0 | *tz = NULL; |
3211 | 0 | } |
3212 | | |
3213 | | /* cache result; use strlcpy to truncate name if necessary */ |
3214 | 0 | strlcpy(tzc->abbrev, lowtoken, TOKMAXLEN + 1); |
3215 | 0 | tzc->ftype = *ftype; |
3216 | 0 | tzc->offset = *offset; |
3217 | 0 | tzc->tz = *tz; |
3218 | 0 | } |
3219 | | |
3220 | 0 | return 0; |
3221 | 0 | } |
3222 | | |
3223 | | /* |
3224 | | * Reset tzabbrevcache after a change in session_timezone. |
3225 | | */ |
3226 | | void |
3227 | | ClearTimeZoneAbbrevCache(void) |
3228 | 2 | { |
3229 | 2 | memset(tzabbrevcache, 0, sizeof(tzabbrevcache)); |
3230 | 2 | } |
3231 | | |
3232 | | |
3233 | | /* DecodeSpecial() |
3234 | | * Decode text string using lookup table. |
3235 | | * |
3236 | | * Recognizes the keywords listed in datetktbl. |
3237 | | * Note: at one time this would also recognize timezone abbreviations, |
3238 | | * but no more; use DecodeTimezoneAbbrev for that. |
3239 | | * |
3240 | | * Given string must be lowercased already. |
3241 | | * |
3242 | | * Implement a cache lookup since it is likely that dates |
3243 | | * will be related in format. |
3244 | | */ |
3245 | | int |
3246 | | DecodeSpecial(int field, const char *lowtoken, int *val) |
3247 | 0 | { |
3248 | 0 | int type; |
3249 | 0 | const datetkn *tp; |
3250 | |
|
3251 | 0 | tp = datecache[field]; |
3252 | | /* use strncmp so that we match truncated tokens */ |
3253 | 0 | if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0) |
3254 | 0 | { |
3255 | 0 | tp = datebsearch(lowtoken, datetktbl, szdatetktbl); |
3256 | 0 | } |
3257 | 0 | if (tp == NULL) |
3258 | 0 | { |
3259 | 0 | type = UNKNOWN_FIELD; |
3260 | 0 | *val = 0; |
3261 | 0 | } |
3262 | 0 | else |
3263 | 0 | { |
3264 | 0 | datecache[field] = tp; |
3265 | 0 | type = tp->type; |
3266 | 0 | *val = tp->value; |
3267 | 0 | } |
3268 | |
|
3269 | 0 | return type; |
3270 | 0 | } |
3271 | | |
3272 | | |
3273 | | /* DecodeTimezoneName() |
3274 | | * Interpret string as a timezone abbreviation or name. |
3275 | | * Throw error if the name is not recognized. |
3276 | | * |
3277 | | * The return value indicates what kind of zone identifier it is: |
3278 | | * TZNAME_FIXED_OFFSET: fixed offset from UTC |
3279 | | * TZNAME_DYNTZ: dynamic timezone abbreviation |
3280 | | * TZNAME_ZONE: full tzdb zone name |
3281 | | * |
3282 | | * For TZNAME_FIXED_OFFSET, *offset receives the UTC offset (in seconds, |
3283 | | * with ISO sign convention: positive is east of Greenwich). |
3284 | | * For the other two cases, *tz receives the timezone struct representing |
3285 | | * the zone name or the abbreviation's underlying zone. |
3286 | | */ |
3287 | | int |
3288 | | DecodeTimezoneName(const char *tzname, int *offset, pg_tz **tz) |
3289 | 0 | { |
3290 | 0 | char *lowzone; |
3291 | 0 | int dterr, |
3292 | 0 | type; |
3293 | 0 | DateTimeErrorExtra extra; |
3294 | | |
3295 | | /* |
3296 | | * First we look in the timezone abbreviation table (to handle cases like |
3297 | | * "EST"), and if that fails, we look in the timezone database (to handle |
3298 | | * cases like "America/New_York"). This matches the order in which |
3299 | | * timestamp input checks the cases; it's important because the timezone |
3300 | | * database unwisely uses a few zone names that are identical to offset |
3301 | | * abbreviations. |
3302 | | */ |
3303 | | |
3304 | | /* DecodeTimezoneAbbrev requires lowercase input */ |
3305 | 0 | lowzone = downcase_truncate_identifier(tzname, |
3306 | 0 | strlen(tzname), |
3307 | 0 | false); |
3308 | |
|
3309 | 0 | dterr = DecodeTimezoneAbbrev(0, lowzone, &type, offset, tz, &extra); |
3310 | 0 | if (dterr) |
3311 | 0 | DateTimeParseError(dterr, &extra, NULL, NULL, NULL); |
3312 | |
|
3313 | 0 | if (type == TZ || type == DTZ) |
3314 | 0 | { |
3315 | | /* fixed-offset abbreviation, return the offset */ |
3316 | 0 | return TZNAME_FIXED_OFFSET; |
3317 | 0 | } |
3318 | 0 | else if (type == DYNTZ) |
3319 | 0 | { |
3320 | | /* dynamic-offset abbreviation, return its referenced timezone */ |
3321 | 0 | return TZNAME_DYNTZ; |
3322 | 0 | } |
3323 | 0 | else |
3324 | 0 | { |
3325 | | /* try it as a full zone name */ |
3326 | 0 | *tz = pg_tzset(tzname); |
3327 | 0 | if (*tz == NULL) |
3328 | 0 | ereport(ERROR, |
3329 | 0 | (errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
3330 | 0 | errmsg("time zone \"%s\" not recognized", tzname))); |
3331 | 0 | return TZNAME_ZONE; |
3332 | 0 | } |
3333 | 0 | } |
3334 | | |
3335 | | /* DecodeTimezoneNameToTz() |
3336 | | * Interpret string as a timezone abbreviation or name. |
3337 | | * Throw error if the name is not recognized. |
3338 | | * |
3339 | | * This is a simple wrapper for DecodeTimezoneName that produces a pg_tz * |
3340 | | * result in all cases. |
3341 | | */ |
3342 | | pg_tz * |
3343 | | DecodeTimezoneNameToTz(const char *tzname) |
3344 | 0 | { |
3345 | 0 | pg_tz *result; |
3346 | 0 | int offset; |
3347 | |
|
3348 | 0 | if (DecodeTimezoneName(tzname, &offset, &result) == TZNAME_FIXED_OFFSET) |
3349 | 0 | { |
3350 | | /* fixed-offset abbreviation, get a pg_tz descriptor for that */ |
3351 | 0 | result = pg_tzset_offset(-offset); /* flip to POSIX sign convention */ |
3352 | 0 | } |
3353 | 0 | return result; |
3354 | 0 | } |
3355 | | |
3356 | | /* DecodeTimezoneAbbrevPrefix() |
3357 | | * Interpret prefix of string as a timezone abbreviation, if possible. |
3358 | | * |
3359 | | * This has roughly the same functionality as DecodeTimezoneAbbrev(), |
3360 | | * but the API is adapted to the needs of formatting.c. Notably, |
3361 | | * we will match the longest possible prefix of the given string |
3362 | | * rather than insisting on a complete match, and downcasing is applied |
3363 | | * here rather than in the caller. |
3364 | | * |
3365 | | * Returns the length of the timezone abbreviation, or -1 if not recognized. |
3366 | | * On success, sets *offset to the GMT offset for the abbreviation if it |
3367 | | * is a fixed-offset abbreviation, or sets *tz to the pg_tz struct for |
3368 | | * a dynamic abbreviation. |
3369 | | */ |
3370 | | int |
3371 | | DecodeTimezoneAbbrevPrefix(const char *str, int *offset, pg_tz **tz) |
3372 | 0 | { |
3373 | 0 | char lowtoken[TOKMAXLEN + 1]; |
3374 | 0 | int len; |
3375 | |
|
3376 | 0 | *offset = 0; /* avoid uninitialized vars on failure */ |
3377 | 0 | *tz = NULL; |
3378 | | |
3379 | | /* Downcase as much of the string as we could need */ |
3380 | 0 | for (len = 0; len < TOKMAXLEN; len++) |
3381 | 0 | { |
3382 | 0 | if (*str == '\0' || !isalpha((unsigned char) *str)) |
3383 | 0 | break; |
3384 | 0 | lowtoken[len] = pg_tolower((unsigned char) *str++); |
3385 | 0 | } |
3386 | 0 | lowtoken[len] = '\0'; |
3387 | | |
3388 | | /* |
3389 | | * We could avoid doing repeated binary searches if we cared to duplicate |
3390 | | * datebsearch here, but it's not clear that such an optimization would be |
3391 | | * worth the trouble. In common cases there's probably not anything after |
3392 | | * the zone abbrev anyway. So just search with successively truncated |
3393 | | * strings. |
3394 | | */ |
3395 | 0 | while (len > 0) |
3396 | 0 | { |
3397 | 0 | bool isfixed; |
3398 | 0 | int isdst; |
3399 | 0 | const datetkn *tp; |
3400 | | |
3401 | | /* See if the current session_timezone recognizes it. */ |
3402 | 0 | if (session_timezone && |
3403 | 0 | TimeZoneAbbrevIsKnown(lowtoken, session_timezone, |
3404 | 0 | &isfixed, offset, &isdst)) |
3405 | 0 | { |
3406 | 0 | if (isfixed) |
3407 | 0 | { |
3408 | | /* flip sign to agree with the convention in zoneabbrevtbl */ |
3409 | 0 | *offset = -(*offset); |
3410 | 0 | } |
3411 | 0 | else |
3412 | 0 | { |
3413 | | /* Caller must resolve the abbrev's current meaning */ |
3414 | 0 | *tz = session_timezone; |
3415 | 0 | } |
3416 | 0 | return len; |
3417 | 0 | } |
3418 | | |
3419 | | /* Known in zoneabbrevtbl? */ |
3420 | 0 | if (zoneabbrevtbl) |
3421 | 0 | tp = datebsearch(lowtoken, zoneabbrevtbl->abbrevs, |
3422 | 0 | zoneabbrevtbl->numabbrevs); |
3423 | 0 | else |
3424 | 0 | tp = NULL; |
3425 | 0 | if (tp != NULL) |
3426 | 0 | { |
3427 | 0 | if (tp->type == DYNTZ) |
3428 | 0 | { |
3429 | 0 | DateTimeErrorExtra extra; |
3430 | 0 | pg_tz *tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp, |
3431 | 0 | &extra); |
3432 | |
|
3433 | 0 | if (tzp != NULL) |
3434 | 0 | { |
3435 | | /* Caller must resolve the abbrev's current meaning */ |
3436 | 0 | *tz = tzp; |
3437 | 0 | return len; |
3438 | 0 | } |
3439 | 0 | } |
3440 | 0 | else |
3441 | 0 | { |
3442 | | /* Fixed-offset zone abbrev, so it's easy */ |
3443 | 0 | *offset = tp->value; |
3444 | 0 | return len; |
3445 | 0 | } |
3446 | 0 | } |
3447 | | |
3448 | | /* Nope, try the next shorter string. */ |
3449 | 0 | lowtoken[--len] = '\0'; |
3450 | 0 | } |
3451 | | |
3452 | | /* Did not find a match */ |
3453 | 0 | return -1; |
3454 | 0 | } |
3455 | | |
3456 | | |
3457 | | /* ClearPgItmIn |
3458 | | * |
3459 | | * Zero out a pg_itm_in |
3460 | | */ |
3461 | | static inline void |
3462 | | ClearPgItmIn(struct pg_itm_in *itm_in) |
3463 | 0 | { |
3464 | 0 | itm_in->tm_usec = 0; |
3465 | 0 | itm_in->tm_mday = 0; |
3466 | 0 | itm_in->tm_mon = 0; |
3467 | 0 | itm_in->tm_year = 0; |
3468 | 0 | } |
3469 | | |
3470 | | |
3471 | | /* DecodeInterval() |
3472 | | * Interpret previously parsed fields for general time interval. |
3473 | | * Returns 0 if successful, DTERR code if bogus input detected. |
3474 | | * dtype and itm_in are output parameters. |
3475 | | * |
3476 | | * Allow "date" field DTK_DATE since this could be just |
3477 | | * an unsigned floating point number. - thomas 1997-11-16 |
3478 | | * |
3479 | | * Allow ISO-style time span, with implicit units on number of days |
3480 | | * preceding an hh:mm:ss field. - thomas 1998-04-30 |
3481 | | * |
3482 | | * itm_in remains undefined for infinite interval values for which dtype alone |
3483 | | * suffices. |
3484 | | */ |
3485 | | int |
3486 | | DecodeInterval(char **field, int *ftype, int nf, int range, |
