1 /*
2 **********************************************************************
3 *   Copyright (C) 1999-2014, International Business Machines
4 *   Corporation and others.  All Rights Reserved.
5 **********************************************************************
6 *   Date        Name        Description
7 *   12/09/99    aliu        Ported from Java.
8 **********************************************************************
9 */
10 
11 #include "unicode/utypes.h"
12 
13 #if !UCONFIG_NO_COLLATION
14 
15 #include "thcoll.h"
16 #include "unicode/utypes.h"
17 #include "unicode/coll.h"
18 #include "unicode/localpointer.h"
19 #include "unicode/sortkey.h"
20 #include "unicode/tblcoll.h"
21 #include "unicode/ustring.h"
22 #include "cstring.h"
23 #include "filestrm.h"
24 #include "textfile.h"
25 
26 /**
27  * The TestDictionary test expects a file of this name, with this
28  * encoding, to be present in the directory $ICU/source/test/testdata.
29  */
30 //#define TEST_FILE           "th18057.txt"
31 
32 /**
33  * This is the most failures we show in TestDictionary.  If this number
34  * is < 0, we show all failures.
35  */
36 #define MAX_FAILURES_TO_SHOW -1
37 
CollationThaiTest()38 CollationThaiTest::CollationThaiTest() {
39     UErrorCode status = U_ZERO_ERROR;
40     coll = Collator::createInstance(Locale("th", "TH", ""), status);
41     if (coll && U_SUCCESS(status)) {
42         //coll->setStrength(Collator::TERTIARY);
43     } else {
44         delete coll;
45         coll = 0;
46     }
47 }
48 
~CollationThaiTest()49 CollationThaiTest::~CollationThaiTest() {
50     delete coll;
51 }
52 
runIndexedTest(int32_t index,UBool exec,const char * & name,char *)53 void CollationThaiTest::runIndexedTest(int32_t index, UBool exec, const char* &name,
54                                        char* /*par*/) {
55 
56     if((!coll) && exec) {
57       dataerrln(__FILE__ " cannot test - failed to create collator.");
58       name = "some test";
59       return;
60     }
61 
62     switch (index) {
63         TESTCASE(0,TestDictionary);
64         TESTCASE(1,TestCornerCases);
65         TESTCASE(2,TestNamesList);
66         TESTCASE(3,TestInvalidThai);
67         TESTCASE(4,TestReordering);
68         default: name = ""; break;
69     }
70 }
71 
72 /**
73  * Read the external names list, and confirms that the collator
74  * gets the same results when comparing lines one to another
75  * using regular and iterative comparison.
76  */
TestNamesList(void)77 void CollationThaiTest::TestNamesList(void) {
78     if (coll == 0) {
79         errln("Error: could not construct Thai collator");
80         return;
81     }
82 
83     UErrorCode ec = U_ZERO_ERROR;
84     TextFile names("TestNames_Thai.txt", "UTF16LE", ec);
85     if (U_FAILURE(ec)) {
86         logln("Can't open TestNames_Thai.txt: %s; skipping test",
87               u_errorName(ec));
88         return;
89     }
90 
91     //
92     // Loop through each word in the dictionary and compare it to the previous
93     // word.  They should be in sorted order.
94     //
95     UnicodeString lastWord, word;
96     //int32_t failed = 0;
97     int32_t wordCount = 0;
98     while (names.readLineSkippingComments(word, ec, FALSE) && U_SUCCESS(ec)) {
99 
100         // Show the first 8 words being compared, so we can see what's happening
101         ++wordCount;
102         if (wordCount <= 8) {
103             UnicodeString str;
104             logln((UnicodeString)"Word " + wordCount + ": " + IntlTest::prettify(word, str));
105         }
106 
107         if (lastWord.length() > 0) {
108             Collator::EComparisonResult result = coll->compare(lastWord, word);
109             doTest(coll, lastWord, word, result);
110         }
111         lastWord = word;
112     }
113 
114     assertSuccess("readLine", ec);
115 
116     logln((UnicodeString)"Words checked: " + wordCount);
117 }
118 
119 /**
120  * Read the external dictionary file, which is already in proper
121  * sorted order, and confirm that the collator compares each line as
122  * preceding the following line.
