src/common/regex.cpp

   1 ///////////////////////////////////////////////////////////////////////////////
   2 // Name:        src/common/regex.cpp
   3 // Purpose:     regular expression matching
   4 // Author:      Karsten Ballüder and Vadim Zeitlin
   5 // Modified by:
   6 // Created:     13.07.01
   7 // RCS-ID:      $Id$
   8 // Copyright:   (c) 2000 Karsten Ballüder <ballueder@gmx.net>
   9 //                  2001 Vadim Zeitlin <vadim@wxwindows.org>
  10 // Licence:     wxWindows licence
  11 ///////////////////////////////////////////////////////////////////////////////
  12
  13 // ============================================================================
  14 // declarations
  15 // ============================================================================
  16
  17 // ----------------------------------------------------------------------------
  18 // headers
  19 // ----------------------------------------------------------------------------
  20
  21 // For compilers that support precompilation, includes "wx.h".
  22 #include "wx/wxprec.h"
  23
  24 #ifdef __BORLANDC__
  25     #pragma hdrstop
  26 #endif
  27
  28 #if wxUSE_REGEX
  29
  30 #ifndef WX_PRECOMP
  31     #include "wx/object.h"
  32     #include "wx/string.h"
  33     #include "wx/log.h"
  34     #include "wx/intl.h"
  35 #endif //WX_PRECOMP
  36
  37 // FreeBSD, Watcom and DMars require this, CW doesn't have nor need it.
  38 // Others also don't seem to need it. If you have an error related to
  39 // (not) including <sys/types.h> please report details to
  40 // wx-dev@lists.wxwindows.org
  41 #if defined(__UNIX__) || defined(__WATCOMC__) || defined(__DIGITALMARS__)
  42 #   include <sys/types.h>
  43 #endif
  44
  45 #include <regex.h>
  46 #include "wx/regex.h"
  47
  48 // WXREGEX_USING_BUILTIN    defined when using the built-in regex lib
  49 // WXREGEX_USING_RE_SEARCH  defined when using re_search in the GNU regex lib
  50 // WXREGEX_IF_NEED_LEN()    wrap the len parameter only used with the built-in
  51 //                          or GNU regex
  52 // WXREGEX_CONVERT_TO_MB    defined when the regex lib is using chars and
  53 //                          wxChar is wide, so conversion must be done
  54 // WXREGEX_CHAR(x)          Convert wxChar to wxRegChar
  55 //
  56 #ifdef __REG_NOFRONT
  57 #   define WXREGEX_USING_BUILTIN
  58 #   define WXREGEX_IF_NEED_LEN(x) ,x
  59 #   define WXREGEX_CHAR(x) x
  60 #else
  61 #   ifdef HAVE_RE_SEARCH
  62 #       define WXREGEX_IF_NEED_LEN(x) ,x
  63 #       define WXREGEX_USING_RE_SEARCH
  64 #   else
  65 #       define WXREGEX_IF_NEED_LEN(x)
  66 #   endif
  67 #   if wxUSE_UNICODE
  68 #       define WXREGEX_CONVERT_TO_MB
  69 #   endif
  70 #   define WXREGEX_CHAR(x) wxConvertWX2MB(x)
  71 #   define wx_regfree regfree
  72 #   define wx_regerror regerror
  73 #endif
  74
  75 // ----------------------------------------------------------------------------
  76 // private classes
  77 // ----------------------------------------------------------------------------
  78
  79 #ifndef WXREGEX_USING_RE_SEARCH
  80
  81 // the array of offsets for the matches, the usual POSIX regmatch_t array.
