regexp.c

来自「开放源码的编译器open watcom 1.6.0版的源代码」· C语言 代码 · 共 1,192 行 · 第 1/3 页

//
// Copyright (C) 1991 Texas Instruments Incorporated.
//
// Permission is granted to any individual or institution to use, copy, modify,
// and distribute this software, provided that this complete copyright and
// permission notice is maintained, intact, in all copies and supporting
// documentation.
//
// Texas Instruments Incorporated provides this software "as is" without
// express or implied warranty.
//
//
// Created: MNF 06/13/89 -- Initial Design and Implementation
// Updated: MBN 09/08/89 -- Added conditional exception handling
// Updated: MBN 12/15/89 -- Sprinkled "const" qualifiers all over the place!
// Updated: MJF 03/12/90 -- Added group names to RAISE
// Updated: DLS 03/22/91 -- New lite version
//
// This file contains the implementation for the member functions of the CoolRegexp
// class.  The  CoolRegexp class is  defined  in the  CoolRegexp.h   header file.  More
// documentation is also available  in that file.   A  significant part of this
// file is derived directly from other work done on regular expressions.   That
// part is so marked.
//

#ifndef REGEXPH                                 // If CoolRegexp class not define
#include <cool/Regexp.h>                        // Include class specification 
#endif

// ~CoolRegexp -- Destructor for CoolRegexp class (not inline because it's virtual)
// Input:     None
// Output:    None

CoolRegexp::~CoolRegexp () {
  delete [] this->program;
}


// CoolRegexp -- Copy constructor duplicates its values and state
// Input:    Reference to a Regular Expression 
// Output:   None

CoolRegexp::CoolRegexp (const CoolRegexp& rxp) {
  int ind; 
  this->progsize = rxp.progsize;                // Copy regular expression size
  this->program = new char[this->progsize];     // Allocate storage
  for(ind=this->progsize; ind-- != 0;)          // Copy regular expresion
    this->program[ind] = rxp.program[ind];
  this->startp[0] = rxp.startp[0];              // Copy pointers into last
  this->endp[0] = rxp.endp[0];                  // Successful "find" operation
  this->regmust = rxp.regmust;                  // Copy field
  if (rxp.regmust != NULL) {
    char* dum = rxp.program;
    ind = 0;
    while (dum != rxp.regmust) {
      ++dum;
      ++ind;
    }
    this->regmust = this->program + ind;
  }
  this->regstart = rxp.regstart;                // Copy starting index
  this->reganch = rxp.reganch;                  // Copy remaining private data
  this->regmlen = rxp.regmlen;                  // Copy remaining private data
}
     

// operator== -- Overload the equality operator for the CoolRegexp class. Two CoolRegexp
//               are == if their programs are the same.
// Input:        A reference to a regular expression.
// Output:       Boolean TRUE/FALSE

Boolean CoolRegexp::operator== (const CoolRegexp& rxp) const {
  int ind = this->progsize;                     // Get regular expression size
  if (ind != rxp.progsize)                      // If different size regexp
    return FALSE;                               // Return failure
  while(ind-- != 0)                             // Else while still characters
    if(this->program[ind] != rxp.program[ind])  // If regexp are different    
      return FALSE;                             // Return failure             
  return TRUE;                                  // Else same, return success  
}


// deep_equal -- Two CoolRegexp objects that are deep_equal are both == and have
//               the same startp[0] and endp[0] pointers. This means that they
//               have the same compiled regular expression and they last
//               matched on the same identical string.
// Input:        A reference to a CoolRegexp object.
// Output:       Boolean TRUE/FALSE

Boolean CoolRegexp::deep_equal (const CoolRegexp& rxp) const {
  int ind = this->progsize;                     // Get regular expression size
  if (ind != rxp.progsize)                      // If different size regexp
    return FALSE;                               // Return failure
  while(ind-- != 0)                             // Else while still characters
    if(this->program[ind] != rxp.program[ind])  // If regexp are different    
      return FALSE;                             // Return failure             
  return (this->startp[0] == rxp.startp[0] &&   // Else if same start/end ptrs,
          this->endp[0] == rxp.endp[0]);        // Return TRUE
}   

// The remaining code in this file is derived from the  regular expression code
// whose  copyright statement appears  below.  It has been  changed to work
// with the class concepts of C++ and COOL.

/*
 * compile and find 
 *
 *      Copyright (c) 1986 by University of Toronto.
 *      Written by Henry Spencer.  Not derived from licensed software.
 *
 *      Permission is granted to anyone to use this software for any
 *      purpose on any computer system, and to redistribute it freely,
 *      subject to the following restrictions:
 *
 *      1. The author is not responsible for the consequences of use of
 *              this software, no matter how awful, even if they arise
 *              from defects in it.
 *
 *      2. The origin of this software must not be misrepresented, either
 *              by explicit claim or by omission.
 *
 *      3. Altered versions must be plainly marked as such, and must not
 *              be misrepresented as being the original software.
 *
 * Beware that some of this code is subtly aware of the way operator
 * precedence is structured in regular expressions.  Serious changes in
 * regular-expression syntax might require a total rethink.
 */