3487 | | int *dtype, struct pg_itm_in *itm_in) |
3488 | 0 | { |
3489 | 0 | bool force_negative = false; |
3490 | 0 | bool is_before = false; |
3491 | 0 | bool parsing_unit_val = false; |
3492 | 0 | char *cp; |
3493 | 0 | int fmask = 0, |
3494 | 0 | tmask, |
3495 | 0 | type, |
3496 | 0 | uval; |
3497 | 0 | int i; |
3498 | 0 | int dterr; |
3499 | 0 | int64 val; |
3500 | 0 | double fval; |
3501 | |
|
3502 | 0 | *dtype = DTK_DELTA; |
3503 | 0 | type = IGNORE_DTF; |
3504 | 0 | ClearPgItmIn(itm_in); |
3505 | | |
3506 | | /*---------- |
3507 | | * The SQL standard defines the interval literal |
3508 | | * '-1 1:00:00' |
3509 | | * to mean "negative 1 days and negative 1 hours", while Postgres |
3510 | | * traditionally treats this as meaning "negative 1 days and positive |
3511 | | * 1 hours". In SQL_STANDARD intervalstyle, we apply the leading sign |
3512 | | * to all fields if there are no other explicit signs. |
3513 | | * |
3514 | | * We leave the signs alone if there are additional explicit signs. |
3515 | | * This protects us against misinterpreting postgres-style dump output, |
3516 | | * since the postgres-style output code has always put an explicit sign on |
3517 | | * all fields following a negative field. But note that SQL-spec output |
3518 | | * is ambiguous and can be misinterpreted on load! (So it's best practice |
3519 | | * to dump in postgres style, not SQL style.) |
3520 | | *---------- |
3521 | | */ |
3522 | 0 | if (IntervalStyle == INTSTYLE_SQL_STANDARD && nf > 0 && *field[0] == '-') |
3523 | 0 | { |
3524 | 0 | force_negative = true; |
3525 | | /* Check for additional explicit signs */ |
3526 | 0 | for (i = 1; i < nf; i++) |
3527 | 0 | { |
3528 | 0 | if (*field[i] == '-' || *field[i] == '+') |
3529 | 0 | { |
3530 | 0 | force_negative = false; |
3531 | 0 | break; |
3532 | 0 | } |
3533 | 0 | } |
3534 | 0 | } |
3535 | | |
3536 | | /* read through list backwards to pick up units before values */ |
3537 | 0 | for (i = nf - 1; i >= 0; i--) |
3538 | 0 | { |
3539 | 0 | switch (ftype[i]) |
3540 | 0 | { |
3541 | 0 | case DTK_TIME: |
3542 | 0 | dterr = DecodeTimeForInterval(field[i], fmask, range, |
3543 | 0 | &tmask, itm_in); |
3544 | 0 | if (dterr) |
3545 | 0 | return dterr; |
3546 | 0 | if (force_negative && |
3547 | 0 | itm_in->tm_usec > 0) |
3548 | 0 | itm_in->tm_usec = -itm_in->tm_usec; |
3549 | 0 | type = DTK_DAY; |
3550 | 0 | parsing_unit_val = false; |
3551 | 0 | break; |
3552 | | |
3553 | 0 | case DTK_TZ: |
3554 | | |
3555 | | /* |
3556 | | * Timezone means a token with a leading sign character and at |
3557 | | * least one digit; there could be ':', '.', '-' embedded in |
3558 | | * it as well. |
3559 | | */ |
3560 | 0 | Assert(*field[i] == '-' || *field[i] == '+'); |
3561 | | |
3562 | | /* |
3563 | | * Check for signed hh:mm or hh:mm:ss. If so, process exactly |
3564 | | * like DTK_TIME case above, plus handling the sign. |
3565 | | */ |
3566 | 0 | if (strchr(field[i] + 1, ':') != NULL && |
3567 | 0 | DecodeTimeForInterval(field[i] + 1, fmask, range, |
3568 | 0 | &tmask, itm_in) == 0) |
3569 | 0 | { |
3570 | 0 | if (*field[i] == '-') |
3571 | 0 | { |
3572 | | /* flip the sign on time field */ |
3573 | 0 | if (itm_in->tm_usec == PG_INT64_MIN) |
3574 | 0 | return DTERR_FIELD_OVERFLOW; |
3575 | 0 | itm_in->tm_usec = -itm_in->tm_usec; |
3576 | 0 | } |
3577 | | |
3578 | 0 | if (force_negative && |
3579 | 0 | itm_in->tm_usec > 0) |
3580 | 0 | itm_in->tm_usec = -itm_in->tm_usec; |
3581 | | |
3582 | | /* |
3583 | | * Set the next type to be a day, if units are not |
3584 | | * specified. This handles the case of '1 +02:03' since we |
3585 | | * are reading right to left. |
3586 | | */ |
3587 | 0 | type = DTK_DAY; |
3588 | 0 | parsing_unit_val = false; |
3589 | 0 | break; |
3590 | 0 | } |
3591 | | |
3592 | | /* |
3593 | | * Otherwise, fall through to DTK_NUMBER case, which can |
3594 | | * handle signed float numbers and signed year-month values. |
3595 | | */ |
3596 | | |
3597 | | /* FALLTHROUGH */ |
3598 | | |
3599 | 0 | case DTK_DATE: |
3600 | 0 | case DTK_NUMBER: |
3601 | 0 | if (type == IGNORE_DTF) |
3602 | 0 | { |
3603 | | /* use typmod to decide what rightmost field is */ |
3604 | 0 | switch (range) |
3605 | 0 | { |
3606 | 0 | case INTERVAL_MASK(YEAR): |
3607 | 0 | type = DTK_YEAR; |
3608 | 0 | break; |
3609 | 0 | case INTERVAL_MASK(MONTH): |
3610 | 0 | case INTERVAL_MASK(YEAR) | INTERVAL_MASK(MONTH): |
3611 | 0 | type = DTK_MONTH; |
3612 | 0 | break; |
3613 | 0 | case INTERVAL_MASK(DAY): |
3614 | 0 | type = DTK_DAY; |
3615 | 0 | break; |
3616 | 0 | case INTERVAL_MASK(HOUR): |
3617 | 0 | case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR): |
3618 | 0 | type = DTK_HOUR; |
3619 | 0 | break; |
3620 | 0 | case INTERVAL_MASK(MINUTE): |
3621 | 0 | case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE): |
3622 | 0 | case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE): |
3623 | 0 | type = DTK_MINUTE; |
3624 | 0 | break; |
3625 | 0 | case INTERVAL_MASK(SECOND): |
3626 | 0 | case INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND): |
3627 | 0 | case INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND): |
3628 | 0 | case INTERVAL_MASK(DAY) | INTERVAL_MASK(HOUR) | INTERVAL_MASK(MINUTE) | INTERVAL_MASK(SECOND): |
3629 | 0 | type = DTK_SECOND; |
3630 | 0 | break; |
3631 | 0 | default: |
3632 | 0 | type = DTK_SECOND; |
3633 | 0 | break; |
3634 | 0 | } |
3635 | 0 | } |
3636 | | |
3637 | 0 | errno = 0; |
3638 | 0 | val = strtoi64(field[i], &cp, 10); |
3639 | 0 | if (errno == ERANGE) |
3640 | 0 | return DTERR_FIELD_OVERFLOW; |
3641 | | |
3642 | 0 | if (*cp == '-') |
3643 | 0 | { |
3644 | | /* SQL "years-months" syntax */ |
3645 | 0 | int val2; |
3646 | |
|
3647 | 0 | val2 = strtoint(cp + 1, &cp, 10); |
3648 | 0 | if (errno == ERANGE || val2 < 0 || val2 >= MONTHS_PER_YEAR) |
3649 | 0 | return DTERR_FIELD_OVERFLOW; |
3650 | 0 | if (*cp != '\0') |
3651 | 0 | return DTERR_BAD_FORMAT; |
3652 | 0 | type = DTK_MONTH; |
3653 | 0 | if (*field[i] == '-') |
3654 | 0 | val2 = -val2; |
3655 | 0 | if (pg_mul_s64_overflow(val, MONTHS_PER_YEAR, &val)) |
3656 | 0 | return DTERR_FIELD_OVERFLOW; |
3657 | 0 | if (pg_add_s64_overflow(val, val2, &val)) |
3658 | 0 | return DTERR_FIELD_OVERFLOW; |
3659 | 0 | fval = 0; |
3660 | 0 | } |
3661 | 0 | else if (*cp == '.') |
3662 | 0 | { |
3663 | 0 | dterr = ParseFraction(cp, &fval); |
3664 | 0 | if (dterr) |
3665 | 0 | return dterr; |
3666 | 0 | if (*field[i] == '-') |
3667 | 0 | fval = -fval; |
3668 | 0 | } |
3669 | 0 | else if (*cp == '\0') |
3670 | 0 | fval = 0; |
3671 | 0 | else |
3672 | 0 | return DTERR_BAD_FORMAT; |
3673 | | |
3674 | 0 | tmask = 0; /* DTK_M(type); */ |
3675 | |
|
3676 | 0 | if (force_negative) |
3677 | 0 | { |
3678 | | /* val and fval should be of same sign, but test anyway */ |
3679 | 0 | if (val > 0) |
3680 | 0 | val = -val; |
3681 | 0 | if (fval > 0) |
3682 | 0 | fval = -fval; |
3683 | 0 | } |
3684 | |
|
3685 | 0 | switch (type) |
3686 | 0 | { |
3687 | 0 | case DTK_MICROSEC: |
3688 | 0 | if (!AdjustMicroseconds(val, fval, 1, itm_in)) |
3689 | 0 | return DTERR_FIELD_OVERFLOW; |
3690 | 0 | tmask = DTK_M(MICROSECOND); |
3691 | 0 | break; |
3692 | | |
3693 | 0 | case DTK_MILLISEC: |
3694 | 0 | if (!AdjustMicroseconds(val, fval, 1000, itm_in)) |
3695 | 0 | return DTERR_FIELD_OVERFLOW; |
3696 | 0 | tmask = DTK_M(MILLISECOND); |
3697 | 0 | break; |
3698 | | |
3699 | 0 | case DTK_SECOND: |
3700 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in)) |
3701 | 0 | return DTERR_FIELD_OVERFLOW; |
3702 | | |
3703 | | /* |
3704 | | * If any subseconds were specified, consider this |
3705 | | * microsecond and millisecond input as well. |
3706 | | */ |
3707 | 0 | if (fval == 0) |
3708 | 0 | tmask = DTK_M(SECOND); |
3709 | 0 | else |
3710 | 0 | tmask = DTK_ALL_SECS_M; |
3711 | 0 | break; |
3712 | | |
3713 | 0 | case DTK_MINUTE: |
3714 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in)) |
3715 | 0 | return DTERR_FIELD_OVERFLOW; |
3716 | 0 | tmask = DTK_M(MINUTE); |
3717 | 0 | break; |
3718 | | |
3719 | 0 | case DTK_HOUR: |
3720 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in)) |
3721 | 0 | return DTERR_FIELD_OVERFLOW; |
3722 | 0 | tmask = DTK_M(HOUR); |
3723 | 0 | type = DTK_DAY; /* set for next field */ |
3724 | 0 | break; |
3725 | | |
3726 | 0 | case DTK_DAY: |
3727 | 0 | if (!AdjustDays(val, 1, itm_in) || |
3728 | 0 | !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in)) |
3729 | 0 | return DTERR_FIELD_OVERFLOW; |
3730 | 0 | tmask = DTK_M(DAY); |
3731 | 0 | break; |
3732 | | |
3733 | 0 | case DTK_WEEK: |
3734 | 0 | if (!AdjustDays(val, 7, itm_in) || |
3735 | 0 | !AdjustFractDays(fval, 7, itm_in)) |
3736 | 0 | return DTERR_FIELD_OVERFLOW; |
3737 | 0 | tmask = DTK_M(WEEK); |
3738 | 0 | break; |
3739 | | |
3740 | 0 | case DTK_MONTH: |
3741 | 0 | if (!AdjustMonths(val, itm_in) || |
3742 | 0 | !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in)) |
3743 | 0 | return DTERR_FIELD_OVERFLOW; |
3744 | 0 | tmask = DTK_M(MONTH); |
3745 | 0 | break; |
3746 | | |
3747 | 0 | case DTK_YEAR: |
3748 | 0 | if (!AdjustYears(val, 1, itm_in) || |
3749 | 0 | !AdjustFractYears(fval, 1, itm_in)) |
3750 | 0 | return DTERR_FIELD_OVERFLOW; |
3751 | 0 | tmask = DTK_M(YEAR); |
3752 | 0 | break; |
3753 | | |
3754 | 0 | case DTK_DECADE: |
3755 | 0 | if (!AdjustYears(val, 10, itm_in) || |
3756 | 0 | !AdjustFractYears(fval, 10, itm_in)) |
3757 | 0 | return DTERR_FIELD_OVERFLOW; |
3758 | 0 | tmask = DTK_M(DECADE); |
3759 | 0 | break; |
3760 | | |
3761 | 0 | case DTK_CENTURY: |
3762 | 0 | if (!AdjustYears(val, 100, itm_in) || |
3763 | 0 | !AdjustFractYears(fval, 100, itm_in)) |
3764 | 0 | return DTERR_FIELD_OVERFLOW; |
3765 | 0 | tmask = DTK_M(CENTURY); |
3766 | 0 | break; |
3767 | | |
3768 | 0 | case DTK_MILLENNIUM: |
3769 | 0 | if (!AdjustYears(val, 1000, itm_in) || |
3770 | 0 | !AdjustFractYears(fval, 1000, itm_in)) |
