1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ********************************************************************** |
5 | * Copyright (C) 1999-2011, International Business Machines |
6 | * Corporation and others. All Rights Reserved. |
7 | ********************************************************************** |
8 | * Date Name Description |
9 | * 11/17/99 aliu Creation. |
10 | ********************************************************************** |
11 | */ |
12 | |
13 | #include "unicode/utypes.h" |
14 | |
15 | #if !UCONFIG_NO_TRANSLITERATION |
16 | |
17 | #include "unicode/unistr.h" |
18 | #include "unicode/uniset.h" |
19 | #include "unicode/utf16.h" |
20 | #include "rbt_set.h" |
21 | #include "rbt_rule.h" |
22 | #include "cmemory.h" |
23 | #include "putilimp.h" |
24 | |
25 | U_CDECL_BEGIN |
26 | static void U_CALLCONV _deleteRule(void *rule) { |
27 | delete (icu::TransliterationRule *)rule; |
28 | } |
29 | U_CDECL_END |
30 | |
31 | //---------------------------------------------------------------------- |
32 | // BEGIN Debugging support |
33 | //---------------------------------------------------------------------- |
34 | |
35 | // #define DEBUG_RBT |
36 | |
37 | #ifdef DEBUG_RBT |
38 | #include <stdio.h> |
39 | #include "charstr.h" |
40 | |
41 | /** |
42 | * @param appendTo result is appended to this param. |
43 | * @param input the string being transliterated |
44 | * @param pos the index struct |
45 | */ |
46 | static UnicodeString& _formatInput(UnicodeString &appendTo, |
47 | const UnicodeString& input, |
48 | const UTransPosition& pos) { |
49 | // Output a string of the form aaa{bbb|ccc|ddd}eee, where |
50 | // the {} indicate the context start and limit, and the || |
51 | // indicate the start and limit. |
52 | if (0 <= pos.contextStart && |
53 | pos.contextStart <= pos.start && |
54 | pos.start <= pos.limit && |
55 | pos.limit <= pos.contextLimit && |
56 | pos.contextLimit <= input.length()) { |
57 | |
58 | UnicodeString a, b, c, d, e; |
59 | input.extractBetween(0, pos.contextStart, a); |
60 | input.extractBetween(pos.contextStart, pos.start, b); |
61 | input.extractBetween(pos.start, pos.limit, c); |
62 | input.extractBetween(pos.limit, pos.contextLimit, d); |
63 | input.extractBetween(pos.contextLimit, input.length(), e); |
64 | appendTo.append(a).append((UChar)123/*{*/).append(b). |
65 | append((UChar)124/*|*/).append(c).append((UChar)124/*|*/).append(d). |
66 | append((UChar)125/*}*/).append(e); |
67 | } else { |
68 | appendTo.append("INVALID UTransPosition" ); |
69 | //appendTo.append((UnicodeString)"INVALID UTransPosition {cs=" + |
70 | // pos.contextStart + ", s=" + pos.start + ", l=" + |
71 | // pos.limit + ", cl=" + pos.contextLimit + "} on " + |
72 | // input); |
73 | } |
74 | return appendTo; |
75 | } |
76 | |
77 | // Append a hex string to the target |
78 | UnicodeString& _appendHex(uint32_t number, |
79 | int32_t digits, |
80 | UnicodeString& target) { |
81 | static const UChar digitString[] = { |
82 | 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, |
83 | 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0 |
84 | }; |
85 | while (digits--) { |
86 | target += digitString[(number >> (digits*4)) & 0xF]; |
87 | } |
88 | return target; |
89 | } |
90 | |
91 | // Replace nonprintable characters with unicode escapes |
