icuSources/i18n/regexst.cpp

   1 //
   2 //  regexst.h
   3 //
   4 //  Copyright (C) 2004-2010, International Business Machines Corporation and others.
   5 //  All Rights Reserved.
   6 //
   7 //  This file contains class RegexStaticSets
   8 //
   9 //  This class is internal to the regular expression implementation.
  10 //  For the public Regular Expression API, see the file "unicode/regex.h"
  11 //
  12 //  RegexStaticSets groups together the common UnicodeSets that are needed
  13 //   for compiling or executing RegularExpressions.  This grouping simplifies
  14 //   the thread safe lazy creation and sharing of these sets across
  15 //   all instances of regular expressions.
  16 //
  17 #include "unicode/utypes.h"
  18
  19 #if !UCONFIG_NO_REGULAR_EXPRESSIONS
  20
  21 #include "unicode/unistr.h"
  22 #include "unicode/uniset.h"
  23 #include "unicode/uchar.h"
  24 #include "unicode/regex.h"
  25 #include "uprops.h"
  26 #include "cmemory.h"
  27 #include "cstring.h"
  28 #include "uassert.h"
  29 #include "ucln_in.h"
  30 #include "umutex.h"
  31
  32 #include "regexcst.h"   // Contains state table for the regex pattern parser.
  33                         //   generated by a Perl script.
  34 #include "regexst.h"
  35
  36
  37
  38 U_NAMESPACE_BEGIN
  39
  40
  41 //------------------------------------------------------------------------------
  42 //
  43 // Unicode Set pattern strings for all of the required constant sets.
  44 //               Initialized with hex values for portability to EBCDIC based machines.
  45 //                Really ugly, but there's no good way to avoid it.
  46 //
  47 //------------------------------------------------------------------------------
  48
  49 // "Rule Char" Characters are those with no special meaning, and therefore do not
  50 //    need to be escaped to appear as literals in a regexp.  Expressed
  51 //    as the inverse of those needing escaping --  [^\*\?\+\[\(\)\{\}\^\$\|\\\.]
  52 static const UChar gRuleSet_rule_char_pattern[]       = {
  53  //   [    ^      \     *     \     ?     \     +     \     [     \     (     /     )
  54     0x5b, 0x5e, 0x5c, 0x2a, 0x5c, 0x3f, 0x5c, 0x2b, 0x5c, 0x5b, 0x5c, 0x28, 0x5c, 0x29,
  55  //   \     {    \     }     \     ^     \     $     \     |     \     \     \     .     ]
  56     0x5c, 0x7b,0x5c, 0x7d, 0x5c, 0x5e, 0x5c, 0x24, 0x5c, 0x7c, 0x5c, 0x5c, 0x5c, 0x2e, 0x5d, 0};
  57
  58
  59 static const UChar gRuleSet_digit_char_pattern[] = {
  60 //    [    0      -    9     ]
  61     0x5b, 0x30, 0x2d, 0x39, 0x5d, 0};
  62
  63 //
  64 //   Here are the backslash escape characters that ICU's unescape() function
  65 //    will handle.
  66 //
  67 static const UChar gUnescapeCharPattern[] = {
  68 //    [     a     c     e     f     n     r     t     u     U     x    ]
  69     0x5b, 0x61, 0x63, 0x65, 0x66, 0x6e, 0x72, 0x74, 0x75, 0x55, 0x78, 0x5d, 0};
  70
  71
  72 //
  73 //  Unicode Set Definitions for Regular Expression  \w
  74 //
  75 static const UChar gIsWordPattern[] = {
  76 //    [     \     p     {    A     l     p     h     a     b     e     t     i      c    }
  77     0x5b, 0x5c, 0x70, 0x7b, 0x61, 0x6c, 0x70, 0x68, 0x61, 0x62, 0x65, 0x74, 0x69, 0x63, 0x7d,
  78 //          \     p     {    M     }                               Mark
  79           0x5c, 0x70, 0x7b, 0x4d, 0x7d,
  80 //          \     p     {    N     d     }                         Digit_Numeric
  81           0x5c, 0x70, 0x7b, 0x4e, 0x64, 0x7d,
  82 //          \     p     {    P     c     }      ]                  Connector_Punctuation
  83           0x5c, 0x70, 0x7b, 0x50, 0x63, 0x7d, 0x5d, 0};
  84
  85
  86 //
  87 //  Unicode Set Definitions for Regular Expression  \s
  88 //
  89 static const UChar gIsSpacePattern[] = {
  90 //        [     \     p     {     W     h     i     t     e     S     p     a     c     e     }     ]
  91         0x5b, 0x5c, 0x70, 0x7b, 0x57, 0x68, 0x69, 0x74, 0x65, 0x53, 0x70, 0x61, 0x63, 0x65, 0x7d, 0x5d, 0};
  92
  93
  94 //
  95 //  UnicodeSets used in implementation of Grapheme Cluster detection, \X
  96 //
  97 static const UChar gGC_ControlPattern[] = {
  98 //    [     [     :     Z     l     :     ]     [     :     Z     p     :     ]
  99     0x5b, 0x5b, 0x3a, 0x5A, 0x6c, 0x3a, 0x5d, 0x5b, 0x3a, 0x5A, 0x70, 0x3a, 0x5d,
 100 //    [     :     C     c     :     ]     [     :     C     f     :     ]     -