3771 | 0 | return DTERR_FIELD_OVERFLOW; |
3772 | 0 | tmask = DTK_M(MILLENNIUM); |
3773 | 0 | break; |
3774 | | |
3775 | 0 | default: |
3776 | 0 | return DTERR_BAD_FORMAT; |
3777 | 0 | } |
3778 | 0 | parsing_unit_val = false; |
3779 | 0 | break; |
3780 | | |
3781 | 0 | case DTK_STRING: |
3782 | 0 | case DTK_SPECIAL: |
3783 | | /* reject consecutive unhandled units */ |
3784 | 0 | if (parsing_unit_val) |
3785 | 0 | return DTERR_BAD_FORMAT; |
3786 | 0 | type = DecodeUnits(i, field[i], &uval); |
3787 | 0 | if (type == UNKNOWN_FIELD) |
3788 | 0 | type = DecodeSpecial(i, field[i], &uval); |
3789 | 0 | if (type == IGNORE_DTF) |
3790 | 0 | continue; |
3791 | | |
3792 | 0 | tmask = 0; /* DTK_M(type); */ |
3793 | 0 | switch (type) |
3794 | 0 | { |
3795 | 0 | case UNITS: |
3796 | 0 | type = uval; |
3797 | 0 | parsing_unit_val = true; |
3798 | 0 | break; |
3799 | | |
3800 | 0 | case AGO: |
3801 | | |
3802 | | /* |
3803 | | * "ago" is only allowed to appear at the end of the |
3804 | | * interval. |
3805 | | */ |
3806 | 0 | if (i != nf - 1) |
3807 | 0 | return DTERR_BAD_FORMAT; |
3808 | 0 | is_before = true; |
3809 | 0 | type = uval; |
3810 | 0 | break; |
3811 | | |
3812 | 0 | case RESERV: |
3813 | 0 | tmask = (DTK_DATE_M | DTK_TIME_M); |
3814 | | |
3815 | | /* |
3816 | | * Only reserved words corresponding to infinite |
3817 | | * intervals are accepted. |
3818 | | */ |
3819 | 0 | if (uval != DTK_LATE && uval != DTK_EARLY) |
3820 | 0 | return DTERR_BAD_FORMAT; |
3821 | | |
3822 | | /* |
3823 | | * Infinity cannot be followed by anything else. We |
3824 | | * could allow "ago" to reverse the sign of infinity |
3825 | | * but using signed infinity is more intuitive. |
3826 | | */ |
3827 | 0 | if (i != nf - 1) |
3828 | 0 | return DTERR_BAD_FORMAT; |
3829 | | |
3830 | 0 | *dtype = uval; |
3831 | 0 | break; |
3832 | | |
3833 | 0 | default: |
3834 | 0 | return DTERR_BAD_FORMAT; |
3835 | 0 | } |
3836 | 0 | break; |
3837 | | |
3838 | 0 | default: |
3839 | 0 | return DTERR_BAD_FORMAT; |
3840 | 0 | } |
3841 | | |
3842 | 0 | if (tmask & fmask) |
3843 | 0 | return DTERR_BAD_FORMAT; |
3844 | 0 | fmask |= tmask; |
3845 | 0 | } |
3846 | | |
3847 | | /* ensure that at least one time field has been found */ |
3848 | 0 | if (fmask == 0) |
3849 | 0 | return DTERR_BAD_FORMAT; |
3850 | | |
3851 | | /* reject if unit appeared and was never handled */ |
3852 | 0 | if (parsing_unit_val) |
3853 | 0 | return DTERR_BAD_FORMAT; |
3854 | | |
3855 | | /* finally, AGO negates everything */ |
3856 | 0 | if (is_before) |
3857 | 0 | { |
3858 | 0 | if (itm_in->tm_usec == PG_INT64_MIN || |
3859 | 0 | itm_in->tm_mday == INT_MIN || |
3860 | 0 | itm_in->tm_mon == INT_MIN || |
3861 | 0 | itm_in->tm_year == INT_MIN) |
3862 | 0 | return DTERR_FIELD_OVERFLOW; |
3863 | | |
3864 | 0 | itm_in->tm_usec = -itm_in->tm_usec; |
3865 | 0 | itm_in->tm_mday = -itm_in->tm_mday; |
3866 | 0 | itm_in->tm_mon = -itm_in->tm_mon; |
3867 | 0 | itm_in->tm_year = -itm_in->tm_year; |
3868 | 0 | } |
3869 | | |
3870 | 0 | return 0; |
3871 | 0 | } |
3872 | | |
3873 | | |
3874 | | /* |
3875 | | * Helper functions to avoid duplicated code in DecodeISO8601Interval. |
3876 | | * |
3877 | | * Parse a decimal value and break it into integer and fractional parts. |
3878 | | * Set *endptr to end+1 of the parsed substring. |
3879 | | * Returns 0 or DTERR code. |
3880 | | */ |
3881 | | static int |
3882 | | ParseISO8601Number(char *str, char **endptr, int64 *ipart, double *fpart) |
3883 | 0 | { |
3884 | 0 | double val; |
3885 | | |
3886 | | /* |
3887 | | * Historically this has accepted anything that strtod() would take, |
3888 | | * notably including "e" notation, so continue doing that. This is |
3889 | | * slightly annoying because the precision of double is less than that of |
3890 | | * int64, so we would lose accuracy for inputs larger than 2^53 or so. |
3891 | | * However, historically we rejected inputs outside the int32 range, |
3892 | | * making that concern moot. What we do now is reject abs(val) above |
3893 | | * 1.0e15 (a round number a bit less than 2^50), so that any accepted |
3894 | | * value will have an exact integer part, and thereby a fraction part with |
3895 | | * abs(*fpart) less than 1. In the absence of field complaints it doesn't |
3896 | | * seem worth working harder. |
3897 | | */ |
3898 | 0 | if (!(isdigit((unsigned char) *str) || *str == '-' || *str == '.')) |
3899 | 0 | return DTERR_BAD_FORMAT; |
3900 | 0 | errno = 0; |
3901 | 0 | val = strtod(str, endptr); |
3902 | | /* did we not see anything that looks like a double? */ |
3903 | 0 | if (*endptr == str || errno != 0) |
3904 | 0 | return DTERR_BAD_FORMAT; |
3905 | | /* watch out for overflow, including infinities; reject NaN too */ |
3906 | 0 | if (isnan(val) || val < -1.0e15 || val > 1.0e15) |
3907 | 0 | return DTERR_FIELD_OVERFLOW; |
3908 | | /* be very sure we truncate towards zero (cf dtrunc()) */ |
3909 | 0 | if (val >= 0) |
3910 | 0 | *ipart = (int64) floor(val); |
3911 | 0 | else |
3912 | 0 | *ipart = (int64) -floor(-val); |
3913 | 0 | *fpart = val - *ipart; |
3914 | | /* Callers expect this to hold */ |
3915 | 0 | Assert(*fpart > -1.0 && *fpart < 1.0); |
3916 | 0 | return 0; |
3917 | 0 | } |
3918 | | |
3919 | | /* |
3920 | | * Determine number of integral digits in a valid ISO 8601 number field |
3921 | | * (we should ignore sign and any fraction part) |
3922 | | */ |
3923 | | static int |
3924 | | ISO8601IntegerWidth(char *fieldstart) |
3925 | 0 | { |
3926 | | /* We might have had a leading '-' */ |
3927 | 0 | if (*fieldstart == '-') |
3928 | 0 | fieldstart++; |
3929 | 0 | return strspn(fieldstart, "0123456789"); |
3930 | 0 | } |
3931 | | |
3932 | | |
3933 | | /* DecodeISO8601Interval() |
3934 | | * Decode an ISO 8601 time interval of the "format with designators" |
3935 | | * (section 4.4.3.2) or "alternative format" (section 4.4.3.3) |
3936 | | * Examples: P1D for 1 day |
3937 | | * PT1H for 1 hour |
3938 | | * P2Y6M7DT1H30M for 2 years, 6 months, 7 days 1 hour 30 min |
3939 | | * P0002-06-07T01:30:00 the same value in alternative format |
3940 | | * |
3941 | | * Returns 0 if successful, DTERR code if bogus input detected. |
3942 | | * Note: error code should be DTERR_BAD_FORMAT if input doesn't look like |
3943 | | * ISO8601, otherwise this could cause unexpected error messages. |
3944 | | * dtype and itm_in are output parameters. |
3945 | | * |
3946 | | * A couple exceptions from the spec: |
3947 | | * - a week field ('W') may coexist with other units |
3948 | | * - allows decimals in fields other than the least significant unit. |
3949 | | */ |
3950 | | int |
3951 | | DecodeISO8601Interval(char *str, |
3952 | | int *dtype, struct pg_itm_in *itm_in) |
3953 | 0 | { |
3954 | 0 | bool datepart = true; |
3955 | 0 | bool havefield = false; |
3956 | |
|
3957 | 0 | *dtype = DTK_DELTA; |
3958 | 0 | ClearPgItmIn(itm_in); |
3959 | |
|
3960 | 0 | if (strlen(str) < 2 || str[0] != 'P') |
3961 | 0 | return DTERR_BAD_FORMAT; |
3962 | | |
3963 | 0 | str++; |
3964 | 0 | while (*str) |
3965 | 0 | { |
3966 | 0 | char *fieldstart; |
3967 | 0 | int64 val; |
3968 | 0 | double fval; |
3969 | 0 | char unit; |
3970 | 0 | int dterr; |
3971 | |
|
3972 | 0 | if (*str == 'T') /* T indicates the beginning of the time part */ |
3973 | 0 | { |
3974 | 0 | datepart = false; |
3975 | 0 | havefield = false; |
3976 | 0 | str++; |
3977 | 0 | continue; |
3978 | 0 | } |
3979 | | |
3980 | 0 | fieldstart = str; |
3981 | 0 | dterr = ParseISO8601Number(str, &str, &val, &fval); |
3982 | 0 | if (dterr) |
3983 | 0 | return dterr; |
3984 | | |
3985 | | /* |
3986 | | * Note: we could step off the end of the string here. Code below |
3987 | | * *must* exit the loop if unit == '\0'. |
3988 | | */ |
3989 | 0 | unit = *str++; |
3990 | |
|
3991 | 0 | if (datepart) |
3992 | 0 | { |
3993 | 0 | switch (unit) /* before T: Y M W D */ |
3994 | 0 | { |
3995 | 0 | case 'Y': |
3996 | 0 | if (!AdjustYears(val, 1, itm_in) || |
3997 | 0 | !AdjustFractYears(fval, 1, itm_in)) |
3998 | 0 | return DTERR_FIELD_OVERFLOW; |
3999 | 0 | break; |
4000 | 0 | case 'M': |
4001 | 0 | if (!AdjustMonths(val, itm_in) || |
4002 | 0 | !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in)) |
4003 | 0 | return DTERR_FIELD_OVERFLOW; |
4004 | 0 | break; |
4005 | 0 | case 'W': |
4006 | 0 | if (!AdjustDays(val, 7, itm_in) || |
4007 | 0 | !AdjustFractDays(fval, 7, itm_in)) |
4008 | 0 | return DTERR_FIELD_OVERFLOW; |
4009 | 0 | break; |
4010 | 0 | case 'D': |
4011 | 0 | if (!AdjustDays(val, 1, itm_in) || |
4012 | 0 | !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in)) |
4013 | 0 | return DTERR_FIELD_OVERFLOW; |
4014 | 0 | break; |
4015 | 0 | case 'T': /* ISO 8601 4.4.3.3 Alternative Format / Basic */ |
4016 | 0 | case '\0': |
4017 | 0 | if (ISO8601IntegerWidth(fieldstart) == 8 && !havefield) |
4018 | 0 | { |
4019 | 0 | if (!AdjustYears(val / 10000, 1, itm_in) || |
4020 | 0 | !AdjustMonths((val / 100) % 100, itm_in) || |
4021 | 0 | !AdjustDays(val % 100, 1, itm_in) || |
4022 | 0 | !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in)) |
4023 | 0 | return DTERR_FIELD_OVERFLOW; |
4024 | 0 | if (unit == '\0') |
4025 | 0 | return 0; |
4026 | 0 | datepart = false; |
4027 | 0 | havefield = false; |
4028 | 0 | continue; |
4029 | 0 | } |
4030 | | /* Else fall through to extended alternative format */ |
4031 | | /* FALLTHROUGH */ |
4032 | 0 | case '-': /* ISO 8601 4.4.3.3 Alternative Format, |
4033 | | * Extended */ |
4034 | 0 | if (havefield) |
4035 | 0 | return DTERR_BAD_FORMAT; |
4036 | | |
4037 | 0 | if (!AdjustYears(val, 1, itm_in) || |
4038 | 0 | !AdjustFractYears(fval, 1, itm_in)) |
4039 | 0 | return DTERR_FIELD_OVERFLOW; |
4040 | 0 | if (unit == '\0') |
4041 | 0 | return 0; |
4042 | 0 | if (unit == 'T') |
4043 | 0 | { |
4044 | 0 | datepart = false; |
4045 | 0 | havefield = false; |
4046 | 0 | continue; |
4047 | 0 | } |
4048 | | |
4049 | 0 | dterr = ParseISO8601Number(str, &str, &val, &fval); |