92 | UnicodeString& _escape(const UnicodeString &source, |
93 | UnicodeString &target) { |
94 | for (int32_t i = 0; i < source.length(); ) { |
95 | UChar32 ch = source.char32At(i); |
96 | i += U16_LENGTH(ch); |
97 | if (ch < 0x09 || (ch > 0x0A && ch < 0x20)|| ch > 0x7E) { |
98 | if (ch <= 0xFFFF) { |
99 | target += "\\u" ; |
100 | _appendHex(ch, 4, target); |
101 | } else { |
102 | target += "\\U" ; |
103 | _appendHex(ch, 8, target); |
104 | } |
105 | } else { |
106 | target += ch; |
107 | } |
108 | } |
109 | return target; |
110 | } |
111 | |
112 | inline void _debugOut(const char* msg, TransliterationRule* rule, |
113 | const Replaceable& theText, UTransPosition& pos) { |
114 | UnicodeString buf(msg, "" ); |
115 | if (rule) { |
116 | UnicodeString r; |
117 | rule->toRule(r, TRUE); |
118 | buf.append((UChar)32).append(r); |
119 | } |
120 | buf.append(UnicodeString(" => " , "" )); |
121 | UnicodeString* text = (UnicodeString*)&theText; |
122 | _formatInput(buf, *text, pos); |
123 | UnicodeString esc; |
124 | _escape(buf, esc); |
125 | CharString cbuf(esc); |
126 | printf("%s\n" , (const char*) cbuf); |
127 | } |
128 | |
129 | #else |
130 | #define _debugOut(msg, rule, theText, pos) |
131 | #endif |
132 | |
133 | //---------------------------------------------------------------------- |
134 | // END Debugging support |
135 | //---------------------------------------------------------------------- |
136 | |
137 | // Fill the precontext and postcontext with the patterns of the rules |
138 | // that are masking one another. |
139 | static void maskingError(const icu::TransliterationRule& rule1, |
140 | const icu::TransliterationRule& rule2, |
141 | UParseError& parseError) { |
142 | icu::UnicodeString r; |
143 | int32_t len; |
144 | |
145 | parseError.line = parseError.offset = -1; |
146 | |
147 | // for pre-context |
148 | rule1.toRule(r, FALSE); |
149 | len = uprv_min(r.length(), U_PARSE_CONTEXT_LEN-1); |
150 | r.extract(0, len, parseError.preContext); |
151 | parseError.preContext[len] = 0; |
152 | |
153 | //for post-context |
154 | r.truncate(0); |
155 | rule2.toRule(r, FALSE); |
156 | len = uprv_min(r.length(), U_PARSE_CONTEXT_LEN-1); |
157 | r.extract(0, len, parseError.postContext); |
158 | parseError.postContext[len] = 0; |
159 | } |
160 | |
161 | U_NAMESPACE_BEGIN |
162 | |
163 | /** |
164 | * Construct a new empty rule set. |
165 | */ |
166 | TransliterationRuleSet::TransliterationRuleSet(UErrorCode& status) : UMemory() { |
167 | ruleVector = new UVector(&_deleteRule, NULL, status); |
168 | if (U_FAILURE(status)) { |
169 | return; |
170 | } |
171 | if (ruleVector == NULL) { |
172 | status = U_MEMORY_ALLOCATION_ERROR; |
173 | } |
174 | rules = NULL; |
175 | maxContextLength = 0; |
176 | } |
177 | |
178 | /** |
179 | * Copy constructor. |
180 | */ |
181 | TransliterationRuleSet::TransliterationRuleSet(const TransliterationRuleSet& other) : |
182 | UMemory(other), |
183 | ruleVector(0), |
184 | rules(0), |
185 | maxContextLength(other.maxContextLength) { |
186 | |
187 | int32_t i, len; |
188 | uprv_memcpy(index, other.index, sizeof(index)); |
189 | UErrorCode status = U_ZERO_ERROR; |
190 | ruleVector = new UVector(&_deleteRule, NULL, status); |
191 | if (other.ruleVector != 0 && ruleVector != 0 && U_SUCCESS(status)) { |
192 | len = other.ruleVector->size(); |
193 | for (i=0; i<len && U_SUCCESS(status); ++i) { |
