/src/icu/source/i18n/number_compact.cpp
Line | Count | Source (jump to first uncovered line) |
1 | | // © 2017 and later: Unicode, Inc. and others. |
2 | | // License & terms of use: http://www.unicode.org/copyright.html |
3 | | |
4 | | #include "unicode/utypes.h" |
5 | | |
6 | | #if !UCONFIG_NO_FORMATTING |
7 | | |
8 | | #include "unicode/ustring.h" |
9 | | #include "unicode/ures.h" |
10 | | #include "cstring.h" |
11 | | #include "charstr.h" |
12 | | #include "resource.h" |
13 | | #include "number_compact.h" |
14 | | #include "number_microprops.h" |
15 | | #include "uresimp.h" |
16 | | |
17 | | using namespace icu; |
18 | | using namespace icu::number; |
19 | | using namespace icu::number::impl; |
20 | | |
21 | | namespace { |
22 | | |
23 | | // A dummy object used when a "0" compact decimal entry is encountered. This is necessary |
24 | | // in order to prevent falling back to root. Object equality ("==") is intended. |
25 | | const UChar *USE_FALLBACK = u"<USE FALLBACK>"; |
26 | | |
27 | | /** Produces a string like "NumberElements/latn/patternsShort/decimalFormat". */ |
28 | | void getResourceBundleKey(const char *nsName, CompactStyle compactStyle, CompactType compactType, |
29 | 0 | CharString &sb, UErrorCode &status) { |
30 | 0 | sb.clear(); |
31 | 0 | sb.append("NumberElements/", status); |
32 | 0 | sb.append(nsName, status); |
33 | 0 | sb.append(compactStyle == CompactStyle::UNUM_SHORT ? "/patternsShort" : "/patternsLong", status); |
34 | 0 | sb.append(compactType == CompactType::TYPE_DECIMAL ? "/decimalFormat" : "/currencyFormat", status); |
35 | 0 | } |
36 | | |
37 | 0 | int32_t getIndex(int32_t magnitude, StandardPlural::Form plural) { |
38 | 0 | return magnitude * StandardPlural::COUNT + plural; |
39 | 0 | } |
40 | | |
41 | 0 | int32_t countZeros(const UChar *patternString, int32_t patternLength) { |
42 | | // NOTE: This strategy for computing the number of zeros is a hack for efficiency. |
43 | | // It could break if there are any 0s that aren't part of the main pattern. |
44 | 0 | int32_t numZeros = 0; |
45 | 0 | for (int32_t i = 0; i < patternLength; i++) { |
46 | 0 | if (patternString[i] == u'0') { |
47 | 0 | numZeros++; |
48 | 0 | } else if (numZeros > 0) { |
49 | 0 | break; // zeros should always be contiguous |
50 | 0 | } |
51 | 0 | } |
52 | 0 | return numZeros; |
53 | 0 | } |
54 | | |
55 | | } // namespace |
56 | | |
57 | | // NOTE: patterns and multipliers both get zero-initialized. |
58 | 0 | CompactData::CompactData() : patterns(), multipliers(), largestMagnitude(0), isEmpty(true) { |
59 | 0 | } |
60 | | |
61 | | void CompactData::populate(const Locale &locale, const char *nsName, CompactStyle compactStyle, |
62 | 0 | CompactType compactType, UErrorCode &status) { |
63 | 0 | CompactDataSink sink(*this); |
64 | 0 | LocalUResourceBundlePointer rb(ures_open(nullptr, locale.getName(), &status)); |
65 | 0 | if (U_FAILURE(status)) { return; } |
66 | | |
67 | 0 | bool nsIsLatn = strcmp(nsName, "latn") == 0; |
68 | 0 | bool compactIsShort = compactStyle == CompactStyle::UNUM_SHORT; |
69 | | |
70 | | // Fall back to latn numbering system and/or short compact style. |
71 | 0 | CharString resourceKey; |
