icuSources/i18n/regexst.cpp

   1 //
   2 //  regexst.h
   3 //
   4 //  Copyright (C) 2004-2012, 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,
  84 //          \     u     2    0     0     c      \     u     2    0     0     d     ]
  85           0x5c, 0x75, 0x32, 0x30, 0x30, 0x63, 0x5c, 0x75, 0x32, 0x30, 0x30, 0x64, 0x5d, 0};
  86
  87
  88 //
  89 //  Unicode Set Definitions for Regular Expression  \s
  90 //
  91 static const UChar gIsSpacePattern[] = {
  92 //        [     \     p     {     W     h     i     t     e     S     p     a     c     e     }     ]
  93         0x5b, 0x5c, 0x70, 0x7b, 0x57, 0x68, 0x69, 0x74, 0x65, 0x53, 0x70, 0x61, 0x63, 0x65, 0x7d, 0x5d, 0};
  94
  95
  96 //
  97 //  UnicodeSets used in implementation of Grapheme Cluster detection, \X
  98 //
  99 static const UChar gGC_ControlPattern[] = {
 100 //    [     [     :     Z     l     :     ]     [     :     Z     p     :     ]
 101     0x5b, 0x5b, 0x3a, 0x5A, 0x6c, 0x3a, 0x5d, 0x5b, 0x3a, 0x5A, 0x70, 0x3a, 0x5d,
 102 //    [     :     C     c     :     ]     [     :     C     f     :     ]     -
 103     0x5b, 0x3a, 0x43, 0x63, 0x3a, 0x5d, 0x5b, 0x3a, 0x43, 0x66, 0x3a, 0x5d, 0x2d,
 104 //    [     :     G     r     a     p     h     e     m     e     _
 105     0x5b, 0x3a, 0x47, 0x72, 0x61, 0x70, 0x68, 0x65, 0x6d, 0x65, 0x5f,
 106 //    E     x     t     e     n     d     :     ]     ]
 107     0x45, 0x78, 0x74, 0x65, 0x6e, 0x64, 0x3a, 0x5d, 0x5d, 0};
 108
 109 static const UChar gGC_ExtendPattern[] = {
 110 //    [     \     p     {     G     r     a     p     h     e     m     e     _
 111     0x5b, 0x5c, 0x70, 0x7b, 0x47, 0x72, 0x61, 0x70, 0x68, 0x65, 0x6d, 0x65, 0x5f,
 112 //    E     x     t     e     n     d     }     ]
 113     0x45, 0x78, 0x74, 0x65, 0x6e, 0x64, 0x7d, 0x5d, 0};
 114
 115 static const UChar gGC_LPattern[] = {
 116 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 117     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 118 //    l     a     b     l     e     _     T     y     p     e     =     L     }     ]
 119     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x7d,  0x5d, 0};
 120
 121 static const UChar gGC_VPattern[] = {
 122 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 123     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 124 //    l     a     b     l     e     _     T     y     p     e     =     V     }     ]
 125     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x56, 0x7d,  0x5d, 0};
 126
 127 static const UChar gGC_TPattern[] = {
 128 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 129     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 130 //    l     a     b     l     e     _     T     y     p     e     =     T     }    ]
 131     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x54, 0x7d, 0x5d, 0};
 132
 133 static const UChar gGC_LVPattern[] = {
 134 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 135     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 136 //    l     a     b     l     e     _     T     y     p     e     =     L     V     }     ]
 137     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x56, 0x7d, 0x5d, 0};
 138
 139 static const UChar gGC_LVTPattern[] = {
 140 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 141     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 142 //    l     a     b     l     e     _     T     y     p     e     =     L     V     T     }     ]
 143     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x56, 0x54, 0x7d, 0x5d, 0};
 144
 145
 146 RegexStaticSets *RegexStaticSets::gStaticSets = NULL;
 147
 148 RegexStaticSets::RegexStaticSets(UErrorCode *status)
 149 :
 150 fUnescapeCharSet(UnicodeString(TRUE, gUnescapeCharPattern, -1), *status),
 151 fRuleDigitsAlias(NULL),
 152 fEmptyText(NULL)
 153 {
 154     // First zero out everything
 155     int i;
 156     for (i=0; i<URX_LAST_SET; i++) {
 157         fPropSets[i] = NULL;
 158     }
 159     // Then init the sets to their correct values.
 160     fPropSets[URX_ISWORD_SET]  = new UnicodeSet(UnicodeString(TRUE, gIsWordPattern, -1),     *status);
 161     fPropSets[URX_ISSPACE_SET] = new UnicodeSet(UnicodeString(TRUE, gIsSpacePattern, -1),    *status);
 162     fPropSets[URX_GC_EXTEND]   = new UnicodeSet(UnicodeString(TRUE, gGC_ExtendPattern, -1),  *status);
 163     fPropSets[URX_GC_CONTROL]  = new UnicodeSet(UnicodeString(TRUE, gGC_ControlPattern, -1), *status);
 164     fPropSets[URX_GC_L]        = new UnicodeSet(UnicodeString(TRUE, gGC_LPattern, -1),       *status);
 165     fPropSets[URX_GC_V]        = new UnicodeSet(UnicodeString(TRUE, gGC_VPattern, -1),       *status);
 166     fPropSets[URX_GC_T]        = new UnicodeSet(UnicodeString(TRUE, gGC_TPattern, -1),       *status);
 167     fPropSets[URX_GC_LV]       = new UnicodeSet(UnicodeString(TRUE, gGC_LVPattern, -1),      *status);
 168     fPropSets[URX_GC_LVT]      = new UnicodeSet(UnicodeString(TRUE, gGC_LVTPattern, -1),     *status);
 169
 170     // Check for null pointers
 171     if (fPropSets[URX_ISWORD_SET] == NULL || fPropSets[URX_ISSPACE_SET] == NULL || fPropSets[URX_GC_EXTEND] == NULL ||
 172         fPropSets[URX_GC_CONTROL] == NULL || fPropSets[URX_GC_L] == NULL || fPropSets[URX_GC_V] == NULL ||
 173         fPropSets[URX_GC_T] == NULL || fPropSets[URX_GC_LV] == NULL || fPropSets[URX_GC_LVT] == NULL) {
 174         goto ExitConstrDeleteAll;
 175     }
 176     if (U_FAILURE(*status)) {
 177         // Bail out if we were unable to create the above sets.
