icuSources/i18n/regexst.cpp

   1 // © 2016 and later: Unicode, Inc. and others.
   2 // License & terms of use: http://www.unicode.org/copyright.html
   3 //
   4 //  regexst.h
   5 //
   6 //  Copyright (C) 2004-2015, International Business Machines Corporation and others.
   7 //  All Rights Reserved.
   8 //
   9 //  This file contains class RegexStaticSets
  10 //
  11 //  This class is internal to the regular expression implementation.
  12 //  For the public Regular Expression API, see the file "unicode/regex.h"
  13 //
  14 //  RegexStaticSets groups together the common UnicodeSets that are needed
  15 //   for compiling or executing RegularExpressions.  This grouping simplifies
  16 //   the thread safe lazy creation and sharing of these sets across
  17 //   all instances of regular expressions.
  18 //
  19 #include "unicode/utypes.h"
  20
  21 #if !UCONFIG_NO_REGULAR_EXPRESSIONS
  22
  23 #include "unicode/unistr.h"
  24 #include "unicode/uniset.h"
  25 #include "unicode/uchar.h"
  26 #include "unicode/regex.h"
  27 #include "uprops.h"
  28 #include "cmemory.h"
  29 #include "cstring.h"
  30 #include "uassert.h"
  31 #include "ucln_in.h"
  32 #include "umutex.h"
  33
  34 #include "regexcst.h"   // Contains state table for the regex pattern parser.
  35                         //   generated by a Perl script.
  36 #include "regexst.h"
  37
  38
  39
  40 U_NAMESPACE_BEGIN
  41
  42
  43 //------------------------------------------------------------------------------
  44 //
  45 // Unicode Set pattern strings for all of the required constant sets.
  46 //               Initialized with hex values for portability to EBCDIC based machines.
  47 //                Really ugly, but there's no good way to avoid it.
  48 //
  49 //------------------------------------------------------------------------------
  50
  51 // "Rule Char" Characters are those with no special meaning, and therefore do not
  52 //    need to be escaped to appear as literals in a regexp.  Expressed
  53 //    as the inverse of those needing escaping --  [^\*\?\+\[\(\)\{\}\^\$\|\\\.]
  54 static const UChar gRuleSet_rule_char_pattern[]       = {
  55  //   [    ^      \     *     \     ?     \     +     \     [     \     (     /     )
  56     0x5b, 0x5e, 0x5c, 0x2a, 0x5c, 0x3f, 0x5c, 0x2b, 0x5c, 0x5b, 0x5c, 0x28, 0x5c, 0x29,
  57  //   \     {    \     }     \     ^     \     $     \     |     \     \     \     .     ]
  58     0x5c, 0x7b,0x5c, 0x7d, 0x5c, 0x5e, 0x5c, 0x24, 0x5c, 0x7c, 0x5c, 0x5c, 0x5c, 0x2e, 0x5d, 0};
  59
  60 //
  61 //   Here are the backslash escape characters that ICU's unescape() function
  62 //    will handle.