  82 class wxRegExMatches
  83 {
  84 public:
  85     typedef regmatch_t *match_type;
  86
  87     wxRegExMatches(size_t n)        { m_matches = new regmatch_t[n]; }
  88     ~wxRegExMatches()               { delete [] m_matches; }
  89
  90     // we just use casts here because the fields of regmatch_t struct may be 64
  91     // bit but we're limited to size_t in our public API and are not going to
  92     // change it because operating on strings longer than 4GB using it is
  93     // absolutely impractical anyhow
  94     size_t Start(size_t n) const
  95     {
  96         return wx_truncate_cast(size_t, m_matches[n].rm_so);
  97     }
  98
  99     size_t End(size_t n) const
 100     {
 101         return wx_truncate_cast(size_t, m_matches[n].rm_eo);
 102     }
 103
 104     regmatch_t *get() const         { return m_matches; }
 105
 106 private:
 107     regmatch_t *m_matches;
 108 };
 109
 110 #else // WXREGEX_USING_RE_SEARCH
 111
 112 // the array of offsets for the matches, the struct used by the GNU lib
 113 class wxRegExMatches
 114 {
 115 public:
 116     typedef re_registers *match_type;
 117
 118     wxRegExMatches(size_t n)
 119     {
 120         m_matches.num_regs = n;
 121         m_matches.start = new regoff_t[n];
 122         m_matches.end = new regoff_t[n];
 123     }
 124
 125     ~wxRegExMatches()
 126     {
 127         delete [] m_matches.start;
 128         delete [] m_matches.end;
 129     }
 130
 131     size_t Start(size_t n) const    { return m_matches.start[n]; }
 132     size_t End(size_t n) const      { return m_matches.end[n]; }
 133
 134     re_registers *get()             { return &m_matches; }
 135
 136 private:
 137     re_registers m_matches;
 138 };
 139
 140 #endif // WXREGEX_USING_RE_SEARCH
 141
 142 // the character type used by the regular expression engine
 143 #ifndef WXREGEX_CONVERT_TO_MB
 144 typedef wxChar wxRegChar;
 145 #else
 146 typedef char wxRegChar;
 147 #endif
 148
 149 // the real implementation of wxRegEx
 150 class wxRegExImpl
 151 {
 152 public:
 153     // ctor and dtor
 154     wxRegExImpl();
 155     ~wxRegExImpl();
 156
 157     // return true if Compile() had been called successfully
 158     bool IsValid() const { return m_isCompiled; }
 159
 160     // RE operations
 161     bool Compile(const wxString& expr, int flags = 0);
 162     bool Matches(const wxRegChar *str, int flags
 163                  WXREGEX_IF_NEED_LEN(size_t len)) const;
 164     bool GetMatch(size_t *start, size_t *len, size_t index = 0) const;
 165     size_t GetMatchCount() const;
 166     int Replace(wxString *pattern, const wxString& replacement,
 167                 size_t maxMatches = 0) const;
 168
 169 private:
 170     // return the string containing the error message for the given err code
 171     wxString GetErrorMsg(int errorcode, bool badconv) const;
 172
 173     // init the members
 174     void Init()
 175     {
 176         m_isCompiled = false;
 177         m_Matches = NULL;
 178         m_nMatches = 0;
 179     }
 180
 181     // free the RE if compiled
 182     void Free()
 183     {
 184         if ( IsValid() )
 185         {
 186             wx_regfree(&m_RegEx);
 187         }
 188
 189         delete m_Matches;
 190     }
 191
 192     // free the RE if any and reinit the members
 193     void Reinit()
 194     {
 195         Free();
 196         Init();
 197     }
 198
 199     // compiled RE
 200     regex_t         m_RegEx;
 201
 202     // the subexpressions data
 203     wxRegExMatches *m_Matches;
 204     size_t          m_nMatches;
 205
 206     // true if m_RegEx is valid
 207     bool            m_isCompiled;
 208 };
 209
 210
 211 // ============================================================================
 212 // implementation
 213 // ============================================================================
 214
 215 // ----------------------------------------------------------------------------
 216 // wxRegExImpl
 217 // ----------------------------------------------------------------------------
 218
 219 wxRegExImpl::wxRegExImpl()
 220 {
 221     Init();
 222 }
 223
 224 wxRegExImpl::~wxRegExImpl()
 225 {
 226     Free();