 178         // The rest of the initialization needs them, so we cannot proceed.
 179         return;
 180     }
 181
 182
 183     //
 184     // The following sets  are dynamically constructed, because their
 185     //   initialization strings would be unreasonable.
 186     //
 187
 188
 189     //
 190     //  "Normal" is the set of characters that don't need special handling
 191     //            when finding grapheme cluster boundaries.
 192     //
 193     fPropSets[URX_GC_NORMAL] = new UnicodeSet(0, UnicodeSet::MAX_VALUE);
 194     // Null pointer check
 195     if (fPropSets[URX_GC_NORMAL] == NULL) {
 196         goto ExitConstrDeleteAll;
 197     }
 198     fPropSets[URX_GC_NORMAL]->remove(0xac00, 0xd7a4);
 199     fPropSets[URX_GC_NORMAL]->removeAll(*fPropSets[URX_GC_CONTROL]);
 200     fPropSets[URX_GC_NORMAL]->removeAll(*fPropSets[URX_GC_L]);
 201     fPropSets[URX_GC_NORMAL]->removeAll(*fPropSets[URX_GC_V]);
 202     fPropSets[URX_GC_NORMAL]->removeAll(*fPropSets[URX_GC_T]);
 203
 204     // Initialize the 8-bit fast bit sets from the parallel full
 205     //   UnicodeSets.
 206     for (i=0; i<URX_LAST_SET; i++) {
 207         if (fPropSets[i]) {
 208             fPropSets[i]->compact();
 209             fPropSets8[i].init(fPropSets[i]);
 210         }
 211     }
 212
 213     // Sets used while parsing rules, but not referenced from the parse state table
 214     fRuleSets[kRuleSet_rule_char-128]   = UnicodeSet(UnicodeString(TRUE, gRuleSet_rule_char_pattern, -1),   *status);
 215     fRuleSets[kRuleSet_digit_char-128]  = UnicodeSet(UnicodeString(TRUE, gRuleSet_digit_char_pattern, -1),  *status);
 216     fRuleDigitsAlias = &fRuleSets[kRuleSet_digit_char-128];
 217     for (i=0; i<(int32_t)(sizeof(fRuleSets)/sizeof(fRuleSets[0])); i++) {
 218         fRuleSets[i].compact();
 219     }
 220
 221     // Finally, initialize an empty string for utility purposes
 222     fEmptyText = utext_openUChars(NULL, NULL, 0, status);
 223
 224     return; // If we reached this point, everything is fine so just exit
 225
 226 ExitConstrDeleteAll: // Remove fPropSets and fRuleSets and return error
 227     for (i=0; i<URX_LAST_SET; i++) {
 228         delete fPropSets[i];
 229         fPropSets[i] = NULL;
 230     }
 231     *status = U_MEMORY_ALLOCATION_ERROR;
 232 }
 233
 234
 235 RegexStaticSets::~RegexStaticSets() {
 236     int32_t i;
 237
 238     for (i=0; i<URX_LAST_SET; i++) {
 239         delete fPropSets[i];
 240         fPropSets[i] = NULL;
 241     }
 242     fRuleDigitsAlias = NULL;
 243
 244     utext_close(fEmptyText);
 245 }
 246
 247
 248 //------------------------------------------------------------------------------
 249 //
 250 //   regex_cleanup      Memory cleanup function, free/delete all
 251 //                      cached memory.  Called by ICU's u_cleanup() function.
 252 //
 253 //------------------------------------------------------------------------------
 254 UBool
 255 RegexStaticSets::cleanup(void) {
 256     delete RegexStaticSets::gStaticSets;
 257     RegexStaticSets::gStaticSets = NULL;
 258     return TRUE;
 259 }
 260
 261 U_CDECL_BEGIN
 262 static UBool U_CALLCONV
 263 regex_cleanup(void) {
 264     return RegexStaticSets::cleanup();
 265 }
 266 U_CDECL_END
 267
 268 void RegexStaticSets::initGlobals(UErrorCode *status) {
 269     RegexStaticSets *p;
 270     UMTX_CHECK(NULL, gStaticSets, p);
 271     if (p == NULL) {
 272         p = new RegexStaticSets(status);
 273         if (p == NULL) {
 274                 *status = U_MEMORY_ALLOCATION_ERROR;
 275                 return;
 276         }
 277         if (U_FAILURE(*status)) {
 278             delete p;
 279             return;
 280         }
 281         umtx_lock(NULL);
 282         if (gStaticSets == NULL) {
 283             gStaticSets = p;
 284             p = NULL;
 285         }
 286         umtx_unlock(NULL);
 287         if (p) {
 288             delete p;
 289         }
 290         ucln_i18n_registerCleanup(UCLN_I18N_REGEX, regex_cleanup);
 291     }
 292 }
 293
 294
 295 U_NAMESPACE_END
 296 #endif  // !UCONFIG_NO_REGULAR_EXPRESSIONS