 101     0x5b, 0x3a, 0x43, 0x63, 0x3a, 0x5d, 0x5b, 0x3a, 0x43, 0x66, 0x3a, 0x5d, 0x2d,
 102 //    [     :     G     r     a     p     h     e     m     e     _
 103     0x5b, 0x3a, 0x47, 0x72, 0x61, 0x70, 0x68, 0x65, 0x6d, 0x65, 0x5f,
 104 //    E     x     t     e     n     d     :     ]     ]
 105     0x45, 0x78, 0x74, 0x65, 0x6e, 0x64, 0x3a, 0x5d, 0x5d, 0};
 106
 107 static const UChar gGC_ExtendPattern[] = {
 108 //    [     \     p     {     G     r     a     p     h     e     m     e     _
 109     0x5b, 0x5c, 0x70, 0x7b, 0x47, 0x72, 0x61, 0x70, 0x68, 0x65, 0x6d, 0x65, 0x5f,
 110 //    E     x     t     e     n     d     }     ]
 111     0x45, 0x78, 0x74, 0x65, 0x6e, 0x64, 0x7d, 0x5d, 0};
 112
 113 static const UChar gGC_LPattern[] = {
 114 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 115     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 116 //    l     a     b     l     e     _     T     y     p     e     =     L     }     ]
 117     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x7d,  0x5d, 0};
 118
 119 static const UChar gGC_VPattern[] = {
 120 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 121     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 122 //    l     a     b     l     e     _     T     y     p     e     =     V     }     ]
 123     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x56, 0x7d,  0x5d, 0};
 124
 125 static const UChar gGC_TPattern[] = {
 126 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 127     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 128 //    l     a     b     l     e     _     T     y     p     e     =     T     }    ]
 129     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x54, 0x7d, 0x5d, 0};
 130
 131 static const UChar gGC_LVPattern[] = {
 132 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 133     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 134 //    l     a     b     l     e     _     T     y     p     e     =     L     V     }     ]
 135     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x56, 0x7d, 0x5d, 0};
 136
 137 static const UChar gGC_LVTPattern[] = {
 138 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 139     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 140 //    l     a     b     l     e     _     T     y     p     e     =     L     V     T     }     ]
 141     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x56, 0x54, 0x7d, 0x5d, 0};
 142
 143
 144 RegexStaticSets *RegexStaticSets::gStaticSets = NULL;
 145
 146 RegexStaticSets::RegexStaticSets(UErrorCode *status)
 147 :
 148 fUnescapeCharSet(UnicodeString(TRUE, gUnescapeCharPattern, -1), *status),
 149 fRuleDigitsAlias(NULL),
 150 fEmptyText(NULL)
 151 {
 152     // First zero out everything
 153     int i;
 154     for (i=0; i<URX_LAST_SET; i++) {
 155         fPropSets[i] = NULL;
 156     }
 157     // Then init the sets to their correct values.
 158     fPropSets[URX_ISWORD_SET]  = new UnicodeSet(UnicodeString(TRUE, gIsWordPattern, -1),     *status);
 159     fPropSets[URX_ISSPACE_SET] = new UnicodeSet(UnicodeString(TRUE, gIsSpacePattern, -1),    *status);
 160     fPropSets[URX_GC_EXTEND]   = new UnicodeSet(UnicodeString(TRUE, gGC_ExtendPattern, -1),  *status);
 161     fPropSets[URX_GC_CONTROL]  = new UnicodeSet(UnicodeString(TRUE, gGC_ControlPattern, -1), *status);
 162     fPropSets[URX_GC_L]        = new UnicodeSet(UnicodeString(TRUE, gGC_LPattern, -1),       *status);
 163     fPropSets[URX_GC_V]        = new UnicodeSet(UnicodeString(TRUE, gGC_VPattern, -1),       *status);
 164     fPropSets[URX_GC_T]        = new UnicodeSet(UnicodeString(TRUE, gGC_TPattern, -1),       *status);
 165     fPropSets[URX_GC_LV]       = new UnicodeSet(UnicodeString(TRUE, gGC_LVPattern, -1),      *status);
 166     fPropSets[URX_GC_LVT]      = new UnicodeSet(UnicodeString(TRUE, gGC_LVTPattern, -1),     *status);
 167
 168     // Check for null pointers
 169     if (fPropSets[URX_ISWORD_SET] == NULL || fPropSets[URX_ISSPACE_SET] == NULL || fPropSets[URX_GC_EXTEND] == NULL ||
 170         fPropSets[URX_GC_CONTROL] == NULL || fPropSets[URX_GC_L] == NULL || fPropSets[URX_GC_V] == NULL ||
 171         fPropSets[URX_GC_T] == NULL || fPropSets[URX_GC_LV] == NULL || fPropSets[URX_GC_LVT] == NULL) {
 172         goto ExitConstrDeleteAll;
 173     }
 174     if (U_FAILURE(*status)) {
 175         // Bail out if we were unable to create the above sets.