72 | 0 | getResourceBundleKey(nsName, compactStyle, compactType, resourceKey, status); |
73 | 0 | UErrorCode localStatus = U_ZERO_ERROR; |
74 | 0 | ures_getAllItemsWithFallback(rb.getAlias(), resourceKey.data(), sink, localStatus); |
75 | 0 | if (isEmpty && !nsIsLatn) { |
76 | 0 | getResourceBundleKey("latn", compactStyle, compactType, resourceKey, status); |
77 | 0 | localStatus = U_ZERO_ERROR; |
78 | 0 | ures_getAllItemsWithFallback(rb.getAlias(), resourceKey.data(), sink, localStatus); |
79 | 0 | } |
80 | 0 | if (isEmpty && !compactIsShort) { |
81 | 0 | getResourceBundleKey(nsName, CompactStyle::UNUM_SHORT, compactType, resourceKey, status); |
82 | 0 | localStatus = U_ZERO_ERROR; |
83 | 0 | ures_getAllItemsWithFallback(rb.getAlias(), resourceKey.data(), sink, localStatus); |
84 | 0 | } |
85 | 0 | if (isEmpty && !nsIsLatn && !compactIsShort) { |
86 | 0 | getResourceBundleKey("latn", CompactStyle::UNUM_SHORT, compactType, resourceKey, status); |
87 | 0 | localStatus = U_ZERO_ERROR; |
88 | 0 | ures_getAllItemsWithFallback(rb.getAlias(), resourceKey.data(), sink, localStatus); |
89 | 0 | } |
90 | | |
91 | | // The last fallback should be guaranteed to return data. |
92 | 0 | if (isEmpty) { |
93 | 0 | status = U_INTERNAL_PROGRAM_ERROR; |
94 | 0 | } |
95 | 0 | } |
96 | | |
97 | 0 | int32_t CompactData::getMultiplier(int32_t magnitude) const { |
98 | 0 | if (magnitude < 0) { |
99 | 0 | return 0; |
100 | 0 | } |
101 | 0 | if (magnitude > largestMagnitude) { |
102 | 0 | magnitude = largestMagnitude; |
103 | 0 | } |
104 | 0 | return multipliers[magnitude]; |
105 | 0 | } |
106 | | |
107 | | const UChar *CompactData::getPattern( |
108 | | int32_t magnitude, |
109 | | const PluralRules *rules, |
110 | 0 | const DecimalQuantity &dq) const { |
111 | 0 | if (magnitude < 0) { |
112 | 0 | return nullptr; |
113 | 0 | } |
114 | 0 | if (magnitude > largestMagnitude) { |
115 | 0 | magnitude = largestMagnitude; |
116 | 0 | } |
117 | 0 | const UChar *patternString = nullptr; |
118 | 0 | if (dq.hasIntegerValue()) { |
119 | 0 | int64_t i = dq.toLong(true); |
120 | 0 | if (i == 0) { |
121 | 0 | patternString = patterns[getIndex(magnitude, StandardPlural::Form::EQ_0)]; |
122 | 0 | } else if (i == 1) { |
123 | 0 | patternString = patterns[getIndex(magnitude, StandardPlural::Form::EQ_1)]; |
124 | 0 | } |
125 | 0 | if (patternString != nullptr) { |
126 | 0 | return patternString; |
127 | 0 | } |
128 | 0 | } |
129 | 0 | StandardPlural::Form plural = utils::getStandardPlural(rules, dq); |
130 | 0 | patternString = patterns[getIndex(magnitude, plural)]; |
131 | 0 | if (patternString == nullptr && plural != StandardPlural::OTHER) { |
132 | | // Fall back to "other" plural variant |
133 | 0 | patternString = patterns[getIndex(magnitude, StandardPlural::OTHER)]; |
134 | 0 | } |
135 | 0 | if (patternString == USE_FALLBACK) { // == is intended |
136 | | // Return null if USE_FALLBACK is present |
137 | 0 | patternString = nullptr; |
138 | 0 | } |
139 | 0 | return patternString; |
140 | 0 | } |
141 | | |
142 | 0 | void CompactData::getUniquePatterns(UVector &output, UErrorCode &status) const { |
143 | 0 | U_ASSERT(output.isEmpty()); |