4050 | 0 | if (dterr) |
4051 | 0 | return dterr; |
4052 | 0 | if (!AdjustMonths(val, itm_in) || |
4053 | 0 | !AdjustFractDays(fval, DAYS_PER_MONTH, itm_in)) |
4054 | 0 | return DTERR_FIELD_OVERFLOW; |
4055 | 0 | if (*str == '\0') |
4056 | 0 | return 0; |
4057 | 0 | if (*str == 'T') |
4058 | 0 | { |
4059 | 0 | datepart = false; |
4060 | 0 | havefield = false; |
4061 | 0 | continue; |
4062 | 0 | } |
4063 | 0 | if (*str != '-') |
4064 | 0 | return DTERR_BAD_FORMAT; |
4065 | 0 | str++; |
4066 | |
|
4067 | 0 | dterr = ParseISO8601Number(str, &str, &val, &fval); |
4068 | 0 | if (dterr) |
4069 | 0 | return dterr; |
4070 | 0 | if (!AdjustDays(val, 1, itm_in) || |
4071 | 0 | !AdjustFractMicroseconds(fval, USECS_PER_DAY, itm_in)) |
4072 | 0 | return DTERR_FIELD_OVERFLOW; |
4073 | 0 | if (*str == '\0') |
4074 | 0 | return 0; |
4075 | 0 | if (*str == 'T') |
4076 | 0 | { |
4077 | 0 | datepart = false; |
4078 | 0 | havefield = false; |
4079 | 0 | continue; |
4080 | 0 | } |
4081 | 0 | return DTERR_BAD_FORMAT; |
4082 | 0 | default: |
4083 | | /* not a valid date unit suffix */ |
4084 | 0 | return DTERR_BAD_FORMAT; |
4085 | 0 | } |
4086 | 0 | } |
4087 | 0 | else |
4088 | 0 | { |
4089 | 0 | switch (unit) /* after T: H M S */ |
4090 | 0 | { |
4091 | 0 | case 'H': |
4092 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in)) |
4093 | 0 | return DTERR_FIELD_OVERFLOW; |
4094 | 0 | break; |
4095 | 0 | case 'M': |
4096 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in)) |
4097 | 0 | return DTERR_FIELD_OVERFLOW; |
4098 | 0 | break; |
4099 | 0 | case 'S': |
4100 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in)) |
4101 | 0 | return DTERR_FIELD_OVERFLOW; |
4102 | 0 | break; |
4103 | 0 | case '\0': /* ISO 8601 4.4.3.3 Alternative Format */ |
4104 | 0 | if (ISO8601IntegerWidth(fieldstart) == 6 && !havefield) |
4105 | 0 | { |
4106 | 0 | if (!AdjustMicroseconds(val / 10000, 0, USECS_PER_HOUR, itm_in) || |
4107 | 0 | !AdjustMicroseconds((val / 100) % 100, 0, USECS_PER_MINUTE, itm_in) || |
4108 | 0 | !AdjustMicroseconds(val % 100, 0, USECS_PER_SEC, itm_in) || |
4109 | 0 | !AdjustFractMicroseconds(fval, 1, itm_in)) |
4110 | 0 | return DTERR_FIELD_OVERFLOW; |
4111 | 0 | return 0; |
4112 | 0 | } |
4113 | | /* Else fall through to extended alternative format */ |
4114 | | /* FALLTHROUGH */ |
4115 | 0 | case ':': /* ISO 8601 4.4.3.3 Alternative Format, |
4116 | | * Extended */ |
4117 | 0 | if (havefield) |
4118 | 0 | return DTERR_BAD_FORMAT; |
4119 | | |
4120 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_HOUR, itm_in)) |
4121 | 0 | return DTERR_FIELD_OVERFLOW; |
4122 | 0 | if (unit == '\0') |
4123 | 0 | return 0; |
4124 | | |
4125 | 0 | dterr = ParseISO8601Number(str, &str, &val, &fval); |
4126 | 0 | if (dterr) |
4127 | 0 | return dterr; |
4128 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_MINUTE, itm_in)) |
4129 | 0 | return DTERR_FIELD_OVERFLOW; |
4130 | 0 | if (*str == '\0') |
4131 | 0 | return 0; |
4132 | 0 | if (*str != ':') |
4133 | 0 | return DTERR_BAD_FORMAT; |
4134 | 0 | str++; |
4135 | |
|
4136 | 0 | dterr = ParseISO8601Number(str, &str, &val, &fval); |
4137 | 0 | if (dterr) |
4138 | 0 | return dterr; |
4139 | 0 | if (!AdjustMicroseconds(val, fval, USECS_PER_SEC, itm_in)) |
4140 | 0 | return DTERR_FIELD_OVERFLOW; |
4141 | 0 | if (*str == '\0') |
4142 | 0 | return 0; |
4143 | 0 | return DTERR_BAD_FORMAT; |
4144 | | |
4145 | 0 | default: |
4146 | | /* not a valid time unit suffix */ |
4147 | 0 | return DTERR_BAD_FORMAT; |
4148 | 0 | } |
4149 | 0 | } |
4150 | | |
4151 | 0 | havefield = true; |
4152 | 0 | } |
4153 | | |
4154 | 0 | return 0; |
4155 | 0 | } |
4156 | | |
4157 | | |
4158 | | /* DecodeUnits() |
4159 | | * Decode text string using lookup table. |
4160 | | * |
4161 | | * This routine recognizes keywords associated with time interval units. |
4162 | | * |
4163 | | * Given string must be lowercased already. |
4164 | | * |
4165 | | * Implement a cache lookup since it is likely that dates |
4166 | | * will be related in format. |
4167 | | */ |
4168 | | int |
4169 | | DecodeUnits(int field, const char *lowtoken, int *val) |
4170 | 0 | { |
4171 | 0 | int type; |
4172 | 0 | const datetkn *tp; |
4173 | |
|
4174 | 0 | tp = deltacache[field]; |
4175 | | /* use strncmp so that we match truncated tokens */ |
4176 | 0 | if (tp == NULL || strncmp(lowtoken, tp->token, TOKMAXLEN) != 0) |
4177 | 0 | { |
4178 | 0 | tp = datebsearch(lowtoken, deltatktbl, szdeltatktbl); |
4179 | 0 | } |
4180 | 0 | if (tp == NULL) |
4181 | 0 | { |
4182 | 0 | type = UNKNOWN_FIELD; |
4183 | 0 | *val = 0; |
4184 | 0 | } |
4185 | 0 | else |
4186 | 0 | { |
4187 | 0 | deltacache[field] = tp; |
4188 | 0 | type = tp->type; |
4189 | 0 | *val = tp->value; |
4190 | 0 | } |
4191 | |
|
4192 | 0 | return type; |
4193 | 0 | } /* DecodeUnits() */ |
4194 | | |
4195 | | /* |
4196 | | * Report an error detected by one of the datetime input processing routines. |
4197 | | * |
4198 | | * dterr is the error code, and *extra contains any auxiliary info we need |
4199 | | * for the error report. extra can be NULL if not needed for the particular |
4200 | | * dterr value. |
4201 | | * |
4202 | | * str is the original input string, and datatype is the name of the datatype |
4203 | | * we were trying to accept. (For some DTERR codes, these are not used and |
4204 | | * can be NULL.) |
4205 | | * |
4206 | | * If escontext points to an ErrorSaveContext node, that is filled instead |
4207 | | * of throwing an error. |
4208 | | * |
4209 | | * Note: it might seem useless to distinguish DTERR_INTERVAL_OVERFLOW and |
4210 | | * DTERR_TZDISP_OVERFLOW from DTERR_FIELD_OVERFLOW, but SQL99 mandates three |
4211 | | * separate SQLSTATE codes, so ... |
4212 | | */ |
4213 | | void |
4214 | | DateTimeParseError(int dterr, DateTimeErrorExtra *extra, |
4215 | | const char *str, const char *datatype, |
4216 | | Node *escontext) |
4217 | 0 | { |
4218 | 0 | switch (dterr) |
4219 | 0 | { |
4220 | 0 | case DTERR_FIELD_OVERFLOW: |
4221 | 0 | errsave(escontext, |
4222 | 0 | (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW), |
4223 | 0 | errmsg("date/time field value out of range: \"%s\"", |
4224 | 0 | str))); |
4225 | 0 | break; |
4226 | 0 | case DTERR_MD_FIELD_OVERFLOW: |
4227 | | /* <nanny>same as above, but add hint about DateStyle</nanny> */ |
4228 | 0 | errsave(escontext, |
4229 | 0 | (errcode(ERRCODE_DATETIME_FIELD_OVERFLOW), |
4230 | 0 | errmsg("date/time field value out of range: \"%s\"", |
4231 | 0 | str), |
4232 | 0 | errhint("Perhaps you need a different \"DateStyle\" setting."))); |
4233 | 0 | break; |
4234 | 0 | case DTERR_INTERVAL_OVERFLOW: |
4235 | 0 | errsave(escontext, |
4236 | 0 | (errcode(ERRCODE_INTERVAL_FIELD_OVERFLOW), |
4237 | 0 | errmsg("interval field value out of range: \"%s\"", |
4238 | 0 | str))); |
4239 | 0 | break; |
4240 | 0 | case DTERR_TZDISP_OVERFLOW: |
4241 | 0 | errsave(escontext, |
4242 | 0 | (errcode(ERRCODE_INVALID_TIME_ZONE_DISPLACEMENT_VALUE), |
4243 | 0 | errmsg("time zone displacement out of range: \"%s\"", |
4244 | 0 | str))); |
4245 | 0 | break; |
4246 | 0 | case DTERR_BAD_TIMEZONE: |
4247 | 0 | errsave(escontext, |
4248 | 0 | (errcode(ERRCODE_INVALID_PARAMETER_VALUE), |
4249 | 0 | errmsg("time zone \"%s\" not recognized", |
4250 | 0 | extra->dtee_timezone))); |
4251 | 0 | break; |
4252 | 0 | case DTERR_BAD_ZONE_ABBREV: |
4253 | 0 | errsave(escontext, |
4254 | 0 | (errcode(ERRCODE_CONFIG_FILE_ERROR), |
4255 | 0 | errmsg("time zone \"%s\" not recognized", |
4256 | 0 | extra->dtee_timezone), |
4257 | 0 | errdetail("This time zone name appears in the configuration file for time zone abbreviation \"%s\".", |
4258 | 0 | extra->dtee_abbrev))); |
4259 | 0 | break; |
4260 | 0 | case DTERR_BAD_FORMAT: |
4261 | 0 | default: |
4262 | 0 | errsave(escontext, |
4263 | 0 | (errcode(ERRCODE_INVALID_DATETIME_FORMAT), |
4264 | 0 | errmsg("invalid input syntax for type %s: \"%s\"", |
4265 | 0 | datatype, str))); |
4266 | 0 | break; |
4267 | 0 | } |
4268 | 0 | } |
4269 | | |
4270 | | /* datebsearch() |
4271 | | * Binary search -- from Knuth (6.2.1) Algorithm B. Special case like this |
4272 | | * is WAY faster than the generic bsearch(). |
4273 | | */ |
4274 | | static const datetkn * |
4275 | | datebsearch(const char *key, const datetkn *base, int nel) |
4276 | 0 | { |
4277 | 0 | if (nel > 0) |
4278 | 0 | { |
4279 | 0 | const datetkn *last = base + nel - 1, |
4280 | 0 | *position; |
4281 | 0 | int result; |
4282 | |
|
4283 | 0 | while (last >= base) |
4284 | 0 | { |
4285 | 0 | position = base + ((last - base) >> 1); |
4286 | | /* precheck the first character for a bit of extra speed */ |
4287 | 0 | result = (int) key[0] - (int) position->token[0]; |
4288 | 0 | if (result == 0) |
4289 | 0 | { |
4290 | | /* use strncmp so that we match truncated tokens */ |
4291 | 0 | result = strncmp(key, position->token, TOKMAXLEN); |
4292 | 0 | if (result == 0) |
4293 | 0 | return position; |
4294 | 0 | } |
4295 | 0 | if (result < 0) |
4296 | 0 | last = position - 1; |
4297 | 0 | else |
4298 | 0 | base = position + 1; |
4299 | 0 | } |
4300 | 0 | } |
4301 | 0 | return NULL; |
4302 | 0 | } |
4303 | | |
4304 | | /* EncodeTimezone() |
4305 | | * Copies representation of a numeric timezone offset to str. |
4306 | | * |
4307 | | * Returns a pointer to the new end of string. No NUL terminator is put |
4308 | | * there; callers are responsible for NUL terminating str themselves. |
4309 | | */ |
4310 | | static char * |
4311 | | EncodeTimezone(char *str, int tz, int style) |
4312 | 0 | { |
4313 | 0 | int hour, |
4314 | 0 | min, |
4315 | 0 | sec; |
4316 | |
|
4317 | 0 | sec = abs(tz); |
4318 | 0 | min = sec / SECS_PER_MINUTE; |
4319 | 0 | sec -= min * SECS_PER_MINUTE; |
4320 | 0 | hour = min / MINS_PER_HOUR; |
4321 | 0 | min -= hour * MINS_PER_HOUR; |
4322 | | |
4323 | | /* TZ is negated compared to sign we wish to display ... */ |
4324 | 0 | *str++ = (tz <= 0 ? '+' : '-'); |
4325 | |
|