123  */
TestDictionary(void)124 void CollationThaiTest::TestDictionary(void) {
125     if (coll == 0) {
126         errln("Error: could not construct Thai collator");
127         return;
128     }
129 
130     UErrorCode ec = U_ZERO_ERROR;
131     TextFile riwords("riwords.txt", "UTF8", ec);
132     if (U_FAILURE(ec)) {
133         logln("Can't open riwords.txt: %s; skipping test",
134               u_errorName(ec));
135         return;
136     }
137 
138     //
139     // Loop through each word in the dictionary and compare it to the previous
140     // word.  They should be in sorted order.
141     //
142     UnicodeString lastWord, word;
143     int32_t failed = 0;
144     int32_t wordCount = 0;
145     while (riwords.readLineSkippingComments(word, ec, FALSE) && U_SUCCESS(ec)) {
146 
147         // Show the first 8 words being compared, so we can see what's happening
148         ++wordCount;
149         if (wordCount <= 8) {
150             UnicodeString str;
151             logln((UnicodeString)"Word " + wordCount + ": " + IntlTest::prettify(word, str));
152         }
153 
154         if (lastWord.length() > 0) {
155             int32_t result = coll->compare(lastWord, word);
156 
157             if (result > 0) {
158                 failed++;
159                 if (MAX_FAILURES_TO_SHOW < 0 || failed <= MAX_FAILURES_TO_SHOW) {
160                     UnicodeString str;
161                     UnicodeString msg =
162                         UnicodeString("--------------------------------------------\n")
163                         + riwords.getLineNumber()
164                         + " compare(" + IntlTest::prettify(lastWord, str);
165                     msg += UnicodeString(", ")
166                         + IntlTest::prettify(word, str) + ") returned " + result
167                         + ", expected -1\n";
168                     UErrorCode status = U_ZERO_ERROR;
169                     CollationKey k1, k2;
170                     coll->getCollationKey(lastWord, k1, status);
171                     coll->getCollationKey(word, k2, status);
172                     if (U_FAILURE(status)) {
173                         errln((UnicodeString)"Fail: getCollationKey returned " + u_errorName(status));
174                         return;
175                     }
176                     msg.append("key1: ").append(prettify(k1, str)).append("\n");
177                     msg.append("key2: ").append(prettify(k2, str));
178                     errln(msg);
179                 }
180             }
181         }
182         lastWord = word;
183     }
184 
185     assertSuccess("readLine", ec);
186 
187     if (failed != 0) {
188         if (failed > MAX_FAILURES_TO_SHOW) {
189             errln((UnicodeString)"Too many failures; only the first " +
190                   MAX_FAILURES_TO_SHOW + " failures were shown");
191         }
192         errln((UnicodeString)"Summary: " + failed + " of " + (riwords.getLineNumber() - 1) +
193               " comparisons failed");
194     }
195 
196     logln((UnicodeString)"Words checked: " + wordCount);
197 }
198 
199 /**
200  * Odd corner conditions taken from "How to Sort Thai Without Rewriting Sort",
201  * by Doug Cooper, http://seasrc.th.net/paper/thaisort.zip
202  */
TestCornerCases(void)203 void CollationThaiTest::TestCornerCases(void) {
204     const char* TESTS[] = {
205         // Shorter words precede longer
206         "\\u0e01",                               "<",    "\\u0e01\\u0e01",
207 
208         // Tone marks are considered after letters (i.e. are primary ignorable)
209         "\\u0e01\\u0e32",                        "<",    "\\u0e01\\u0e49\\u0e32",
210 
211         // ditto for other over-marks
212         "\\u0e01\\u0e32",                        "<",    "\\u0e01\\u0e32\\u0e4c",
213 
214         // commonly used mark-in-context order.
215         // In effect, marks are sorted after each syllable.
216         "\\u0e01\\u0e32\\u0e01\\u0e49\\u0e32",   "<",    "\\u0e01\\u0e48\\u0e32\\u0e01\\u0e49\\u0e32",
217 
218         // Hyphens and other punctuation follow whitespace but come before letters
219         "\\u0e01\\u0e32",                        "=",    "\\u0e01\\u0e32-",
220         "\\u0e01\\u0e32-",                       "<",    "\\u0e01\\u0e32\\u0e01\\u0e32",
221 
222         // Doubler follows an indentical word without the doubler
223         "\\u0e01\\u0e32",                        "=",    "\\u0e01\\u0e32\\u0e46",
224         "\\u0e01\\u0e32\\u0e46",                 "<",    "\\u0e01\\u0e32\\u0e01\\u0e32",
225 
226 
227         // \\u0e45 after either \\u0e24 or \\u0e26 is treated as a single
228         // combining character, similar to "c < ch" in traditional spanish.