144 | | // NOTE: In C++, this is done more manually with a UVector. |
145 | | // In Java, we can take advantage of JDK HashSet. |
146 | 0 | for (auto pattern : patterns) { |
147 | 0 | if (pattern == nullptr || pattern == USE_FALLBACK) { |
148 | 0 | continue; |
149 | 0 | } |
150 | | |
151 | | // Insert pattern into the UVector if the UVector does not already contain the pattern. |
152 | | // Search the UVector from the end since identical patterns are likely to be adjacent. |
153 | 0 | for (int32_t i = output.size() - 1; i >= 0; i--) { |
154 | 0 | if (u_strcmp(pattern, static_cast<const UChar *>(output[i])) == 0) { |
155 | 0 | goto continue_outer; |
156 | 0 | } |
157 | 0 | } |
158 | | |
159 | | // The string was not found; add it to the UVector. |
160 | | // ANDY: This requires a const_cast. Why? |
161 | 0 | output.addElementX(const_cast<UChar *>(pattern), status); |
162 | |
|
163 | 0 | continue_outer: |
164 | 0 | continue; |
165 | 0 | } |
166 | 0 | } |
167 | | |
168 | | void CompactData::CompactDataSink::put(const char *key, ResourceValue &value, UBool /*noFallback*/, |
169 | 0 | UErrorCode &status) { |
170 | | // traverse into the table of powers of ten |
171 | 0 | ResourceTable powersOfTenTable = value.getTable(status); |
172 | 0 | if (U_FAILURE(status)) { return; } |
173 | 0 | for (int i3 = 0; powersOfTenTable.getKeyAndValue(i3, key, value); ++i3) { |
174 | | |
175 | | // Assumes that the keys are always of the form "10000" where the magnitude is the |
176 | | // length of the key minus one. We expect magnitudes to be less than MAX_DIGITS. |
177 | 0 | auto magnitude = static_cast<int8_t> (strlen(key) - 1); |
178 | 0 | int8_t multiplier = data.multipliers[magnitude]; |
179 | 0 | U_ASSERT(magnitude < COMPACT_MAX_DIGITS); |
180 | | |
181 | | // Iterate over the plural variants ("one", "other", etc) |
182 | 0 | ResourceTable pluralVariantsTable = value.getTable(status); |
183 | 0 | if (U_FAILURE(status)) { return; } |
184 | 0 | for (int i4 = 0; pluralVariantsTable.getKeyAndValue(i4, key, value); ++i4) { |
185 | | // Skip this magnitude/plural if we already have it from a child locale. |
186 | | // Note: This also skips USE_FALLBACK entries. |
187 | 0 | StandardPlural::Form plural = StandardPlural::fromString(key, status); |
188 | 0 | if (U_FAILURE(status)) { return; } |
189 | 0 | if (data.patterns[getIndex(magnitude, plural)] != nullptr) { |
190 | 0 | continue; |
191 | 0 | } |
192 | | |
193 | | // The value "0" means that we need to use the default pattern and not fall back |
194 | | // to parent locales. Example locale where this is relevant: 'it'. |
195 | 0 | int32_t patternLength; |
196 | 0 | const UChar *patternString = value.getString(patternLength, status); |
197 | 0 | if (U_FAILURE(status)) { return; } |
198 | 0 | if (u_strcmp(patternString, u"0") == 0) { |
199 | 0 | patternString = USE_FALLBACK; |
200 | 0 | patternLength = 0; |
201 | 0 | } |
202 | | |
203 | | // Save the pattern string. We will parse it lazily. |
204 | 0 | data.patterns[getIndex(magnitude, plural)] = patternString; |
205 | | |
206 | | // If necessary, compute the multiplier: the difference between the magnitude |
207 | | // and the number of zeros in the pattern. |