  63 //
  64 static const UChar gUnescapeCharPattern[] = {
  65 //    [     a     c     e     f     n     r     t     u     U     x    ]
  66     0x5b, 0x61, 0x63, 0x65, 0x66, 0x6e, 0x72, 0x74, 0x75, 0x55, 0x78, 0x5d, 0};
  67
  68
  69 //
  70 //  Unicode Set Definitions for Regular Expression  \w
  71 //
  72 static const UChar gIsWordPattern[] = {
  73 //    [     \     p     {    A     l     p     h     a     b     e     t     i      c    }
  74     0x5b, 0x5c, 0x70, 0x7b, 0x61, 0x6c, 0x70, 0x68, 0x61, 0x62, 0x65, 0x74, 0x69, 0x63, 0x7d,
  75 //          \     p     {    M     }                               Mark
  76           0x5c, 0x70, 0x7b, 0x4d, 0x7d,
  77 //          \     p     {    N     d     }                         Digit_Numeric
  78           0x5c, 0x70, 0x7b, 0x4e, 0x64, 0x7d,
  79 //          \     p     {    P     c     }                         Connector_Punctuation
  80           0x5c, 0x70, 0x7b, 0x50, 0x63, 0x7d,
  81 //          \     u     2    0     0     c      \     u     2    0     0     d     ]
  82           0x5c, 0x75, 0x32, 0x30, 0x30, 0x63, 0x5c, 0x75, 0x32, 0x30, 0x30, 0x64, 0x5d, 0};
  83
  84
  85 //
  86 //  Unicode Set Definitions for Regular Expression  \s
  87 //
  88 static const UChar gIsSpacePattern[] = {
  89 //        [     \     p     {     W     h     i     t     e     S     p     a     c     e     }     ]
  90         0x5b, 0x5c, 0x70, 0x7b, 0x57, 0x68, 0x69, 0x74, 0x65, 0x53, 0x70, 0x61, 0x63, 0x65, 0x7d, 0x5d, 0};
  91
  92
  93 //
  94 //  UnicodeSets used in implementation of Grapheme Cluster detection, \X
  95 //
  96 static const UChar gGC_ControlPattern[] = {
  97 //    [     [     :     Z     l     :     ]     [     :     Z     p     :     ]
  98     0x5b, 0x5b, 0x3a, 0x5A, 0x6c, 0x3a, 0x5d, 0x5b, 0x3a, 0x5A, 0x70, 0x3a, 0x5d,
  99 //    [     :     C     c     :     ]     [     :     C     f     :     ]     -
 100     0x5b, 0x3a, 0x43, 0x63, 0x3a, 0x5d, 0x5b, 0x3a, 0x43, 0x66, 0x3a, 0x5d, 0x2d,
 101 //    [     :     G     r     a     p     h     e     m     e     _
 102     0x5b, 0x3a, 0x47, 0x72, 0x61, 0x70, 0x68, 0x65, 0x6d, 0x65, 0x5f,
 103 //    E     x     t     e     n     d     :     ]     ]
 104     0x45, 0x78, 0x74, 0x65, 0x6e, 0x64, 0x3a, 0x5d, 0x5d, 0};
 105
 106 static const UChar gGC_ExtendPattern[] = {
 107 //    [     \     p     {     G     r     a     p     h     e     m     e     _
 108     0x5b, 0x5c, 0x70, 0x7b, 0x47, 0x72, 0x61, 0x70, 0x68, 0x65, 0x6d, 0x65, 0x5f,
 109 //    E     x     t     e     n     d     }     ]
 110     0x45, 0x78, 0x74, 0x65, 0x6e, 0x64, 0x7d, 0x5d, 0};
 111
 112 static const UChar gGC_LPattern[] = {
 113 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 114     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 115 //    l     a     b     l     e     _     T     y     p     e     =     L     }     ]
 116     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x7d,  0x5d, 0};
 117
 118 static const UChar gGC_VPattern[] = {
 119 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 120     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 121 //    l     a     b     l     e     _     T     y     p     e     =     V     }     ]
 122     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x56, 0x7d,  0x5d, 0};
 123
 124 static const UChar gGC_TPattern[] = {
 125 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 126     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 127 //    l     a     b     l     e     _     T     y     p     e     =     T     }    ]