194 | TransliterationRule *tempTranslitRule = new TransliterationRule(*(TransliterationRule*)other.ruleVector->elementAt(i)); |
195 | // Null pointer test |
196 | if (tempTranslitRule == NULL) { |
197 | status = U_MEMORY_ALLOCATION_ERROR; |
198 | break; |
199 | } |
200 | ruleVector->addElement(tempTranslitRule, status); |
201 | if (U_FAILURE(status)) { |
202 | break; |
203 | } |
204 | } |
205 | } |
206 | if (other.rules != 0 && U_SUCCESS(status)) { |
207 | UParseError p; |
208 | freeze(p, status); |
209 | } |
210 | } |
211 | |
212 | /** |
213 | * Destructor. |
214 | */ |
215 | TransliterationRuleSet::~TransliterationRuleSet() { |
216 | delete ruleVector; // This deletes the contained rules |
217 | uprv_free(rules); |
218 | } |
219 | |
220 | void TransliterationRuleSet::setData(const TransliterationRuleData* d) { |
221 | /** |
222 | * We assume that the ruleset has already been frozen. |
223 | */ |
224 | int32_t len = index[256]; // see freeze() |
225 | for (int32_t i=0; i<len; ++i) { |
226 | rules[i]->setData(d); |
227 | } |
228 | } |
229 | |
230 | /** |
231 | * Return the maximum context length. |
232 | * @return the length of the longest preceding context. |
233 | */ |
234 | int32_t TransliterationRuleSet::getMaximumContextLength(void) const { |
235 | return maxContextLength; |
236 | } |
237 | |
238 | /** |
239 | * Add a rule to this set. Rules are added in order, and order is |
240 | * significant. The last call to this method must be followed by |
241 | * a call to <code>freeze()</code> before the rule set is used. |
242 | * |
243 | * <p>If freeze() has already been called, calling addRule() |
244 | * unfreezes the rules, and freeze() must be called again. |
245 | * |
246 | * @param adoptedRule the rule to add |
247 | */ |
248 | void TransliterationRuleSet::addRule(TransliterationRule* adoptedRule, |
249 | UErrorCode& status) { |
250 | if (U_FAILURE(status)) { |
251 | delete adoptedRule; |
252 | return; |
253 | } |
254 | ruleVector->addElement(adoptedRule, status); |
255 | |
256 | int32_t len; |
257 | if ((len = adoptedRule->getContextLength()) > maxContextLength) { |
258 | maxContextLength = len; |
259 | } |
260 | |
261 | uprv_free(rules); |
262 | rules = 0; |
263 | } |
264 | |
265 | /** |
266 | * Check this for masked rules and index it to optimize performance. |
267 | * The sequence of operations is: (1) add rules to a set using |
268 | * <code>addRule()</code>; (2) freeze the set using |
269 | * <code>freeze()</code>; (3) use the rule set. If |
270 | * <code>addRule()</code> is called after calling this method, it |
271 | * invalidates this object, and this method must be called again. |
272 | * That is, <code>freeze()</code> may be called multiple times, |
273 | * although for optimal performance it shouldn't be. |
274 | */ |
275 | void TransliterationRuleSet::freeze(UParseError& parseError,UErrorCode& status) { |
276 | /* Construct the rule array and index table. We reorder the |
277 | * rules by sorting them into 256 bins. Each bin contains all |
278 | * rules matching the index value for that bin. A rule |
279 | * matches an index value if string whose first key character |
280 | * has a low byte equal to the index value can match the rule. |
281 | * |
282 | * Each bin contains zero or more rules, in the same order |