 227 }
 228
 229 wxString wxRegExImpl::GetErrorMsg(int errorcode, bool badconv) const
 230 {
 231 #ifdef WXREGEX_CONVERT_TO_MB
 232     // currently only needed when using system library in Unicode mode
 233     if ( badconv )
 234     {
 235         return _("conversion to 8-bit encoding failed");
 236     }
 237 #else
 238     // 'use' badconv to avoid a compiler warning
 239     (void)badconv;
 240 #endif
 241
 242     wxString szError;
 243
 244     // first get the string length needed
 245     int len = wx_regerror(errorcode, &m_RegEx, NULL, 0);
 246     if ( len > 0 )
 247     {
 248         char* szcmbError = new char[++len];
 249
 250         (void)wx_regerror(errorcode, &m_RegEx, szcmbError, len);
 251
 252         szError = wxConvLibc.cMB2WX(szcmbError);
 253         delete [] szcmbError;
 254     }
 255     else // regerror() returned 0
 256     {
 257         szError = _("unknown error");
 258     }
 259
 260     return szError;
 261 }
 262
 263 bool wxRegExImpl::Compile(const wxString& expr, int flags)
 264 {
 265     Reinit();
 266
 267 #ifdef WX_NO_REGEX_ADVANCED
 268 #   define FLAVORS wxRE_BASIC
 269 #else
 270 #   define FLAVORS (wxRE_ADVANCED | wxRE_BASIC)
 271     wxASSERT_MSG( (flags & FLAVORS) != FLAVORS,
 272                   _T("incompatible flags in wxRegEx::Compile") );
 273 #endif
 274     wxASSERT_MSG( !(flags & ~(FLAVORS | wxRE_ICASE | wxRE_NOSUB | wxRE_NEWLINE)),
 275                   _T("unrecognized flags in wxRegEx::Compile") );
 276
 277     // translate our flags to regcomp() ones
 278     int flagsRE = 0;
 279     if ( !(flags & wxRE_BASIC) )
 280 #ifndef WX_NO_REGEX_ADVANCED
 281         if (flags & wxRE_ADVANCED)
 282             flagsRE |= REG_ADVANCED;
 283         else
 284 #endif
 285             flagsRE |= REG_EXTENDED;
 286     if ( flags & wxRE_ICASE )
 287         flagsRE |= REG_ICASE;
 288     if ( flags & wxRE_NOSUB )
 289         flagsRE |= REG_NOSUB;
 290     if ( flags & wxRE_NEWLINE )
 291         flagsRE |= REG_NEWLINE;
 292
 293     // compile it
 294 #ifdef WXREGEX_USING_BUILTIN
 295     bool conv = true;
 296     int errorcode = wx_re_comp(&m_RegEx, expr, expr.length(), flagsRE);
 297 #else
 298     const wxWX2MBbuf conv = expr.mbc_str();
 299     int errorcode = conv ? regcomp(&m_RegEx, conv, flagsRE) : REG_BADPAT;
 300 #endif
 301
 302     if ( errorcode )
 303     {
 304         wxLogError(_("Invalid regular expression '%s': %s"),
 305                    expr.c_str(), GetErrorMsg(errorcode, !conv).c_str());
 306
 307         m_isCompiled = false;
 308     }
 309     else // ok
 310     {
 311         // don't allocate the matches array now, but do it later if necessary
 312         if ( flags & wxRE_NOSUB )
 313         {
 314             // we don't need it at all
 315             m_nMatches = 0;
 316         }
 317         else
 318         {
 319             // we will alloc the array later (only if really needed) but count
 320             // the number of sub-expressions in the regex right now
 321
 322             // there is always one for the whole expression
 323             m_nMatches = 1;
 324
 325             // and some more for bracketed subexperessions
 326             for ( const wxChar *cptr = expr.c_str(); *cptr; cptr++ )
 327             {
 328                 if ( *cptr == _T('\\') )
 329                 {
 330                     // in basic RE syntax groups are inside \(...\)
 331                     if ( *++cptr == _T('(') && (flags & wxRE_BASIC) )
 332                     {
 333                         m_nMatches++;
 334                     }
 335                 }
 336                 else if ( *cptr == _T('(') && !(flags & wxRE_BASIC) )
 337                 {
 338                     // we know that the previous character is not an unquoted
 339                     // backslash because it would have been eaten above, so we
 340                     // have a bare '(' and this indicates a group start for the
 341                     // extended syntax. '(?' is used for extensions by perl-
 342                     // like REs (e.g. advanced), and is not valid for POSIX
 343                     // extended, so ignore them always.