229         // TODO: beef up this case
230         "\\u0e24\\u0e29\\u0e35",                 "<",    "\\u0e24\\u0e45\\u0e29\\u0e35",
231         "\\u0e26\\u0e29\\u0e35",                 "<",    "\\u0e26\\u0e45\\u0e29\\u0e35",
232 
233         // Vowels reorder, should compare \\u0e2d and \\u0e34
234         "\\u0e40\\u0e01\\u0e2d",                 "<",    "\\u0e40\\u0e01\\u0e34",
235 
236         // Tones are compared after the rest of the word (e.g. primary ignorable)
237         "\\u0e01\\u0e32\\u0e01\\u0e48\\u0e32",   "<",    "\\u0e01\\u0e49\\u0e32\\u0e01\\u0e32",
238 
239         // Periods are ignored entirely
240         "\\u0e01.\\u0e01.",                      "<",    "\\u0e01\\u0e32",
241     };
242     const int32_t TESTS_length = (int32_t)(sizeof(TESTS)/sizeof(TESTS[0]));
243 
244     if (coll == 0) {
245         errln("Error: could not construct Thai collator");
246         return;
247     }
248     compareArray(*coll, TESTS, TESTS_length);
249 }
250 
251 //------------------------------------------------------------------------
252 // Internal utilities
253 //------------------------------------------------------------------------
254 
compareArray(Collator & c,const char * tests[],int32_t testsLength)255 void CollationThaiTest::compareArray(Collator& c, const char* tests[],
256                                      int32_t testsLength) {
257     for (int32_t i = 0; i < testsLength; i += 3) {
258 
259         Collator::EComparisonResult expect;
260         if (tests[i+1][0] == '<') {
261           expect = Collator::LESS;
262         } else if (tests[i+1][0] == '>') {
263           expect = Collator::GREATER;
264         } else if (tests[i+1][0] == '=') {
265           expect = Collator::EQUAL;
266         } else {
267             // expect = Integer.decode(tests[i+1]).intValue();
268             errln((UnicodeString)"Error: unknown operator " + tests[i+1]);
269             return;
270         }
271 
272         UnicodeString s1, s2;
273         parseChars(s1, tests[i]);
274         parseChars(s2, tests[i+2]);
275 
276         doTest(&c, s1, s2, expect);
277 #if 0
278         UErrorCode status = U_ZERO_ERROR;
279         int32_t result = c.compare(s1, s2);
280         if (sign(result) != sign(expect))
281         {
282             UnicodeString t1, t2;
283             errln(UnicodeString("") +
284                   i/3 + ": compare(" + IntlTest::prettify(s1, t1)
285                   + " , " + IntlTest::prettify(s2, t2)
286                   + ") got " + result + "; expected " + expect);
287 
288             CollationKey k1, k2;
289             c.getCollationKey(s1, k1, status);
290             c.getCollationKey(s2, k2, status);
291             if (U_FAILURE(status)) {
292                 errln((UnicodeString)"Fail: getCollationKey returned " + u_errorName(status));
293                 return;
294             }
295             errln((UnicodeString)"  key1: " + prettify(k1, t1) );
296             errln((UnicodeString)"  key2: " + prettify(k2, t2) );
297         }
298         else
299         {
300             // Collator.compare worked OK; now try the collation keys
301             CollationKey k1, k2;
302             c.getCollationKey(s1, k1, status);
303             c.getCollationKey(s2, k2, status);
304             if (U_FAILURE(status)) {
305                 errln((UnicodeString)"Fail: getCollationKey returned " + u_errorName(status));
306                 return;
307             }
308 
309             result = k1.compareTo(k2);
310             if (sign(result) != sign(expect)) {
311                 UnicodeString t1, t2;
312                 errln(UnicodeString("") +
313                       i/3 + ": key(" + IntlTest::prettify(s1, t1)
314                       + ").compareTo(key(" + IntlTest::prettify(s2, t2)
315                       + ")) got " + result + "; expected " + expect);
316 
317                 errln((UnicodeString)"  " + prettify(k1, t1) + " vs. " + prettify(k2, t2));
318             }
319         }
320 #endif
321     }
322 }
323 
sign(int32_t i)324 int8_t CollationThaiTest::sign(int32_t i) {
325     if (i < 0) return -1;
326     if (i > 0) return 1;
327     return 0;
328 }
329 
330 /**
331  * Set a UnicodeString corresponding to the given string.  Use
332  * UnicodeString and the default converter, unless we see the sequence
333  * "\\u", in which case we interpret the subsequent escape.