283 | * they were found originally. However, the total rules in |
284 | * the bins may exceed the number in the original vector, |
285 | * since rules that have a variable as their first key |
286 | * character will generally fall into more than one bin. |
287 | * |
288 | * That is, each bin contains all rules that either have that |
289 | * first index value as their first key character, or have |
290 | * a set containing the index value as their first character. |
291 | */ |
292 | int32_t n = ruleVector->size(); |
293 | int32_t j; |
294 | int16_t x; |
295 | UVector v(2*n, status); // heuristic; adjust as needed |
296 | |
297 | if (U_FAILURE(status)) { |
298 | return; |
299 | } |
300 | |
301 | /* Precompute the index values. This saves a LOT of time. |
302 | * Be careful not to call malloc(0). |
303 | */ |
304 | int16_t* indexValue = (int16_t*) uprv_malloc( sizeof(int16_t) * (n > 0 ? n : 1) ); |
305 | /* test for NULL */ |
306 | if (indexValue == 0) { |
307 | status = U_MEMORY_ALLOCATION_ERROR; |
308 | return; |
309 | } |
310 | for (j=0; j<n; ++j) { |
311 | TransliterationRule* r = (TransliterationRule*) ruleVector->elementAt(j); |
312 | indexValue[j] = r->getIndexValue(); |
313 | } |
314 | for (x=0; x<256; ++x) { |
315 | index[x] = v.size(); |
316 | for (j=0; j<n; ++j) { |
317 | if (indexValue[j] >= 0) { |
318 | if (indexValue[j] == x) { |
319 | v.addElement(ruleVector->elementAt(j), status); |
320 | } |
321 | } else { |
322 | // If the indexValue is < 0, then the first key character is |
323 | // a set, and we must use the more time-consuming |
324 | // matchesIndexValue check. In practice this happens |
325 | // rarely, so we seldom tread this code path. |
326 | TransliterationRule* r = (TransliterationRule*) ruleVector->elementAt(j); |
327 | if (r->matchesIndexValue((uint8_t)x)) { |
328 | v.addElement(r, status); |
329 | } |
330 | } |
331 | } |
332 | } |
333 | uprv_free(indexValue); |
334 | index[256] = v.size(); |
335 | |
336 | /* Freeze things into an array. |
337 | */ |
338 | uprv_free(rules); // Contains alias pointers |
339 | |
340 | /* You can't do malloc(0)! */ |
341 | if (v.size() == 0) { |
342 | rules = NULL; |
343 | return; |
344 | } |
345 | rules = (TransliterationRule **)uprv_malloc(v.size() * sizeof(TransliterationRule *)); |
346 | /* test for NULL */ |
347 | if (rules == 0) { |
348 | status = U_MEMORY_ALLOCATION_ERROR; |
349 | return; |
350 | } |
351 | for (j=0; j<v.size(); ++j) { |
352 | rules[j] = (TransliterationRule*) v.elementAt(j); |
353 | } |
354 | |
355 | // TODO Add error reporting that indicates the rules that |
356 | // are being masked. |
357 | //UnicodeString errors; |
358 | |
359 | /* Check for masking. This is MUCH faster than our old check, |
360 | * which was each rule against each following rule, since we |
361 | * only have to check for masking within each bin now. It's |
362 | * 256*O(n2^2) instead of O(n1^2), where n1 is the total rule |
363 | * count, and n2 is the per-bin rule count. But n2<<n1, so |
364 | * it's a big win. |
365 | */ |
366 | for (x=0; x<256; ++x) { |
367 | for (j=index[x]; j<index[x+1]-1; ++j) { |
368 | TransliterationRule* r1 = rules[j]; |
369 | for (int32_t k=j+1; k<index[x+1]; ++k) { |
370 | TransliterationRule* r2 = rules[k]; |
371 | if (r1->masks(*r2)) { |
372 | //| if (errors == null) { |
373 | //| errors = new StringBuffer(); |