 344                     if ( cptr[1] != _T('?') )
 345                         m_nMatches++;
 346                 }
 347             }
 348         }
 349
 350         m_isCompiled = true;
 351     }
 352
 353     return IsValid();
 354 }
 355
 356 #ifdef WXREGEX_USING_RE_SEARCH
 357
 358 // On GNU, regexec is implemented as a wrapper around re_search. re_search
 359 // requires a length parameter which the POSIX regexec does not have,
 360 // therefore regexec must do a strlen on the search text each time it is
 361 // called. This can drastically affect performance when matching is done in
 362 // a loop along a string, such as during a search and replace. Therefore if
 363 // re_search is detected by configure, it is used directly.
 364 //
 365 static int ReSearch(const regex_t *preg,
 366                     const char *text,
 367                     size_t len,
 368                     re_registers *matches,
 369                     int eflags)
 370 {
 371     regex_t *pattern = wx_const_cast(regex_t*, preg);
 372
 373     pattern->not_bol = (eflags & REG_NOTBOL) != 0;
 374     pattern->not_eol = (eflags & REG_NOTEOL) != 0;
 375     pattern->regs_allocated = REGS_FIXED;
 376
 377     int ret = re_search(pattern, text, len, 0, len, matches);
 378     return ret >= 0 ? 0 : REG_NOMATCH;
 379 }
 380
 381 #endif // WXREGEX_USING_RE_SEARCH
 382
 383 bool wxRegExImpl::Matches(const wxRegChar *str,
 384                           int flags
 385                           WXREGEX_IF_NEED_LEN(size_t len)) const
 386 {
 387     wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
 388
 389     // translate our flags to regexec() ones
 390     wxASSERT_MSG( !(flags & ~(wxRE_NOTBOL | wxRE_NOTEOL)),
 391                   _T("unrecognized flags in wxRegEx::Matches") );
 392
 393     int flagsRE = 0;
 394     if ( flags & wxRE_NOTBOL )
 395         flagsRE |= REG_NOTBOL;
 396     if ( flags & wxRE_NOTEOL )
 397         flagsRE |= REG_NOTEOL;
 398
 399     // allocate matches array if needed
 400     wxRegExImpl *self = wxConstCast(this, wxRegExImpl);
 401     if ( !m_Matches && m_nMatches )
 402     {
 403         self->m_Matches = new wxRegExMatches(m_nMatches);
 404     }
 405
 406     wxRegExMatches::match_type matches = m_Matches ? m_Matches->get() : NULL;
 407
 408     // do match it
 409 #if defined WXREGEX_USING_BUILTIN
 410     int rc = wx_re_exec(&self->m_RegEx, str, len, NULL, m_nMatches, matches, flagsRE);
 411 #elif defined WXREGEX_USING_RE_SEARCH
 412     int rc = str ? ReSearch(&self->m_RegEx, str, len, matches, flagsRE) : REG_BADPAT;
 413 #else
 414     int rc = str ? regexec(&self->m_RegEx, str, m_nMatches, matches, flagsRE) : REG_BADPAT;
 415 #endif
 416
 417     switch ( rc )
 418     {
 419         case 0:
 420             // matched successfully
 421             return true;
 422
 423         default:
 424             // an error occurred
 425             wxLogError(_("Failed to find match for regular expression: %s"),
 426                        GetErrorMsg(rc, !str).c_str());
 427             // fall through
 428
 429         case REG_NOMATCH:
 430             // no match
 431             return false;
 432     }
 433 }
 434
 435 bool wxRegExImpl::GetMatch(size_t *start, size_t *len, size_t index) const
 436 {
 437     wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
 438     wxCHECK_MSG( m_nMatches, false, _T("can't use with wxRE_NOSUB") );
 439     wxCHECK_MSG( m_Matches, false, _T("must call Matches() first") );
 440     wxCHECK_MSG( index < m_nMatches, false, _T("invalid match index") );
 441
 442     if ( start )
 443         *start = m_Matches->Start(index);