4326 | 0 | if (sec != 0) |
4327 | 0 | { |
4328 | 0 | str = pg_ultostr_zeropad(str, hour, 2); |
4329 | 0 | *str++ = ':'; |
4330 | 0 | str = pg_ultostr_zeropad(str, min, 2); |
4331 | 0 | *str++ = ':'; |
4332 | 0 | str = pg_ultostr_zeropad(str, sec, 2); |
4333 | 0 | } |
4334 | 0 | else if (min != 0 || style == USE_XSD_DATES) |
4335 | 0 | { |
4336 | 0 | str = pg_ultostr_zeropad(str, hour, 2); |
4337 | 0 | *str++ = ':'; |
4338 | 0 | str = pg_ultostr_zeropad(str, min, 2); |
4339 | 0 | } |
4340 | 0 | else |
4341 | 0 | str = pg_ultostr_zeropad(str, hour, 2); |
4342 | 0 | return str; |
4343 | 0 | } |
4344 | | |
4345 | | /* EncodeDateOnly() |
4346 | | * Encode date as local time. |
4347 | | */ |
4348 | | void |
4349 | | EncodeDateOnly(struct pg_tm *tm, int style, char *str) |
4350 | 0 | { |
4351 | 0 | Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR); |
4352 | |
|
4353 | 0 | switch (style) |
4354 | 0 | { |
4355 | 0 | case USE_ISO_DATES: |
4356 | 0 | case USE_XSD_DATES: |
4357 | | /* compatible with ISO date formats */ |
4358 | 0 | str = pg_ultostr_zeropad(str, |
4359 | 0 | (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4); |
4360 | 0 | *str++ = '-'; |
4361 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4362 | 0 | *str++ = '-'; |
4363 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4364 | 0 | break; |
4365 | | |
4366 | 0 | case USE_SQL_DATES: |
4367 | | /* compatible with Oracle/Ingres date formats */ |
4368 | 0 | if (DateOrder == DATEORDER_DMY) |
4369 | 0 | { |
4370 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4371 | 0 | *str++ = '/'; |
4372 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4373 | 0 | } |
4374 | 0 | else |
4375 | 0 | { |
4376 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4377 | 0 | *str++ = '/'; |
4378 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4379 | 0 | } |
4380 | 0 | *str++ = '/'; |
4381 | 0 | str = pg_ultostr_zeropad(str, |
4382 | 0 | (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4); |
4383 | 0 | break; |
4384 | | |
4385 | 0 | case USE_GERMAN_DATES: |
4386 | | /* German-style date format */ |
4387 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4388 | 0 | *str++ = '.'; |
4389 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4390 | 0 | *str++ = '.'; |
4391 | 0 | str = pg_ultostr_zeropad(str, |
4392 | 0 | (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4); |
4393 | 0 | break; |
4394 | | |
4395 | 0 | case USE_POSTGRES_DATES: |
4396 | 0 | default: |
4397 | | /* traditional date-only style for Postgres */ |
4398 | 0 | if (DateOrder == DATEORDER_DMY) |
4399 | 0 | { |
4400 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4401 | 0 | *str++ = '-'; |
4402 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4403 | 0 | } |
4404 | 0 | else |
4405 | 0 | { |
4406 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4407 | 0 | *str++ = '-'; |
4408 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4409 | 0 | } |
4410 | 0 | *str++ = '-'; |
4411 | 0 | str = pg_ultostr_zeropad(str, |
4412 | 0 | (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4); |
4413 | 0 | break; |
4414 | 0 | } |
4415 | | |
4416 | 0 | if (tm->tm_year <= 0) |
4417 | 0 | { |
4418 | 0 | memcpy(str, " BC", 3); /* Don't copy NUL */ |
4419 | 0 | str += 3; |
4420 | 0 | } |
4421 | 0 | *str = '\0'; |
4422 | 0 | } |
4423 | | |
4424 | | |
4425 | | /* EncodeTimeOnly() |
4426 | | * Encode time fields only. |
4427 | | * |
4428 | | * tm and fsec are the value to encode, print_tz determines whether to include |
4429 | | * a time zone (the difference between time and timetz types), tz is the |
4430 | | * numeric time zone offset, style is the date style, str is where to write the |
4431 | | * output. |
4432 | | */ |
4433 | | void |
4434 | | EncodeTimeOnly(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, int style, char *str) |
4435 | 0 | { |
4436 | 0 | str = pg_ultostr_zeropad(str, tm->tm_hour, 2); |
4437 | 0 | *str++ = ':'; |
4438 | 0 | str = pg_ultostr_zeropad(str, tm->tm_min, 2); |
4439 | 0 | *str++ = ':'; |
4440 | 0 | str = AppendSeconds(str, tm->tm_sec, fsec, MAX_TIME_PRECISION, true); |
4441 | 0 | if (print_tz) |
4442 | 0 | str = EncodeTimezone(str, tz, style); |
4443 | 0 | *str = '\0'; |
4444 | 0 | } |
4445 | | |
4446 | | |
4447 | | /* EncodeDateTime() |
4448 | | * Encode date and time interpreted as local time. |
4449 | | * |
4450 | | * tm and fsec are the value to encode, print_tz determines whether to include |
4451 | | * a time zone (the difference between timestamp and timestamptz types), tz is |
4452 | | * the numeric time zone offset, tzn is the textual time zone, which if |
4453 | | * specified will be used instead of tz by some styles, style is the date |
4454 | | * style, str is where to write the output. |
4455 | | * |
4456 | | * Supported date styles: |
4457 | | * Postgres - day mon hh:mm:ss yyyy tz |
4458 | | * SQL - mm/dd/yyyy hh:mm:ss.ss tz |
4459 | | * ISO - yyyy-mm-dd hh:mm:ss+/-tz |
4460 | | * German - dd.mm.yyyy hh:mm:ss tz |
4461 | | * XSD - yyyy-mm-ddThh:mm:ss.ss+/-tz |
4462 | | */ |
4463 | | void |
4464 | | EncodeDateTime(struct pg_tm *tm, fsec_t fsec, bool print_tz, int tz, const char *tzn, int style, char *str) |
4465 | 0 | { |
4466 | 0 | int day; |
4467 | |
|
4468 | 0 | Assert(tm->tm_mon >= 1 && tm->tm_mon <= MONTHS_PER_YEAR); |
4469 | | |
4470 | | /* |
4471 | | * Negative tm_isdst means we have no valid time zone translation. |
4472 | | */ |
4473 | 0 | if (tm->tm_isdst < 0) |
4474 | 0 | print_tz = false; |
4475 | |
|
4476 | 0 | switch (style) |
4477 | 0 | { |
4478 | 0 | case USE_ISO_DATES: |
4479 | 0 | case USE_XSD_DATES: |
4480 | | /* Compatible with ISO-8601 date formats */ |
4481 | 0 | str = pg_ultostr_zeropad(str, |
4482 | 0 | (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4); |
4483 | 0 | *str++ = '-'; |
4484 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4485 | 0 | *str++ = '-'; |
4486 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4487 | 0 | *str++ = (style == USE_ISO_DATES) ? ' ' : 'T'; |
4488 | 0 | str = pg_ultostr_zeropad(str, tm->tm_hour, 2); |
4489 | 0 | *str++ = ':'; |
4490 | 0 | str = pg_ultostr_zeropad(str, tm->tm_min, 2); |
4491 | 0 | *str++ = ':'; |
4492 | 0 | str = AppendTimestampSeconds(str, tm, fsec); |
4493 | 0 | if (print_tz) |
4494 | 0 | str = EncodeTimezone(str, tz, style); |
4495 | 0 | break; |
4496 | | |
4497 | 0 | case USE_SQL_DATES: |
4498 | | /* Compatible with Oracle/Ingres date formats */ |
4499 | 0 | if (DateOrder == DATEORDER_DMY) |
4500 | 0 | { |
4501 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4502 | 0 | *str++ = '/'; |
4503 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4504 | 0 | } |
4505 | 0 | else |
4506 | 0 | { |
4507 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4508 | 0 | *str++ = '/'; |
4509 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4510 | 0 | } |
4511 | 0 | *str++ = '/'; |
4512 | 0 | str = pg_ultostr_zeropad(str, |
4513 | 0 | (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4); |
4514 | 0 | *str++ = ' '; |
4515 | 0 | str = pg_ultostr_zeropad(str, tm->tm_hour, 2); |
4516 | 0 | *str++ = ':'; |
4517 | 0 | str = pg_ultostr_zeropad(str, tm->tm_min, 2); |
4518 | 0 | *str++ = ':'; |
4519 | 0 | str = AppendTimestampSeconds(str, tm, fsec); |
4520 | | |
4521 | | /* |
4522 | | * Note: the uses of %.*s in this function would be risky if the |
4523 | | * timezone names ever contain non-ASCII characters, since we are |
4524 | | * not being careful to do encoding-aware clipping. However, all |
4525 | | * TZ abbreviations in the IANA database are plain ASCII. |
4526 | | */ |
4527 | 0 | if (print_tz) |
4528 | 0 | { |
4529 | 0 | if (tzn) |
4530 | 0 | { |
4531 | 0 | sprintf(str, " %.*s", MAXTZLEN, tzn); |
4532 | 0 | str += strlen(str); |
4533 | 0 | } |
4534 | 0 | else |
4535 | 0 | str = EncodeTimezone(str, tz, style); |
4536 | 0 | } |
4537 | 0 | break; |
4538 | | |
4539 | 0 | case USE_GERMAN_DATES: |
4540 | | /* German variant on European style */ |
4541 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4542 | 0 | *str++ = '.'; |
4543 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mon, 2); |
4544 | 0 | *str++ = '.'; |
4545 | 0 | str = pg_ultostr_zeropad(str, |
4546 | 0 | (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4); |
4547 | 0 | *str++ = ' '; |
4548 | 0 | str = pg_ultostr_zeropad(str, tm->tm_hour, 2); |
4549 | 0 | *str++ = ':'; |
4550 | 0 | str = pg_ultostr_zeropad(str, tm->tm_min, 2); |
4551 | 0 | *str++ = ':'; |
4552 | 0 | str = AppendTimestampSeconds(str, tm, fsec); |
4553 | |
|
4554 | 0 | if (print_tz) |
4555 | 0 | { |
4556 | 0 | if (tzn) |
4557 | 0 | { |
4558 | 0 | sprintf(str, " %.*s", MAXTZLEN, tzn); |
4559 | 0 | str += strlen(str); |
4560 | 0 | } |
4561 | 0 | else |
4562 | 0 | str = EncodeTimezone(str, tz, style); |
4563 | 0 | } |
4564 | 0 | break; |
4565 | | |
4566 | 0 | case USE_POSTGRES_DATES: |
4567 | 0 | default: |
4568 | | /* Backward-compatible with traditional Postgres abstime dates */ |
4569 | 0 | day = date2j(tm->tm_year, tm->tm_mon, tm->tm_mday); |
4570 | 0 | tm->tm_wday = j2day(day); |
4571 | 0 | memcpy(str, days[tm->tm_wday], 3); |
4572 | 0 | str += 3; |
4573 | 0 | *str++ = ' '; |
4574 | 0 | if (DateOrder == DATEORDER_DMY) |
4575 | 0 | { |
4576 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4577 | 0 | *str++ = ' '; |
4578 | 0 | memcpy(str, months[tm->tm_mon - 1], 3); |
4579 | 0 | str += 3; |
4580 | 0 | } |
4581 | 0 | else |
4582 | 0 | { |
4583 | 0 | memcpy(str, months[tm->tm_mon - 1], 3); |
4584 | 0 | str += 3; |
4585 | 0 | *str++ = ' '; |
4586 | 0 | str = pg_ultostr_zeropad(str, tm->tm_mday, 2); |
4587 | 0 | } |
4588 | 0 | *str++ = ' '; |
4589 | 0 | str = pg_ultostr_zeropad(str, tm->tm_hour, 2); |
4590 | 0 | *str++ = ':'; |
4591 | 0 | str = pg_ultostr_zeropad(str, tm->tm_min, 2); |
4592 | 0 | *str++ = ':'; |
4593 | 0 | str = AppendTimestampSeconds(str, tm, fsec); |
4594 | 0 | *str++ = ' '; |
4595 | 0 | str = pg_ultostr_zeropad(str, |