374 | //| } else { |
375 | //| errors.append("\n"); |
376 | //| } |
377 | //| errors.append("Rule " + r1 + " masks " + r2); |
378 | status = U_RULE_MASK_ERROR; |
379 | maskingError(*r1, *r2, parseError); |
380 | return; |
381 | } |
382 | } |
383 | } |
384 | } |
385 | |
386 | //if (errors != null) { |
387 | // throw new IllegalArgumentException(errors.toString()); |
388 | //} |
389 | } |
390 | |
391 | /** |
392 | * Transliterate the given text with the given UTransPosition |
393 | * indices. Return TRUE if the transliteration should continue |
394 | * or FALSE if it should halt (because of a U_PARTIAL_MATCH match). |
395 | * Note that FALSE is only ever returned if isIncremental is TRUE. |
396 | * @param text the text to be transliterated |
397 | * @param pos the position indices, which will be updated |
398 | * @param incremental if TRUE, assume new text may be inserted |
399 | * at index.limit, and return FALSE if thre is a partial match. |
400 | * @return TRUE unless a U_PARTIAL_MATCH has been obtained, |
401 | * indicating that transliteration should stop until more text |
402 | * arrives. |
403 | */ |
404 | UBool TransliterationRuleSet::transliterate(Replaceable& text, |
405 | UTransPosition& pos, |
406 | UBool incremental) { |
407 | int16_t indexByte = (int16_t) (text.char32At(pos.start) & 0xFF); |
408 | for (int32_t i=index[indexByte]; i<index[indexByte+1]; ++i) { |
409 | UMatchDegree m = rules[i]->matchAndReplace(text, pos, incremental); |
410 | switch (m) { |
411 | case U_MATCH: |
412 | _debugOut("match" , rules[i], text, pos); |
413 | return TRUE; |
414 | case U_PARTIAL_MATCH: |
415 | _debugOut("partial match" , rules[i], text, pos); |
416 | return FALSE; |
417 | default: /* Ram: added default to make GCC happy */ |
418 | break; |
419 | } |
420 | } |
421 | // No match or partial match from any rule |
422 | pos.start += U16_LENGTH(text.char32At(pos.start)); |
423 | _debugOut("no match" , NULL, text, pos); |
424 | return TRUE; |
425 | } |
426 | |
427 | /** |
428 | * Create rule strings that represents this rule set. |
429 | */ |
430 | UnicodeString& TransliterationRuleSet::toRules(UnicodeString& ruleSource, |
431 | UBool escapeUnprintable) const { |
432 | int32_t i; |
433 | int32_t count = ruleVector->size(); |
434 | ruleSource.truncate(0); |
435 | for (i=0; i<count; ++i) { |
436 | if (i != 0) { |
437 | ruleSource.append((UChar) 0x000A /*\n*/); |
438 | } |
439 | TransliterationRule *r = |
440 | (TransliterationRule*) ruleVector->elementAt(i); |
441 | r->toRule(ruleSource, escapeUnprintable); |
442 | } |
443 | return ruleSource; |
444 | } |
445 | |
446 | /** |
447 | * Return the set of all characters that may be modified |
448 | * (getTarget=false) or emitted (getTarget=true) by this set. |
449 | */ |
450 | UnicodeSet& TransliterationRuleSet::getSourceTargetSet(UnicodeSet& result, |
451 | UBool getTarget) const |
452 | { |
453 | result.clear(); |
454 | int32_t count = ruleVector->size(); |
455 | for (int32_t i=0; i<count; ++i) { |
456 | TransliterationRule* r = |
457 | (TransliterationRule*) ruleVector->elementAt(i); |
458 | if (getTarget) { |
459 | r->addTargetSetTo(result); |
460 | } else { |
461 | r->addSourceSetTo(result); |
462 | } |
463 | } |
464 | return result; |
465 | } |
466 | |
467 | U_NAMESPACE_END |
468 | |
469 | #endif /* #if !UCONFIG_NO_TRANSLITERATION */ |
470 | |