 444     if ( len )
 445         *len = m_Matches->End(index) - m_Matches->Start(index);
 446
 447     return true;
 448 }
 449
 450 size_t wxRegExImpl::GetMatchCount() const
 451 {
 452     wxCHECK_MSG( IsValid(), 0, _T("must successfully Compile() first") );
 453     wxCHECK_MSG( m_nMatches, 0, _T("can't use with wxRE_NOSUB") );
 454
 455     return m_nMatches;
 456 }
 457
 458 int wxRegExImpl::Replace(wxString *text,
 459                          const wxString& replacement,
 460                          size_t maxMatches) const
 461 {
 462     wxCHECK_MSG( text, wxNOT_FOUND, _T("NULL text in wxRegEx::Replace") );
 463     wxCHECK_MSG( IsValid(), wxNOT_FOUND, _T("must successfully Compile() first") );
 464
 465     // the input string
 466 #ifndef WXREGEX_CONVERT_TO_MB
 467     const wxChar *textstr = text->c_str();
 468     size_t textlen = text->length();
 469 #else
 470     const wxWX2MBbuf textstr = WXREGEX_CHAR(*text);
 471     if (!textstr)
 472     {
 473         wxLogError(_("Failed to find match for regular expression: %s"),
 474                    GetErrorMsg(0, true).c_str());
 475         return 0;
 476     }
 477     size_t textlen = strlen(textstr);
 478     text->clear();
 479 #endif
 480
 481     // the replacement text
 482     wxString textNew;
 483
 484     // the result, allow 25% extra
 485     wxString result;
 486     result.reserve(5 * textlen / 4);
 487
 488     // attempt at optimization: don't iterate over the string if it doesn't
 489     // contain back references at all
 490     bool mayHaveBackrefs =
 491         replacement.find_first_of(_T("\\&")) != wxString::npos;
 492
 493     if ( !mayHaveBackrefs )
 494     {
 495         textNew = replacement;
 496     }
 497
 498     // the position where we start looking for the match
 499     size_t matchStart = 0;
 500
 501     // number of replacement made: we won't make more than maxMatches of them
 502     // (unless maxMatches is 0 which doesn't limit the number of replacements)
 503     size_t countRepl = 0;
 504
 505     // note that "^" shouldn't match after the first call to Matches() so we
 506     // use wxRE_NOTBOL to prevent it from happening
 507     while ( (!maxMatches || countRepl < maxMatches) &&
 508             Matches(textstr + matchStart,
 509                     countRepl ? wxRE_NOTBOL : 0
 510                     WXREGEX_IF_NEED_LEN(textlen - matchStart)) )
 511     {
 512         // the string possibly contains back references: we need to calculate
 513         // the replacement text anew after each match
 514         if ( mayHaveBackrefs )
 515         {
 516             mayHaveBackrefs = false;
 517             textNew.clear();
 518             textNew.reserve(replacement.length());
 519
 520             for ( const wxChar *p = replacement.c_str(); *p; p++ )
 521             {
 522                 size_t index = (size_t)-1;
 523
 524                 if ( *p == _T('\\') )
 525                 {
 526                     if ( wxIsdigit(*++p) )
 527                     {
 528                         // back reference
 529                         wxChar *end;
 530                         index = (size_t)wxStrtoul(p, &end, 10);
 531                         p = end - 1; // -1 to compensate for p++ in the loop
 532                     }
 533                     //else: backslash used as escape character
 534                 }
 535                 else if ( *p == _T('&') )
 536                 {
 537                     // treat this as "\0" for compatbility with ed and such
 538                     index = 0;
 539                 }
 540
 541                 // do we have a back reference?