4596 | 0 | (tm->tm_year > 0) ? tm->tm_year : -(tm->tm_year - 1), 4); |
4597 | |
|
4598 | 0 | if (print_tz) |
4599 | 0 | { |
4600 | 0 | if (tzn) |
4601 | 0 | { |
4602 | 0 | sprintf(str, " %.*s", MAXTZLEN, tzn); |
4603 | 0 | str += strlen(str); |
4604 | 0 | } |
4605 | 0 | else |
4606 | 0 | { |
4607 | | /* |
4608 | | * We have a time zone, but no string version. Use the |
4609 | | * numeric form, but be sure to include a leading space to |
4610 | | * avoid formatting something which would be rejected by |
4611 | | * the date/time parser later. - thomas 2001-10-19 |
4612 | | */ |
4613 | 0 | *str++ = ' '; |
4614 | 0 | str = EncodeTimezone(str, tz, style); |
4615 | 0 | } |
4616 | 0 | } |
4617 | 0 | break; |
4618 | 0 | } |
4619 | | |
4620 | 0 | if (tm->tm_year <= 0) |
4621 | 0 | { |
4622 | 0 | memcpy(str, " BC", 3); /* Don't copy NUL */ |
4623 | 0 | str += 3; |
4624 | 0 | } |
4625 | 0 | *str = '\0'; |
4626 | 0 | } |
4627 | | |
4628 | | |
4629 | | /* |
4630 | | * Helper functions to avoid duplicated code in EncodeInterval. |
4631 | | */ |
4632 | | |
4633 | | /* Append an ISO-8601-style interval field, but only if value isn't zero */ |
4634 | | static char * |
4635 | | AddISO8601IntPart(char *cp, int64 value, char units) |
4636 | 0 | { |
4637 | 0 | if (value == 0) |
4638 | 0 | return cp; |
4639 | 0 | sprintf(cp, "%" PRId64 "%c", value, units); |
4640 | 0 | return cp + strlen(cp); |
4641 | 0 | } |
4642 | | |
4643 | | /* Append a postgres-style interval field, but only if value isn't zero */ |
4644 | | static char * |
4645 | | AddPostgresIntPart(char *cp, int64 value, const char *units, |
4646 | | bool *is_zero, bool *is_before) |
4647 | 0 | { |
4648 | 0 | if (value == 0) |
4649 | 0 | return cp; |
4650 | 0 | sprintf(cp, "%s%s%" PRId64 " %s%s", |
4651 | 0 | (!*is_zero) ? " " : "", |
4652 | 0 | (*is_before && value > 0) ? "+" : "", |
4653 | 0 | value, |
4654 | 0 | units, |
4655 | 0 | (value != 1) ? "s" : ""); |
4656 | | |
4657 | | /* |
4658 | | * Each nonzero field sets is_before for (only) the next one. This is a |
4659 | | * tad bizarre but it's how it worked before... |
4660 | | */ |
4661 | 0 | *is_before = (value < 0); |
4662 | 0 | *is_zero = false; |
4663 | 0 | return cp + strlen(cp); |
4664 | 0 | } |
4665 | | |
4666 | | /* Append a verbose-style interval field, but only if value isn't zero */ |
4667 | | static char * |
4668 | | AddVerboseIntPart(char *cp, int64 value, const char *units, |
4669 | | bool *is_zero, bool *is_before) |
4670 | 0 | { |
4671 | 0 | if (value == 0) |
4672 | 0 | return cp; |
4673 | | /* first nonzero value sets is_before */ |
4674 | 0 | if (*is_zero) |
4675 | 0 | { |
4676 | 0 | *is_before = (value < 0); |
4677 | 0 | value = i64abs(value); |
4678 | 0 | } |
4679 | 0 | else if (*is_before) |
4680 | 0 | value = -value; |
4681 | 0 | sprintf(cp, " %" PRId64 " %s%s", value, units, (value == 1) ? "" : "s"); |
4682 | 0 | *is_zero = false; |
4683 | 0 | return cp + strlen(cp); |
4684 | 0 | } |
4685 | | |
4686 | | |
4687 | | /* EncodeInterval() |
4688 | | * Interpret time structure as a delta time and convert to string. |
4689 | | * |
4690 | | * Support "traditional Postgres" and ISO-8601 styles. |
4691 | | * Actually, afaik ISO does not address time interval formatting, |
4692 | | * but this looks similar to the spec for absolute date/time. |
4693 | | * - thomas 1998-04-30 |
4694 | | * |
4695 | | * Actually, afaik, ISO 8601 does specify formats for "time |
4696 | | * intervals...[of the]...format with time-unit designators", which |
4697 | | * are pretty ugly. The format looks something like |
4698 | | * P1Y1M1DT1H1M1.12345S |
4699 | | * but useful for exchanging data with computers instead of humans. |
4700 | | * - ron 2003-07-14 |
4701 | | * |
4702 | | * And ISO's SQL 2008 standard specifies standards for |
4703 | | * "year-month literal"s (that look like '2-3') and |
4704 | | * "day-time literal"s (that look like ('4 5:6:7') |
4705 | | */ |
4706 | | void |
4707 | | EncodeInterval(struct pg_itm *itm, int style, char *str) |
4708 | 0 | { |
4709 | 0 | char *cp = str; |
4710 | 0 | int year = itm->tm_year; |
4711 | 0 | int mon = itm->tm_mon; |
4712 | 0 | int64 mday = itm->tm_mday; /* tm_mday could be INT_MIN */ |
4713 | 0 | int64 hour = itm->tm_hour; |
4714 | 0 | int min = itm->tm_min; |
4715 | 0 | int sec = itm->tm_sec; |
4716 | 0 | int fsec = itm->tm_usec; |
4717 | 0 | bool is_before = false; |
4718 | 0 | bool is_zero = true; |
4719 | | |
4720 | | /* |
4721 | | * The sign of year and month are guaranteed to match, since they are |
4722 | | * stored internally as "month". But we'll need to check for is_before and |
4723 | | * is_zero when determining the signs of day and hour/minute/seconds |
4724 | | * fields. |
4725 | | */ |
4726 | 0 | switch (style) |
4727 | 0 | { |
4728 | | /* SQL Standard interval format */ |
4729 | 0 | case INTSTYLE_SQL_STANDARD: |
4730 | 0 | { |
4731 | 0 | bool has_negative = year < 0 || mon < 0 || |
4732 | 0 | mday < 0 || hour < 0 || |
4733 | 0 | min < 0 || sec < 0 || fsec < 0; |
4734 | 0 | bool has_positive = year > 0 || mon > 0 || |
4735 | 0 | mday > 0 || hour > 0 || |
4736 | 0 | min > 0 || sec > 0 || fsec > 0; |
4737 | 0 | bool has_year_month = year != 0 || mon != 0; |
4738 | 0 | bool has_day_time = mday != 0 || hour != 0 || |
4739 | 0 | min != 0 || sec != 0 || fsec != 0; |
4740 | 0 | bool has_day = mday != 0; |
4741 | 0 | bool sql_standard_value = !(has_negative && has_positive) && |
4742 | 0 | !(has_year_month && has_day_time); |
4743 | | |
4744 | | /* |
4745 | | * SQL Standard wants only 1 "<sign>" preceding the whole |
4746 | | * interval ... but can't do that if mixed signs. |
4747 | | */ |
4748 | 0 | if (has_negative && sql_standard_value) |
4749 | 0 | { |
4750 | 0 | *cp++ = '-'; |
4751 | 0 | year = -year; |
4752 | 0 | mon = -mon; |
4753 | 0 | mday = -mday; |
4754 | 0 | hour = -hour; |
4755 | 0 | min = -min; |
4756 | 0 | sec = -sec; |
4757 | 0 | fsec = -fsec; |
4758 | 0 | } |
4759 | |
|
4760 | 0 | if (!has_negative && !has_positive) |
4761 | 0 | { |
4762 | 0 | sprintf(cp, "0"); |
4763 | 0 | } |
4764 | 0 | else if (!sql_standard_value) |
4765 | 0 | { |
4766 | | /* |
4767 | | * For non sql-standard interval values, force outputting |
4768 | | * the signs to avoid ambiguities with intervals with |
4769 | | * mixed sign components. |
4770 | | */ |
4771 | 0 | char year_sign = (year < 0 || mon < 0) ? '-' : '+'; |
4772 | 0 | char day_sign = (mday < 0) ? '-' : '+'; |
4773 | 0 | char sec_sign = (hour < 0 || min < 0 || |
4774 | 0 | sec < 0 || fsec < 0) ? '-' : '+'; |
4775 | |
|
4776 | 0 | sprintf(cp, "%c%d-%d %c%" PRId64 " %c%" PRId64 ":%02d:", |
4777 | 0 | year_sign, abs(year), abs(mon), |
4778 | 0 | day_sign, i64abs(mday), |
4779 | 0 | sec_sign, i64abs(hour), abs(min)); |
4780 | 0 | cp += strlen(cp); |
4781 | 0 | cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true); |
4782 | 0 | *cp = '\0'; |
4783 | 0 | } |
4784 | 0 | else if (has_year_month) |
4785 | 0 | { |
4786 | 0 | sprintf(cp, "%d-%d", year, mon); |
4787 | 0 | } |
4788 | 0 | else if (has_day) |
4789 | 0 | { |
4790 | 0 | sprintf(cp, "%" PRId64 " %" PRId64 ":%02d:", |
4791 | 0 | mday, hour, min); |
4792 | 0 | cp += strlen(cp); |
4793 | 0 | cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true); |
4794 | 0 | *cp = '\0'; |
4795 | 0 | } |
4796 | 0 | else |
4797 | 0 | { |
4798 | 0 | sprintf(cp, "%" PRId64 ":%02d:", hour, min); |
4799 | 0 | cp += strlen(cp); |
4800 | 0 | cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true); |
4801 | 0 | *cp = '\0'; |
4802 | 0 | } |
4803 | 0 | } |
4804 | 0 | break; |
4805 | | |
4806 | | /* ISO 8601 "time-intervals by duration only" */ |
4807 | 0 | case INTSTYLE_ISO_8601: |
4808 | | /* special-case zero to avoid printing nothing */ |
4809 | 0 | if (year == 0 && mon == 0 && mday == 0 && |
4810 | 0 | hour == 0 && min == 0 && sec == 0 && fsec == 0) |
4811 | 0 | { |
4812 | 0 | sprintf(cp, "PT0S"); |
4813 | 0 | break; |
4814 | 0 | } |
4815 | 0 | *cp++ = 'P'; |
4816 | 0 | cp = AddISO8601IntPart(cp, year, 'Y'); |
4817 | 0 | cp = AddISO8601IntPart(cp, mon, 'M'); |
4818 | 0 | cp = AddISO8601IntPart(cp, mday, 'D'); |
4819 | 0 | if (hour != 0 || min != 0 || sec != 0 || fsec != 0) |
4820 | 0 | *cp++ = 'T'; |
4821 | 0 | cp = AddISO8601IntPart(cp, hour, 'H'); |
4822 | 0 | cp = AddISO8601IntPart(cp, min, 'M'); |
4823 | 0 | if (sec != 0 || fsec != 0) |
4824 | 0 | { |
4825 | 0 | if (sec < 0 || fsec < 0) |
4826 | 0 | *cp++ = '-'; |
4827 | 0 | cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false); |
4828 | 0 | *cp++ = 'S'; |
4829 | 0 | *cp++ = '\0'; |
4830 | 0 | } |
4831 | 0 | break; |
4832 | | |
4833 | | /* Compatible with postgresql < 8.4 when DateStyle = 'iso' */ |
4834 | 0 | case INTSTYLE_POSTGRES: |
4835 | 0 | cp = AddPostgresIntPart(cp, year, "year", &is_zero, &is_before); |
4836 | | |
4837 | | /* |
4838 | | * Ideally we should spell out "month" like we do for "year" and |
4839 | | * "day". However, for backward compatibility, we can't easily |
4840 | | * fix this. bjm 2011-05-24 |
4841 | | */ |
4842 | 0 | cp = AddPostgresIntPart(cp, mon, "mon", &is_zero, &is_before); |
4843 | 0 | cp = AddPostgresIntPart(cp, mday, "day", &is_zero, &is_before); |
4844 | 0 | if (is_zero || hour != 0 || min != 0 || sec != 0 || fsec != 0) |
4845 | 0 | { |
4846 | 0 | bool minus = (hour < 0 || min < 0 || sec < 0 || fsec < 0); |
4847 | |
|
4848 | 0 | sprintf(cp, "%s%s%02" PRId64 ":%02d:", |
4849 | 0 | is_zero ? "" : " ", |
4850 | 0 | (minus ? "-" : (is_before ? "+" : "")), |
4851 | 0 | i64abs(hour), abs(min)); |
4852 | 0 | cp += strlen(cp); |
4853 | 0 | cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, true); |
4854 | 0 | *cp = '\0'; |
4855 | 0 | } |
4856 | 0 | break; |
4857 | | |
4858 | | /* Compatible with postgresql < 8.4 when DateStyle != 'iso' */ |
4859 | 0 | case INTSTYLE_POSTGRES_VERBOSE: |
4860 | 0 | default: |
4861 | 0 | strcpy(cp, "@"); |
4862 | 0 | cp++; |
4863 | 0 | cp = AddVerboseIntPart(cp, year, "year", &is_zero, &is_before); |