334  */
parseChars(UnicodeString & result,const char * chars)335 UnicodeString& CollationThaiTest::parseChars(UnicodeString& result,
336                                              const char* chars) {
337     return result = CharsToUnicodeString(chars);
338 }
339 
340 UCollator *thaiColl = NULL;
341 
342 U_CDECL_BEGIN
343 static int U_CALLCONV
StrCmp(const void * p1,const void * p2)344 StrCmp(const void *p1, const void *p2) {
345   return ucol_strcoll(thaiColl, *(UChar **) p1, -1,  *(UChar **)p2, -1);
346 }
347 U_CDECL_END
348 
349 
350 #define LINES 6
351 
TestInvalidThai(void)352 void CollationThaiTest::TestInvalidThai(void) {
353   const char *tests[LINES] = {
354     "\\u0E44\\u0E01\\u0E44\\u0E01",
355     "\\u0E44\\u0E01\\u0E01\\u0E44",
356     "\\u0E01\\u0E44\\u0E01\\u0E44",
357     "\\u0E01\\u0E01\\u0E44\\u0E44",
358     "\\u0E44\\u0E44\\u0E01\\u0E01",
359     "\\u0E01\\u0E44\\u0E44\\u0E01",
360   };
361 
362   UChar strings[LINES][20];
363 
364   UChar *toSort[LINES];
365 
366   int32_t i = 0, j = 0, len = 0;
367 
368   UErrorCode coll_status = U_ZERO_ERROR;
369   UnicodeString iteratorText;
370 
371   thaiColl = ucol_open ("th_TH", &coll_status);
372   if (U_FAILURE(coll_status)) {
373     errln("Error opening Thai collator: %s", u_errorName(coll_status));
374     return;
375   }
376 
377   CollationElementIterator* c = ((RuleBasedCollator *)coll)->createCollationElementIterator( iteratorText );
378 
379   for(i = 0; i < (int32_t)(sizeof(tests)/sizeof(tests[0])); i++) {
380     len = u_unescape(tests[i], strings[i], 20);
381     strings[i][len] = 0;
382     toSort[i] = strings[i];
383   }
384 
385   qsort (toSort, LINES, sizeof (UChar *), StrCmp);
386 
387   for (i=0; i < LINES; i++)
388   {
389     logln("%i", i);
390       for (j=i+1; j < LINES; j++) {
391           if (ucol_strcoll (thaiColl, toSort[i], -1, toSort[j], -1) == UCOL_GREATER)
392           {
393               // inconsistency ordering found!
394             errln("Inconsistent ordering between strings %i and %i", i, j);
395           }
396       }
397       iteratorText.setTo(toSort[i]);
398       c->setText(iteratorText, coll_status);
399       backAndForth(*c);
400   }
401 
402 
403   ucol_close(thaiColl);
404   delete c;
405 }
406 
TestReordering(void)407 void CollationThaiTest::TestReordering(void) {
408   // Until UCA 4.1, the collation code swapped Thai/Lao prevowels with the following consonants,
409   // resulting in consonant+prevowel == prevowel+consonant.
410   // From UCA 5.0 on, there are order-reversing contractions for prevowel+consonant.
411   // From UCA 5.0 until UCA 6.1, there was a tertiary difference between
412   // consonant+prevowel and prevowel+consonant.
413   // In UCA 6.2, they compare equal again.
414   // The test was modified to using a collator with strength=secondary,
415   // ignoring possible tertiary differences.