208 | 0 | if (multiplier == 0) { |
209 | 0 | int32_t numZeros = countZeros(patternString, patternLength); |
210 | 0 | if (numZeros > 0) { // numZeros==0 in certain cases, like Somali "Kun" |
211 | 0 | multiplier = static_cast<int8_t> (numZeros - magnitude - 1); |
212 | 0 | } |
213 | 0 | } |
214 | 0 | } |
215 | | |
216 | | // Save the multiplier. |
217 | 0 | if (data.multipliers[magnitude] == 0) { |
218 | 0 | data.multipliers[magnitude] = multiplier; |
219 | 0 | if (magnitude > data.largestMagnitude) { |
220 | 0 | data.largestMagnitude = magnitude; |
221 | 0 | } |
222 | 0 | data.isEmpty = false; |
223 | 0 | } else { |
224 | 0 | U_ASSERT(data.multipliers[magnitude] == multiplier); |
225 | 0 | } |
226 | 0 | } |
227 | 0 | } |
228 | | |
229 | | /////////////////////////////////////////////////////////// |
230 | | /// END OF CompactData.java; BEGIN CompactNotation.java /// |
231 | | /////////////////////////////////////////////////////////// |
232 | | |
233 | | CompactHandler::CompactHandler( |
234 | | CompactStyle compactStyle, |
235 | | const Locale &locale, |
236 | | const char *nsName, |
237 | | CompactType compactType, |
238 | | const PluralRules *rules, |
239 | | MutablePatternModifier *buildReference, |
240 | | bool safe, |
241 | | const MicroPropsGenerator *parent, |
242 | | UErrorCode &status) |
243 | 0 | : rules(rules), parent(parent), safe(safe) { |
244 | 0 | data.populate(locale, nsName, compactStyle, compactType, status); |
245 | 0 | if (safe) { |
246 | | // Safe code path |
247 | 0 | precomputeAllModifiers(*buildReference, status); |
248 | 0 | } else { |
249 | | // Unsafe code path |
250 | | // Store the MutablePatternModifier reference. |
251 | 0 | unsafePatternModifier = buildReference; |
252 | 0 | } |
253 | 0 | } |
254 | | |
255 | 0 | CompactHandler::~CompactHandler() { |
256 | 0 | for (int32_t i = 0; i < precomputedModsLength; i++) { |
257 | 0 | delete precomputedMods[i].mod; |
258 | 0 | } |
259 | 0 | } |
260 | | |
261 | 0 | void CompactHandler::precomputeAllModifiers(MutablePatternModifier &buildReference, UErrorCode &status) { |
262 | 0 | if (U_FAILURE(status)) { return; } |
263 | | |
264 | | // Initial capacity of 12 for 0K, 00K, 000K, ...M, ...B, and ...T |
265 | 0 | UVector allPatterns(12, status); |
266 | 0 | if (U_FAILURE(status)) { return; } |
267 | 0 | data.getUniquePatterns(allPatterns, status); |
268 | 0 | if (U_FAILURE(status)) { return; } |
269 | | |
270 | | // C++ only: ensure that precomputedMods has room. |
271 | 0 | precomputedModsLength = allPatterns.size(); |
272 | 0 | if (precomputedMods.getCapacity() < precomputedModsLength) { |
273 | 0 | precomputedMods.resize(allPatterns.size(), status); |
274 | 0 | if (U_FAILURE(status)) { return; } |
275 | 0 | } |
276 | | |
277 | 0 | for (int32_t i = 0; i < precomputedModsLength; i++) { |
278 | 0 | auto patternString = static_cast<const UChar *>(allPatterns[i]); |
279 | 0 | UnicodeString hello(patternString); |
280 | 0 | CompactModInfo &info = precomputedMods[i]; |
281 | 0 | ParsedPatternInfo patternInfo; |
282 | 0 | PatternParser::parseToPatternInfo(UnicodeString(patternString), patternInfo, status); |