4864 | 0 | cp = AddVerboseIntPart(cp, mon, "mon", &is_zero, &is_before); |
4865 | 0 | cp = AddVerboseIntPart(cp, mday, "day", &is_zero, &is_before); |
4866 | 0 | cp = AddVerboseIntPart(cp, hour, "hour", &is_zero, &is_before); |
4867 | 0 | cp = AddVerboseIntPart(cp, min, "min", &is_zero, &is_before); |
4868 | 0 | if (sec != 0 || fsec != 0) |
4869 | 0 | { |
4870 | 0 | *cp++ = ' '; |
4871 | 0 | if (sec < 0 || (sec == 0 && fsec < 0)) |
4872 | 0 | { |
4873 | 0 | if (is_zero) |
4874 | 0 | is_before = true; |
4875 | 0 | else if (!is_before) |
4876 | 0 | *cp++ = '-'; |
4877 | 0 | } |
4878 | 0 | else if (is_before) |
4879 | 0 | *cp++ = '-'; |
4880 | 0 | cp = AppendSeconds(cp, sec, fsec, MAX_INTERVAL_PRECISION, false); |
4881 | | /* We output "ago", not negatives, so use abs(). */ |
4882 | 0 | sprintf(cp, " sec%s", |
4883 | 0 | (abs(sec) != 1 || fsec != 0) ? "s" : ""); |
4884 | 0 | is_zero = false; |
4885 | 0 | } |
4886 | | /* identically zero? then put in a unitless zero... */ |
4887 | 0 | if (is_zero) |
4888 | 0 | strcat(cp, " 0"); |
4889 | 0 | if (is_before) |
4890 | 0 | strcat(cp, " ago"); |
4891 | 0 | break; |
4892 | 0 | } |
4893 | 0 | } |
4894 | | |
4895 | | |
4896 | | /* |
4897 | | * We've been burnt by stupid errors in the ordering of the datetkn tables |
4898 | | * once too often. Arrange to check them during postmaster start. |
4899 | | */ |
4900 | | static bool |
4901 | | CheckDateTokenTable(const char *tablename, const datetkn *base, int nel) |
4902 | 0 | { |
4903 | 0 | bool ok = true; |
4904 | 0 | int i; |
4905 | |
|
4906 | 0 | for (i = 0; i < nel; i++) |
4907 | 0 | { |
4908 | | /* check for token strings that don't fit */ |
4909 | 0 | if (strlen(base[i].token) > TOKMAXLEN) |
4910 | 0 | { |
4911 | | /* %.*s is safe since all our tokens are ASCII */ |
4912 | 0 | elog(LOG, "token too long in %s table: \"%.*s\"", |
4913 | 0 | tablename, |
4914 | 0 | TOKMAXLEN + 1, base[i].token); |
4915 | 0 | ok = false; |
4916 | 0 | break; /* don't risk applying strcmp */ |
4917 | 0 | } |
4918 | | /* check for out of order */ |
4919 | 0 | if (i > 0 && |
4920 | 0 | strcmp(base[i - 1].token, base[i].token) >= 0) |
4921 | 0 | { |
4922 | 0 | elog(LOG, "ordering error in %s table: \"%s\" >= \"%s\"", |
4923 | 0 | tablename, |
4924 | 0 | base[i - 1].token, |
4925 | 0 | base[i].token); |
4926 | 0 | ok = false; |
4927 | 0 | } |
4928 | 0 | } |
4929 | 0 | return ok; |
4930 | 0 | } |
4931 | | |
4932 | | bool |
4933 | | CheckDateTokenTables(void) |
4934 | 0 | { |
4935 | 0 | bool ok = true; |
4936 | |
|
4937 | 0 | Assert(UNIX_EPOCH_JDATE == date2j(1970, 1, 1)); |
4938 | 0 | Assert(POSTGRES_EPOCH_JDATE == date2j(2000, 1, 1)); |
4939 | |
|
4940 | 0 | ok &= CheckDateTokenTable("datetktbl", datetktbl, szdatetktbl); |
4941 | 0 | ok &= CheckDateTokenTable("deltatktbl", deltatktbl, szdeltatktbl); |
4942 | 0 | return ok; |
4943 | 0 | } |
4944 | | |
4945 | | /* |
4946 | | * Common code for temporal prosupport functions: simplify, if possible, |
4947 | | * a call to a temporal type's length-coercion function. |
4948 | | * |
4949 | | * Types time, timetz, timestamp and timestamptz each have a range of allowed |
4950 | | * precisions. An unspecified precision is rigorously equivalent to the |
4951 | | * highest specifiable precision. We can replace the function call with a |
4952 | | * no-op RelabelType if it is coercing to the same or higher precision as the |
4953 | | * input is known to have. |
4954 | | * |
4955 | | * The input Node is always a FuncExpr, but to reduce the #include footprint |
4956 | | * of datetime.h, we declare it as Node *. |
4957 | | * |
4958 | | * Note: timestamp_scale throws an error when the typmod is out of range, but |
4959 | | * we can't get there from a cast: our typmodin will have caught it already. |
4960 | | */ |
4961 | | Node * |
4962 | | TemporalSimplify(int32 max_precis, Node *node) |
4963 | 0 | { |
4964 | 0 | FuncExpr *expr = castNode(FuncExpr, node); |
4965 | 0 | Node *ret = NULL; |
4966 | 0 | Node *typmod; |
4967 | |
|
4968 | 0 | Assert(list_length(expr->args) >= 2); |
4969 | |
|
4970 | 0 | typmod = (Node *) lsecond(expr->args); |
4971 | |
|
4972 | 0 | if (IsA(typmod, Const) && !((Const *) typmod)->constisnull) |
4973 | 0 | { |
4974 | 0 | Node *source = (Node *) linitial(expr->args); |
4975 | 0 | int32 old_precis = exprTypmod(source); |
4976 | 0 | int32 new_precis = DatumGetInt32(((Const *) typmod)->constvalue); |
4977 | |
|
4978 | 0 | if (new_precis < 0 || new_precis == max_precis || |
4979 | 0 | (old_precis >= 0 && new_precis >= old_precis)) |
4980 | 0 | ret = relabel_to_typmod(source, new_precis); |
4981 | 0 | } |
4982 | |
|
4983 | 0 | return ret; |
4984 | 0 | } |
4985 | | |
4986 | | /* |
4987 | | * This function gets called during timezone config file load or reload |
4988 | | * to create the final array of timezone tokens. The argument array |
4989 | | * is already sorted in name order. |
4990 | | * |
4991 | | * The result is a TimeZoneAbbrevTable (which must be a single guc_malloc'd |
4992 | | * chunk) or NULL on alloc failure. No other error conditions are defined. |
4993 | | */ |
4994 | | TimeZoneAbbrevTable * |
4995 | | ConvertTimeZoneAbbrevs(struct tzEntry *abbrevs, int n) |
4996 | 0 | { |
4997 | 0 | TimeZoneAbbrevTable *tbl; |
4998 | 0 | Size tbl_size; |
4999 | 0 | int i; |
5000 | | |
5001 | | /* Space for fixed fields and datetkn array */ |
5002 | 0 | tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) + |
5003 | 0 | n * sizeof(datetkn); |
5004 | 0 | tbl_size = MAXALIGN(tbl_size); |
5005 | | /* Count up space for dynamic abbreviations */ |
5006 | 0 | for (i = 0; i < n; i++) |
5007 | 0 | { |
5008 | 0 | struct tzEntry *abbr = abbrevs + i; |
5009 | |
|
5010 | 0 | if (abbr->zone != NULL) |
5011 | 0 | { |
5012 | 0 | Size dsize; |
5013 | |
|
5014 | 0 | dsize = offsetof(DynamicZoneAbbrev, zone) + |
5015 | 0 | strlen(abbr->zone) + 1; |
5016 | 0 | tbl_size += MAXALIGN(dsize); |
5017 | 0 | } |
5018 | 0 | } |
5019 | | |
5020 | | /* Alloc the result ... */ |
5021 | 0 | tbl = guc_malloc(LOG, tbl_size); |
5022 | 0 | if (!tbl) |
5023 | 0 | return NULL; |
5024 | | |
5025 | | /* ... and fill it in */ |
5026 | 0 | tbl->tblsize = tbl_size; |
5027 | 0 | tbl->numabbrevs = n; |
5028 | | /* in this loop, tbl_size reprises the space calculation above */ |
5029 | 0 | tbl_size = offsetof(TimeZoneAbbrevTable, abbrevs) + |
5030 | 0 | n * sizeof(datetkn); |
5031 | 0 | tbl_size = MAXALIGN(tbl_size); |
5032 | 0 | for (i = 0; i < n; i++) |
5033 | 0 | { |
5034 | 0 | struct tzEntry *abbr = abbrevs + i; |
5035 | 0 | datetkn *dtoken = tbl->abbrevs + i; |
5036 | | |
5037 | | /* use strlcpy to truncate name if necessary */ |
5038 | 0 | strlcpy(dtoken->token, abbr->abbrev, TOKMAXLEN + 1); |
5039 | 0 | if (abbr->zone != NULL) |
5040 | 0 | { |
5041 | | /* Allocate a DynamicZoneAbbrev for this abbreviation */ |
5042 | 0 | DynamicZoneAbbrev *dtza; |
5043 | 0 | Size dsize; |
5044 | |
|
5045 | 0 | dtza = (DynamicZoneAbbrev *) ((char *) tbl + tbl_size); |
5046 | 0 | dtza->tz = NULL; |
5047 | 0 | strcpy(dtza->zone, abbr->zone); |
5048 | |
|
5049 | 0 | dtoken->type = DYNTZ; |
5050 | | /* value is offset from table start to DynamicZoneAbbrev */ |
5051 | 0 | dtoken->value = (int32) tbl_size; |
5052 | |
|
5053 | 0 | dsize = offsetof(DynamicZoneAbbrev, zone) + |
5054 | 0 | strlen(abbr->zone) + 1; |
5055 | 0 | tbl_size += MAXALIGN(dsize); |
5056 | 0 | } |
5057 | 0 | else |
5058 | 0 | { |
5059 | 0 | dtoken->type = abbr->is_dst ? DTZ : TZ; |
5060 | 0 | dtoken->value = abbr->offset; |
5061 | 0 | } |
5062 | 0 | } |
5063 | | |
5064 | | /* Assert the two loops above agreed on size calculations */ |
5065 | 0 | Assert(tbl->tblsize == tbl_size); |
5066 | | |
5067 | | /* Check the ordering, if testing */ |
5068 | 0 | Assert(CheckDateTokenTable("timezone abbreviations", tbl->abbrevs, n)); |
5069 | |
|
5070 | 0 | return tbl; |
5071 | 0 | } |
5072 | | |
5073 | | /* |
5074 | | * Install a TimeZoneAbbrevTable as the active table. |
5075 | | * |
5076 | | * Caller is responsible that the passed table doesn't go away while in use. |
5077 | | */ |
5078 | | void |
5079 | | InstallTimeZoneAbbrevs(TimeZoneAbbrevTable *tbl) |
5080 | 0 | { |
5081 | 0 | zoneabbrevtbl = tbl; |
5082 | | /* reset tzabbrevcache, which may contain results from old table */ |
5083 | 0 | memset(tzabbrevcache, 0, sizeof(tzabbrevcache)); |
5084 | 0 | } |
5085 | | |
5086 | | /* |
5087 | | * Helper subroutine to locate pg_tz timezone for a dynamic abbreviation. |
5088 | | * |
5089 | | * On failure, returns NULL and fills *extra for a DTERR_BAD_ZONE_ABBREV error. |
5090 | | */ |
5091 | | static pg_tz * |
5092 | | FetchDynamicTimeZone(TimeZoneAbbrevTable *tbl, const datetkn *tp, |
5093 | | DateTimeErrorExtra *extra) |
5094 | 0 | { |
5095 | 0 | DynamicZoneAbbrev *dtza; |
5096 | | |
5097 | | /* Just some sanity checks to prevent indexing off into nowhere */ |
5098 | 0 | Assert(tp->type == DYNTZ); |
5099 | 0 | Assert(tp->value > 0 && tp->value < tbl->tblsize); |
5100 | |
|
5101 | 0 | dtza = (DynamicZoneAbbrev *) ((char *) tbl + tp->value); |
5102 | | |
5103 | | /* Look up the underlying zone if we haven't already */ |
5104 | 0 | if (dtza->tz == NULL) |
5105 | 0 | { |
5106 | 0 | dtza->tz = pg_tzset(dtza->zone); |
5107 | 0 | if (dtza->tz == NULL) |
5108 | 0 | { |
5109 | | /* Ooops, bogus zone name in config file entry */ |
5110 | 0 | extra->dtee_timezone = dtza->zone; |
5111 | 0 | extra->dtee_abbrev = tp->token; |
5112 | 0 | } |
5113 | 0 | } |
5114 | 0 | return dtza->tz; |
5115 | 0 | } |
5116 | | |
5117 | | |
5118 | | /* |
5119 | | * This set-returning function reads all the time zone abbreviations |
5120 | | * defined by the IANA data for the current timezone setting, |
5121 | | * and returns a set of (abbrev, utc_offset, is_dst). |
5122 | | */ |
5123 | | Datum |
5124 | | pg_timezone_abbrevs_zone(PG_FUNCTION_ARGS) |
5125 | 0 | { |
5126 | 0 | FuncCallContext *funcctx; |