416   const char *tests[] = {
417     "\\u0E41c\\u0301",       "=", "\\u0E41\\u0107", // composition
418     "\\u0E41\\U0001D7CE",    "<", "\\u0E41\\U0001D7CF", // supplementaries
419     "\\u0E41\\U0001D15F",    "=", "\\u0E41\\U0001D158\\U0001D165", // supplementary composition decomps to supplementary
420     "\\u0E41\\U0002F802",    "=", "\\u0E41\\u4E41", // supplementary composition decomps to BMP
421     "\\u0E41\\u0301",        "=", "\\u0E41\\u0301", // unsafe (just checking backwards iteration)
422     "\\u0E41\\u0301\\u0316", "=", "\\u0E41\\u0316\\u0301",
423 
424     "\\u0e24\\u0e41",        "=", "\\u0e41\\u0e24", // exiting contraction bug
425     "\\u0e3f\\u0e3f\\u0e24\\u0e41", "=", "\\u0e3f\\u0e3f\\u0e41\\u0e24",
426 
427     "abc\\u0E41c\\u0301",       "=", "abc\\u0E41\\u0107", // composition
428     "abc\\u0E41\\U0001D000",    "<", "abc\\u0E41\\U0001D001", // supplementaries
429     "abc\\u0E41\\U0001D15F",    "=", "abc\\u0E41\\U0001D158\\U0001D165", // supplementary composition decomps to supplementary
430     "abc\\u0E41\\U0002F802",    "=", "abc\\u0E41\\u4E41", // supplementary composition decomps to BMP
431     "abc\\u0E41\\u0301",        "=", "abc\\u0E41\\u0301", // unsafe (just checking backwards iteration)
432     "abc\\u0E41\\u0301\\u0316", "=", "abc\\u0E41\\u0316\\u0301",
433 
434     "\\u0E41c\\u0301abc",       "=", "\\u0E41\\u0107abc", // composition
435     "\\u0E41\\U0001D000abc",    "<", "\\u0E41\\U0001D001abc", // supplementaries
436     "\\u0E41\\U0001D15Fabc",    "=", "\\u0E41\\U0001D158\\U0001D165abc", // supplementary composition decomps to supplementary
437     "\\u0E41\\U0002F802abc",    "=", "\\u0E41\\u4E41abc", // supplementary composition decomps to BMP
438     "\\u0E41\\u0301abc",        "=", "\\u0E41\\u0301abc", // unsafe (just checking backwards iteration)
439     "\\u0E41\\u0301\\u0316abc", "=", "\\u0E41\\u0316\\u0301abc",
440 
441     "abc\\u0E41c\\u0301abc",       "=", "abc\\u0E41\\u0107abc", // composition
442     "abc\\u0E41\\U0001D000abc",    "<", "abc\\u0E41\\U0001D001abc", // supplementaries
443     "abc\\u0E41\\U0001D15Fabc",    "=", "abc\\u0E41\\U0001D158\\U0001D165abc", // supplementary composition decomps to supplementary
444     "abc\\u0E41\\U0002F802abc",    "=", "abc\\u0E41\\u4E41abc", // supplementary composition decomps to BMP
445     "abc\\u0E41\\u0301abc",        "=", "abc\\u0E41\\u0301abc", // unsafe (just checking backwards iteration)
446     "abc\\u0E41\\u0301\\u0316abc", "=", "abc\\u0E41\\u0316\\u0301abc",
447   };
448 
449   LocalPointer<Collator> coll2(coll->clone());
450   UErrorCode status = U_ZERO_ERROR;
451   coll2->setAttribute(UCOL_STRENGTH, UCOL_SECONDARY, status);
452   if(U_FAILURE(status)) {
453     errln("Unable to set the Thai collator clone to secondary strength");
454     return;
455   }
456   compareArray(*coll2, tests, sizeof(tests)/sizeof(tests[0]));
457 
458   const char *rule = "& c < ab";
459   const char *testcontraction[] = { "\\u0E41ab", ">", "\\u0E41c"}; // After UCA 4.1 Thai are normal so won't break a contraction
460   UnicodeString rules;
461   parseChars(rules, rule);
462   LocalPointer<RuleBasedCollator> rcoll(new RuleBasedCollator(rules, status), status);
463   if(U_SUCCESS(status)) {
464     compareArray(*rcoll, testcontraction, 3);
465   } else {
466     errln("Couldn't instantiate collator from rules");
467   }
468 
469 }
470 
471 
472 #endif /* #if !UCONFIG_NO_COLLATION */
473