 542                 if ( index != (size_t)-1 )
 543                 {
 544                     // yes, get its text
 545                     size_t start, len;
 546                     if ( !GetMatch(&start, &len, index) )
 547                     {
 548                         wxFAIL_MSG( _T("invalid back reference") );
 549
 550                         // just eat it...
 551                     }
 552                     else
 553                     {
 554                         textNew += wxString(textstr + matchStart + start,
 555                                             *wxConvCurrent, len);
 556
 557                         mayHaveBackrefs = true;
 558                     }
 559                 }
 560                 else // ordinary character
 561                 {
 562                     textNew += *p;
 563                 }
 564             }
 565         }
 566
 567         size_t start, len;
 568         if ( !GetMatch(&start, &len) )
 569         {
 570             // we did have match as Matches() returned true above!
 571             wxFAIL_MSG( _T("internal logic error in wxRegEx::Replace") );
 572
 573             return wxNOT_FOUND;
 574         }
 575
 576         // an insurance against implementations that don't grow exponentially
 577         // to ensure building the result takes linear time
 578         if (result.capacity() < result.length() + start + textNew.length())
 579             result.reserve(2 * result.length());
 580
 581 #ifndef WXREGEX_CONVERT_TO_MB
 582         result.append(*text, matchStart, start);
 583 #else
 584         result.append(wxString(textstr + matchStart, *wxConvCurrent, start));
 585 #endif
 586         matchStart += start;
 587         result.append(textNew);
 588
 589         countRepl++;
 590
 591         matchStart += len;
 592     }
 593
 594 #ifndef WXREGEX_CONVERT_TO_MB
 595     result.append(*text, matchStart, wxString::npos);
 596 #else
 597     result.append(wxString(textstr + matchStart, *wxConvCurrent));
 598 #endif
 599     *text = result;
 600
 601     return countRepl;
 602 }
 603
 604 // ----------------------------------------------------------------------------
 605 // wxRegEx: all methods are mostly forwarded to wxRegExImpl
 606 // ----------------------------------------------------------------------------
 607
 608 void wxRegEx::Init()
 609 {
 610     m_impl = NULL;
 611 }
 612
 613 wxRegEx::~wxRegEx()
 614 {
 615     delete m_impl;
 616 }
 617
 618 bool wxRegEx::Compile(const wxString& expr, int flags)
 619 {
 620     if ( !m_impl )
 621     {
 622         m_impl = new wxRegExImpl;
 623     }
 624
 625     if ( !m_impl->Compile(expr, flags) )
 626     {
 627         // error message already given in wxRegExImpl::Compile
 628         delete m_impl;
 629         m_impl = NULL;
 630
 631         return false;
 632     }
 633
 634     return true;
 635 }
 636
 637 bool wxRegEx::Matches(const wxString& str, int flags) const
 638 {
 639     wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
 640
 641     return m_impl->Matches(WXREGEX_CHAR(str), flags
 642                             WXREGEX_IF_NEED_LEN(str.length()));
 643 }
 644
 645 bool wxRegEx::GetMatch(size_t *start, size_t *len, size_t index) const
 646 {
 647     wxCHECK_MSG( IsValid(), false, _T("must successfully Compile() first") );
 648
 649     return m_impl->GetMatch(start, len, index);
 650 }
 651
 652 wxString wxRegEx::GetMatch(const wxString& text, size_t index) const
 653 {
 654     size_t start, len;
 655     if ( !GetMatch(&start, &len, index) )
 656         return wxEmptyString;
 657
 658     return text.Mid(start, len);
 659 }
 660
 661 size_t wxRegEx::GetMatchCount() const
 662 {
 663     wxCHECK_MSG( IsValid(), 0, _T("must successfully Compile() first") );
 664
 665     return m_impl->GetMatchCount();
 666 }
 667
 668 int wxRegEx::Replace(wxString *pattern,
 669                      const wxString& replacement,
 670                      size_t maxMatches) const
 671 {
 672     wxCHECK_MSG( IsValid(), wxNOT_FOUND, _T("must successfully Compile() first") );
 673
 674     return m_impl->Replace(pattern, replacement, maxMatches);
 675 }
 676
 677 #endif // wxUSE_REGEX