 176         // The rest of the initialization needs them, so we cannot proceed.
 177         return;
 178     }
 179
 180
 181     //
 182     // The following sets  are dynamically constructed, because their
 183     //   initialization strings would be unreasonable.
 184     //
 185
 186
 187     //
 188     //  "Normal" is the set of characters that don't need special handling
 189     //            when finding grapheme cluster boundaries.
 190     //
 191     fPropSets[URX_GC_NORMAL] = new UnicodeSet(0, UnicodeSet::MAX_VALUE);
 192     // Null pointer check
 193     if (fPropSets[URX_GC_NORMAL] == NULL) {
 194         goto ExitConstrDeleteAll;
 195     }
 196     fPropSets[URX_GC_NORMAL]->remove(0xac00, 0xd7a4);
 197     fPropSets[URX_GC_NORMAL]->removeAll(*fPropSets[URX_GC_CONTROL]);
 198     fPropSets[URX_GC_NORMAL]->removeAll(*fPropSets[URX_GC_L]);
 199     fPropSets[URX_GC_NORMAL]->removeAll(*fPropSets[URX_GC_V]);
 200     fPropSets[URX_GC_NORMAL]->removeAll(*fPropSets[URX_GC_T]);
 201
 202     // Initialize the 8-bit fast bit sets from the parallel full
 203     //   UnicodeSets.
 204     for (i=0; i<URX_LAST_SET; i++) {
 205         if (fPropSets[i]) {
 206             fPropSets[i]->compact();
 207             fPropSets8[i].init(fPropSets[i]);
 208         }
 209     }
 210
 211     // Sets used while parsing rules, but not referenced from the parse state table
 212     fRuleSets[kRuleSet_rule_char-128]   = UnicodeSet(UnicodeString(TRUE, gRuleSet_rule_char_pattern, -1),   *status);
 213     fRuleSets[kRuleSet_digit_char-128]  = UnicodeSet(UnicodeString(TRUE, gRuleSet_digit_char_pattern, -1),  *status);
 214     fRuleDigitsAlias = &fRuleSets[kRuleSet_digit_char-128];
 215     for (i=0; i<(int32_t)(sizeof(fRuleSets)/sizeof(fRuleSets[0])); i++) {
 216         fRuleSets[i].compact();
 217     }
 218
 219     // Finally, initialize an empty string for utility purposes
 220     fEmptyText = utext_openUChars(NULL, NULL, 0, status);
 221
 222     return; // If we reached this point, everything is fine so just exit
 223
 224 ExitConstrDeleteAll: // Remove fPropSets and fRuleSets and return error
 225     for (i=0; i<URX_LAST_SET; i++) {
 226         delete fPropSets[i];
 227         fPropSets[i] = NULL;
 228     }
 229     *status = U_MEMORY_ALLOCATION_ERROR;
 230 }
 231
 232
 233 RegexStaticSets::~RegexStaticSets() {
 234     int32_t i;
 235
 236     for (i=0; i<URX_LAST_SET; i++) {
 237         delete fPropSets[i];
 238         fPropSets[i] = NULL;
 239     }
 240     fRuleDigitsAlias = NULL;
 241
 242     utext_close(fEmptyText);
 243 }
 244
 245
 246 //------------------------------------------------------------------------------
 247 //
 248 //   regex_cleanup      Memory cleanup function, free/delete all
 249 //                      cached memory.  Called by ICU's u_cleanup() function.
 250 //
 251 //------------------------------------------------------------------------------
 252 UBool
 253 RegexStaticSets::cleanup(void) {
 254     delete RegexStaticSets::gStaticSets;
 255     RegexStaticSets::gStaticSets = NULL;
 256     return TRUE;
 257 }
 258
 259 U_CDECL_BEGIN
 260 static UBool U_CALLCONV
 261 regex_cleanup(void) {
 262     return RegexStaticSets::cleanup();
 263 }
 264 U_CDECL_END
 265
 266 void RegexStaticSets::initGlobals(UErrorCode *status) {
 267     RegexStaticSets *p;
 268     UMTX_CHECK(NULL, gStaticSets, p);
 269     if (p == NULL) {
 270         p = new RegexStaticSets(status);
 271         if (p == NULL) {
 272                 *status = U_MEMORY_ALLOCATION_ERROR;
 273                 return;
 274         }
 275         if (U_FAILURE(*status)) {
 276             delete p;
 277             return;
 278         }
 279         umtx_lock(NULL);
 280         if (gStaticSets == NULL) {
 281             gStaticSets = p;
 282             p = NULL;
 283         }
 284         umtx_unlock(NULL);
 285         if (p) {
 286             delete p;
 287         }
 288         ucln_i18n_registerCleanup(UCLN_I18N_REGEX, regex_cleanup);
 289     }
 290 }
 291
 292
 293 U_NAMESPACE_END
 294 #endif  // !UCONFIG_NO_REGULAR_EXPRESSIONS