 128     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x54, 0x7d, 0x5d, 0};
 129
 130 static const UChar gGC_LVPattern[] = {
 131 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 132     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 133 //    l     a     b     l     e     _     T     y     p     e     =     L     V     }     ]
 134     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x56, 0x7d, 0x5d, 0};
 135
 136 static const UChar gGC_LVTPattern[] = {
 137 //    [     \     p     {     H     a     n     g     u     l     _     S     y     l
 138     0x5b, 0x5c, 0x70, 0x7b, 0x48, 0x61, 0x6e, 0x67, 0x75, 0x6c, 0x5f, 0x53, 0x79, 0x6c,
 139 //    l     a     b     l     e     _     T     y     p     e     =     L     V     T     }     ]
 140     0x6c, 0x61, 0x62, 0x6c, 0x65, 0x5f, 0x54, 0x79, 0x70, 0x65, 0x3d, 0x4c, 0x56, 0x54, 0x7d, 0x5d, 0};
 141
 142
 143 RegexStaticSets *RegexStaticSets::gStaticSets = NULL;
 144 UInitOnce gStaticSetsInitOnce = U_INITONCE_INITIALIZER;
 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].add((UChar)0x30, (UChar)0x39);    // [0-9]
 214     fRuleSets[kRuleSet_ascii_letter-128].add((UChar)0x41, (UChar)0x5A);  // [A-Z]
 215     fRuleSets[kRuleSet_ascii_letter-128].add((UChar)0x61, (UChar)0x7A);  // [a-z]
 216     fRuleDigitsAlias = &fRuleSets[kRuleSet_digit_char-128];
 217     for (i=0; i<UPRV_LENGTHOF(fRuleSets); 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     if (U_SUCCESS(*status)) {
 225         return;
 226     }
 227
 228 ExitConstrDeleteAll: // Remove fPropSets and fRuleSets and return error
 229     for (i=0; i<URX_LAST_SET; i++) {
 230         delete fPropSets[i];
 231         fPropSets[i] = NULL;
 232     }
 233     if (U_SUCCESS(*status)) {
 234         *status = U_MEMORY_ALLOCATION_ERROR;
 235     }
 236 }
 237
 238
 239 RegexStaticSets::~RegexStaticSets() {
 240     int32_t i;
 241
 242     for (i=0; i<URX_LAST_SET; i++) {
 243         delete fPropSets[i];
 244         fPropSets[i] = NULL;
 245     }
 246     fRuleDigitsAlias = NULL;
 247
 248     utext_close(fEmptyText);
 249 }
 250
 251
 252 //------------------------------------------------------------------------------
 253 //
 254 //   regex_cleanup      Memory cleanup function, free/delete all
 255 //                      cached memory.  Called by ICU's u_cleanup() function.
 256 //
 257 //------------------------------------------------------------------------------
 258 UBool
 259 RegexStaticSets::cleanup(void) {
 260     delete RegexStaticSets::gStaticSets;
 261     RegexStaticSets::gStaticSets = NULL;
 262     gStaticSetsInitOnce.reset();
 263     return TRUE;
 264 }
 265
 266 U_CDECL_BEGIN
 267 static UBool U_CALLCONV
 268 regex_cleanup(void) {
 269     return RegexStaticSets::cleanup();
 270 }
 271
 272 static void U_CALLCONV initStaticSets(UErrorCode &status) {
 273     U_ASSERT(RegexStaticSets::gStaticSets == NULL);
 274     ucln_i18n_registerCleanup(UCLN_I18N_REGEX, regex_cleanup);
 275     RegexStaticSets::gStaticSets = new RegexStaticSets(&status);
 276     if (U_FAILURE(status)) {
 277         delete RegexStaticSets::gStaticSets;
 278         RegexStaticSets::gStaticSets = NULL;
 279     }
 280     if (RegexStaticSets::gStaticSets == NULL && U_SUCCESS(status)) {
 281         status = U_MEMORY_ALLOCATION_ERROR;
 282     }
 283 }
 284 U_CDECL_END
 285
 286 void RegexStaticSets::initGlobals(UErrorCode *status) {
 287     umtx_initOnce(gStaticSetsInitOnce, &initStaticSets, *status);
 288 }
 289
 290 U_NAMESPACE_END
 291 #endif  // !UCONFIG_NO_REGULAR_EXPRESSIONS