283 | 0 | if (U_FAILURE(status)) { return; } |
284 | 0 | buildReference.setPatternInfo(&patternInfo, {UFIELD_CATEGORY_NUMBER, UNUM_COMPACT_FIELD}); |
285 | 0 | info.mod = buildReference.createImmutable(status); |
286 | 0 | if (U_FAILURE(status)) { return; } |
287 | 0 | info.patternString = patternString; |
288 | 0 | } |
289 | 0 | } |
290 | | |
291 | | void CompactHandler::processQuantity(DecimalQuantity &quantity, MicroProps µs, |
292 | 0 | UErrorCode &status) const { |
293 | 0 | parent->processQuantity(quantity, micros, status); |
294 | 0 | if (U_FAILURE(status)) { return; } |
295 | | |
296 | | // Treat zero, NaN, and infinity as if they had magnitude 0 |
297 | 0 | int32_t magnitude; |
298 | 0 | int32_t multiplier = 0; |
299 | 0 | if (quantity.isZeroish()) { |
300 | 0 | magnitude = 0; |
301 | 0 | micros.rounder.apply(quantity, status); |
302 | 0 | } else { |
303 | | // TODO: Revisit chooseMultiplierAndApply |
304 | 0 | multiplier = micros.rounder.chooseMultiplierAndApply(quantity, data, status); |
305 | 0 | magnitude = quantity.isZeroish() ? 0 : quantity.getMagnitude(); |
306 | 0 | magnitude -= multiplier; |
307 | 0 | } |
308 | |
|
309 | 0 | const UChar *patternString = data.getPattern(magnitude, rules, quantity); |
310 | 0 | if (patternString == nullptr) { |
311 | | // Use the default (non-compact) modifier. |
312 | | // No need to take any action. |
313 | 0 | } else if (safe) { |
314 | | // Safe code path. |
315 | | // Java uses a hash set here for O(1) lookup. C++ uses a linear search. |
316 | | // TODO: Benchmark this and maybe change to a binary search or hash table. |
317 | 0 | int32_t i = 0; |
318 | 0 | for (; i < precomputedModsLength; i++) { |
319 | 0 | const CompactModInfo &info = precomputedMods[i]; |
320 | 0 | if (u_strcmp(patternString, info.patternString) == 0) { |
321 | 0 | info.mod->applyToMicros(micros, quantity, status); |
322 | 0 | break; |
323 | 0 | } |
324 | 0 | } |
325 | | // It should be guaranteed that we found the entry. |
326 | 0 | U_ASSERT(i < precomputedModsLength); |
327 | 0 | } else { |
328 | | // Unsafe code path. |
329 | | // Overwrite the PatternInfo in the existing modMiddle. |
330 | | // C++ Note: Use unsafePatternInfo for proper lifecycle. |
331 | 0 | ParsedPatternInfo &patternInfo = const_cast<CompactHandler *>(this)->unsafePatternInfo; |
332 | 0 | PatternParser::parseToPatternInfo(UnicodeString(patternString), patternInfo, status); |
333 | 0 | unsafePatternModifier->setPatternInfo( |
334 | 0 | &unsafePatternInfo, |
335 | 0 | {UFIELD_CATEGORY_NUMBER, UNUM_COMPACT_FIELD}); |
336 | 0 | unsafePatternModifier->setNumberProperties(quantity.signum(), StandardPlural::Form::COUNT); |
337 | 0 | micros.modMiddle = unsafePatternModifier; |
338 | 0 | } |
339 | | |
340 | | // Change the exponent only after we select appropriate plural form |
341 | | // for formatting purposes so that we preserve expected formatted |
342 | | // string behavior. |
343 | 0 | quantity.adjustExponent(-1 * multiplier); |
344 | | |
345 | | // We already performed rounding. Do not perform it again. |
346 | 0 | micros.rounder = RoundingImpl::passThrough(); |
347 | 0 | } |
348 | | |
349 | | #endif /* #if !UCONFIG_NO_FORMATTING */ |