5127 | 0 | int *pindex; |
5128 | 0 | Datum result; |
5129 | 0 | HeapTuple tuple; |
5130 | 0 | Datum values[3]; |
5131 | 0 | bool nulls[3] = {0}; |
5132 | 0 | TimestampTz now = GetCurrentTransactionStartTimestamp(); |
5133 | 0 | pg_time_t t = timestamptz_to_time_t(now); |
5134 | 0 | const char *abbrev; |
5135 | 0 | long int gmtoff; |
5136 | 0 | int isdst; |
5137 | 0 | struct pg_itm_in itm_in; |
5138 | 0 | Interval *resInterval; |
5139 | | |
5140 | | /* stuff done only on the first call of the function */ |
5141 | 0 | if (SRF_IS_FIRSTCALL()) |
5142 | 0 | { |
5143 | 0 | TupleDesc tupdesc; |
5144 | 0 | MemoryContext oldcontext; |
5145 | | |
5146 | | /* create a function context for cross-call persistence */ |
5147 | 0 | funcctx = SRF_FIRSTCALL_INIT(); |
5148 | | |
5149 | | /* |
5150 | | * switch to memory context appropriate for multiple function calls |
5151 | | */ |
5152 | 0 | oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx); |
5153 | | |
5154 | | /* allocate memory for user context */ |
5155 | 0 | pindex = (int *) palloc(sizeof(int)); |
5156 | 0 | *pindex = 0; |
5157 | 0 | funcctx->user_fctx = pindex; |
5158 | |
|
5159 | 0 | if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE) |
5160 | 0 | elog(ERROR, "return type must be a row type"); |
5161 | 0 | funcctx->tuple_desc = tupdesc; |
5162 | |
|
5163 | 0 | MemoryContextSwitchTo(oldcontext); |
5164 | 0 | } |
5165 | | |
5166 | | /* stuff done on every call of the function */ |
5167 | 0 | funcctx = SRF_PERCALL_SETUP(); |
5168 | 0 | pindex = (int *) funcctx->user_fctx; |
5169 | |
|
5170 | 0 | while ((abbrev = pg_get_next_timezone_abbrev(pindex, |
5171 | 0 | session_timezone)) != NULL) |
5172 | 0 | { |
5173 | | /* Ignore abbreviations that aren't all-alphabetic */ |
5174 | 0 | if (strspn(abbrev, "ABCDEFGHIJKLMNOPQRSTUVWXYZ") != strlen(abbrev)) |
5175 | 0 | continue; |
5176 | | |
5177 | | /* Determine the current meaning of the abbrev */ |
5178 | 0 | if (!pg_interpret_timezone_abbrev(abbrev, |
5179 | 0 | &t, |
5180 | 0 | &gmtoff, |
5181 | 0 | &isdst, |
5182 | 0 | session_timezone)) |
5183 | 0 | continue; /* hm, not actually used in this zone? */ |
5184 | | |
5185 | 0 | values[0] = CStringGetTextDatum(abbrev); |
5186 | | |
5187 | | /* Convert offset (in seconds) to an interval; can't overflow */ |
5188 | 0 | MemSet(&itm_in, 0, sizeof(struct pg_itm_in)); |
5189 | 0 | itm_in.tm_usec = (int64) gmtoff * USECS_PER_SEC; |
5190 | 0 | resInterval = (Interval *) palloc(sizeof(Interval)); |
5191 | 0 | (void) itmin2interval(&itm_in, resInterval); |
5192 | 0 | values[1] = IntervalPGetDatum(resInterval); |
5193 | |
|
5194 | 0 | values[2] = BoolGetDatum(isdst); |
5195 | |
|
5196 | 0 | tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls); |
5197 | 0 | result = HeapTupleGetDatum(tuple); |
5198 | |
|
5199 | 0 | SRF_RETURN_NEXT(funcctx, result); |
5200 | 0 | } |
5201 | | |
5202 | 0 | SRF_RETURN_DONE(funcctx); |
5203 | 0 | } |
5204 | | |
5205 | | /* |
5206 | | * This set-returning function reads all the time zone abbreviations |
5207 | | * defined by the timezone_abbreviations setting, |
5208 | | * and returns a set of (abbrev, utc_offset, is_dst). |
5209 | | */ |
5210 | | Datum |
5211 | | pg_timezone_abbrevs_abbrevs(PG_FUNCTION_ARGS) |
5212 | 0 | { |
5213 | 0 | FuncCallContext *funcctx; |
5214 | 0 | int *pindex; |
5215 | 0 | Datum result; |
5216 | 0 | HeapTuple tuple; |
5217 | 0 | Datum values[3]; |
5218 | 0 | bool nulls[3] = {0}; |
5219 | 0 | const datetkn *tp; |
5220 | 0 | char buffer[TOKMAXLEN + 1]; |
5221 | 0 | int gmtoffset; |
5222 | 0 | bool is_dst; |
5223 | 0 | unsigned char *p; |
5224 | 0 | struct pg_itm_in itm_in; |
5225 | 0 | Interval *resInterval; |
5226 | | |
5227 | | /* stuff done only on the first call of the function */ |
5228 | 0 | if (SRF_IS_FIRSTCALL()) |
5229 | 0 | { |
5230 | 0 | TupleDesc tupdesc; |
5231 | 0 | MemoryContext oldcontext; |
5232 | | |
5233 | | /* create a function context for cross-call persistence */ |
5234 | 0 | funcctx = SRF_FIRSTCALL_INIT(); |
5235 | | |
5236 | | /* |
5237 | | * switch to memory context appropriate for multiple function calls |
5238 | | */ |
5239 | 0 | oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx); |
5240 | | |
5241 | | /* allocate memory for user context */ |
5242 | 0 | pindex = (int *) palloc(sizeof(int)); |
5243 | 0 | *pindex = 0; |
5244 | 0 | funcctx->user_fctx = pindex; |
5245 | |
|
5246 | 0 | if (get_call_result_type(fcinfo, NULL, &tupdesc) != TYPEFUNC_COMPOSITE) |
5247 | 0 | elog(ERROR, "return type must be a row type"); |
5248 | 0 | funcctx->tuple_desc = tupdesc; |
5249 | |
|
5250 | 0 | MemoryContextSwitchTo(oldcontext); |
5251 | 0 | } |
5252 | | |
5253 | | /* stuff done on every call of the function */ |
5254 | 0 | funcctx = SRF_PERCALL_SETUP(); |
5255 | 0 | pindex = (int *) funcctx->user_fctx; |
5256 | |
|
5257 | 0 | if (zoneabbrevtbl == NULL || |
5258 | 0 | *pindex >= zoneabbrevtbl->numabbrevs) |
5259 | 0 | SRF_RETURN_DONE(funcctx); |
5260 | | |
5261 | 0 | tp = zoneabbrevtbl->abbrevs + *pindex; |
5262 | |
|
5263 | 0 | switch (tp->type) |
5264 | 0 | { |
5265 | 0 | case TZ: |
5266 | 0 | gmtoffset = tp->value; |
5267 | 0 | is_dst = false; |
5268 | 0 | break; |
5269 | 0 | case DTZ: |
5270 | 0 | gmtoffset = tp->value; |
5271 | 0 | is_dst = true; |
5272 | 0 | break; |
5273 | 0 | case DYNTZ: |
5274 | 0 | { |
5275 | | /* Determine the current meaning of the abbrev */ |
5276 | 0 | pg_tz *tzp; |
5277 | 0 | DateTimeErrorExtra extra; |
5278 | 0 | TimestampTz now; |
5279 | 0 | int isdst; |
5280 | |
|
5281 | 0 | tzp = FetchDynamicTimeZone(zoneabbrevtbl, tp, &extra); |
5282 | 0 | if (tzp == NULL) |
5283 | 0 | DateTimeParseError(DTERR_BAD_ZONE_ABBREV, &extra, |
5284 | 0 | NULL, NULL, NULL); |
5285 | 0 | now = GetCurrentTransactionStartTimestamp(); |
5286 | 0 | gmtoffset = -DetermineTimeZoneAbbrevOffsetTS(now, |
5287 | 0 | tp->token, |
5288 | 0 | tzp, |
5289 | 0 | &isdst); |
5290 | 0 | is_dst = (bool) isdst; |
5291 | 0 | break; |
5292 | 0 | } |
5293 | 0 | default: |
5294 | 0 | elog(ERROR, "unrecognized timezone type %d", (int) tp->type); |
5295 | 0 | gmtoffset = 0; /* keep compiler quiet */ |
5296 | 0 | is_dst = false; |
5297 | 0 | break; |
5298 | 0 | } |
5299 | | |
5300 | | /* |
5301 | | * Convert name to text, using upcasing conversion that is the inverse of |
5302 | | * what ParseDateTime() uses. |
5303 | | */ |
5304 | 0 | strlcpy(buffer, tp->token, sizeof(buffer)); |
5305 | 0 | for (p = (unsigned char *) buffer; *p; p++) |
5306 | 0 | *p = pg_toupper(*p); |
5307 | |
|
5308 | 0 | values[0] = CStringGetTextDatum(buffer); |
5309 | | |
5310 | | /* Convert offset (in seconds) to an interval; can't overflow */ |
5311 | 0 | MemSet(&itm_in, 0, sizeof(struct pg_itm_in)); |
5312 | 0 | itm_in.tm_usec = (int64) gmtoffset * USECS_PER_SEC; |
5313 | 0 | resInterval = (Interval *) palloc(sizeof(Interval)); |
5314 | 0 | (void) itmin2interval(&itm_in, resInterval); |
5315 | 0 | values[1] = IntervalPGetDatum(resInterval); |
5316 | |
|
5317 | 0 | values[2] = BoolGetDatum(is_dst); |
5318 | |
|
5319 | 0 | (*pindex)++; |
5320 | |
|
5321 | 0 | tuple = heap_form_tuple(funcctx->tuple_desc, values, nulls); |
5322 | 0 | result = HeapTupleGetDatum(tuple); |
5323 | |
|
5324 | 0 | SRF_RETURN_NEXT(funcctx, result); |
5325 | 0 | } |
5326 | | |
5327 | | /* |
5328 | | * This set-returning function reads all the available full time zones |
5329 | | * and returns a set of (name, abbrev, utc_offset, is_dst). |
5330 | | */ |
5331 | | Datum |
5332 | | pg_timezone_names(PG_FUNCTION_ARGS) |
5333 | 0 | { |
5334 | 0 | ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo; |
5335 | 0 | pg_tzenum *tzenum; |
5336 | 0 | pg_tz *tz; |
5337 | 0 | Datum values[4]; |
5338 | 0 | bool nulls[4] = {0}; |
5339 | 0 | int tzoff; |
5340 | 0 | struct pg_tm tm; |
5341 | 0 | fsec_t fsec; |
5342 | 0 | const char *tzn; |
5343 | 0 | Interval *resInterval; |
5344 | 0 | struct pg_itm_in itm_in; |
5345 | |
|
5346 | 0 | InitMaterializedSRF(fcinfo, 0); |
5347 | | |
5348 | | /* initialize timezone scanning code */ |
5349 | 0 | tzenum = pg_tzenumerate_start(); |
5350 | | |
5351 | | /* search for another zone to display */ |
5352 | 0 | for (;;) |
5353 | 0 | { |
5354 | 0 | tz = pg_tzenumerate_next(tzenum); |
5355 | 0 | if (!tz) |
5356 | 0 | break; |
5357 | | |
5358 | | /* Convert now() to local time in this zone */ |
5359 | 0 | if (timestamp2tm(GetCurrentTransactionStartTimestamp(), |
5360 | 0 | &tzoff, &tm, &fsec, &tzn, tz) != 0) |
5361 | 0 | continue; /* ignore if conversion fails */ |
5362 | | |
5363 | | /* |
5364 | | * IANA's rather silly "Factory" time zone used to emit ridiculously |
5365 | | * long "abbreviations" such as "Local time zone must be set--see zic |
5366 | | * manual page" or "Local time zone must be set--use tzsetup". While |
5367 | | * modern versions of tzdb emit the much saner "-00", it seems some |
5368 | | * benighted packagers are hacking the IANA data so that it continues |
5369 | | * to produce these strings. To prevent producing a weirdly wide |
5370 | | * abbrev column, reject ridiculously long abbreviations. |
5371 | | */ |
5372 | 0 | if (tzn && strlen(tzn) > 31) |
5373 | 0 | continue; |
5374 | | |
5375 | 0 | values[0] = CStringGetTextDatum(pg_get_timezone_name(tz)); |
5376 | 0 | values[1] = CStringGetTextDatum(tzn ? tzn : ""); |
5377 | | |
5378 | | /* Convert tzoff to an interval; can't overflow */ |
5379 | 0 | MemSet(&itm_in, 0, sizeof(struct pg_itm_in)); |
5380 | 0 | itm_in.tm_usec = (int64) -tzoff * USECS_PER_SEC; |
5381 | 0 | resInterval = (Interval *) palloc(sizeof(Interval)); |
5382 | 0 | (void) itmin2interval(&itm_in, resInterval); |
5383 | 0 | values[2] = IntervalPGetDatum(resInterval); |
5384 | |
|
5385 | 0 | values[3] = BoolGetDatum(tm.tm_isdst > 0); |
5386 | |
|
5387 | 0 | tuplestore_putvalues(rsinfo->setResult, rsinfo->setDesc, values, nulls); |
5388 | 0 | } |
5389 | |
|
5390 | 0 | pg_tzenumerate_end(tzenum); |
5391 | 0 | return (Datum) 0; |
5392 | 0 | } |