/*
 * The "internal use only" fields in regexp.h are present to pass info from
 * compile to execute that permits the execute phase to run lots faster on
 * simple cases.  They are:
 *
 * regstart     char that must begin a match; '\0' if none obvious
 * reganch      is the match anchored (at beginning-of-line only)?
 * regmust      string (pointer into program) that match must include, or NULL
 * regmlen      length of regmust string
 *
 * Regstart and reganch permit very fast decisions on suitable starting points
 * for a match, cutting down the work a lot.  Regmust permits fast rejection
 * of lines that cannot possibly match.  The regmust tests are costly enough
 * that compile() supplies a regmust only if the r.e. contains something
 * potentially expensive (at present, the only such thing detected is * or +
 * at the start of the r.e., which can involve a lot of backup).  Regmlen is
 * supplied because the test in find() needs it and compile() is computing
 * it anyway.
 */

/*
 * Structure for regexp "program".  This is essentially a linear encoding
 * of a nondeterministic finite-state machine (aka syntax charts or
 * "railroad normal form" in parsing technology).  Each node is an opcode
 * plus a "next" pointer, possibly plus an operand.  "Next" pointers of
 * all nodes except BRANCH implement concatenation; a "next" pointer with
 * a BRANCH on both ends of it is connecting two alternatives.  (Here we
 * have one of the subtle syntax dependencies:  an individual BRANCH (as
 * opposed to a collection of them) is never concatenated with anything
 * because of operator precedence.)  The operand of some types of node is
 * a literal string; for others, it is a node leading into a sub-FSM.  In
 * particular, the operand of a BRANCH node is the first node of the branch.
 * (NB this is *not* a tree structure:  the tail of the branch connects
 * to the thing following the set of BRANCHes.)  The opcodes are:
 */

// definition   number  opnd?   meaning
#define END     0               // no   End of program.
#define BOL     1               // no   Match "" at beginning of line.
#define EOL     2               // no   Match "" at end of line.
#define ANY     3               // no   Match any one character.
#define ANYOF   4               // str  Match any character in this string.
#define ANYBUT  5               // str  Match any character not in this
                                // string.
#define BRANCH  6               // node Match this alternative, or the
                                // next...
#define BACK    7               // no   Match "", "next" ptr points backward.
#define EXACTLY 8               // str  Match this string.
#define NOTHING 9               // no   Match empty string.
#define STAR    10              // node Match this (simple) thing 0 or more
                                // times.
#define PLUS    11              // node Match this (simple) thing 1 or more
                                // times.
#define OPEN    20              // no   Mark this point in input as start of
                                // #n.
// OPEN+1 is number 1, etc.
#define CLOSE   30              // no   Analogous to OPEN.

/*
 * Opcode notes:
 *
 * BRANCH       The set of branches constituting a single choice are hooked
 *              together with their "next" pointers, since precedence prevents
 *              anything being concatenated to any individual branch.  The
 *              "next" pointer of the last BRANCH in a choice points to the
 *              thing following the whole choice.  This is also where the
 *              final "next" pointer of each individual branch points; each
 *              branch starts with the operand node of a BRANCH node.
 *
 * BACK         Normal "next" pointers all implicitly point forward; BACK
 *              exists to make loop structures possible.
 *
 * STAR,PLUS    '?', and complex '*' and '+', are implemented as circular
 *              BRANCH structures using BACK.  Simple cases (one character
 *              per match) are implemented with STAR and PLUS for speed
 *              and to minimize recursive plunges.
 *
 * OPEN,CLOSE   ...are numbered at compile time.
 */

/*
 * A node is one char of opcode followed by two chars of "next" pointer.
 * "Next" pointers are stored as two 8-bit pieces, high order first.  The
 * value is a positive offset from the opcode of the node containing it.
 * An operand, if any, simply follows the node.  (Note that much of the
 * code generation knows about this implicit relationship.)
 *
 * Using two bytes for the "next" pointer is vast overkill for most things,
 * but allows patterns to get big without disasters.
 */

#define OP(p)           (*(p))
#define NEXT(p)         (((*((p)+1)&0377)<<8) + (*((p)+2)&0377))
#define OPERAND(p)      ((p) + 3)


/*
 * Utility definitions.
 */
#ifndef CHARBITS
#define UCHARAT(p)      ((int)*(const unsigned char*)(p))
#else
#define UCHARAT(p)      ((int)*(p)&CHARBITS)
#endif

#define FAIL(m) { regerror(m); return(NULL); }
#define ISMULT(c)       ((c) == '*' || (c) == '+' || (c) == '?')
#define META    "^$.[()|?+*\\"


/*
 * Flags to be passed up and down.
 */
#define HASWIDTH        01      // Known never to match null string.
#define SIMPLE          02      // Simple enough to be STAR/PLUS operand.
#define SPSTART         04      // Starts with * or +.
#define WORST           0       // Worst case.



/////////////////////////////////////////////////////////////////////////
//
//  COMPILE AND ASSOCIATED FUNCTIONS
//
/////////////////////////////////////////////////////////////////////////


/*
 * Global work variables for compile().
 */
static const char* regparse;    // Input-scan pointer.
static       int   regnpar;     // () count.
static       char  regdummy;
static       char* regcode;     // Code-emit pointer; &regdummy = don't.
static       long  regsize;     // Code size.

/*
 * Forward declarations for compile()'s friends.
 */
#ifndef STATIC
#define STATIC  static
#endif
STATIC       char* reg (int, int*);
STATIC       char* regbranch (int*);
STATIC       char* regpiece (int*);
STATIC       char* regatom (int*);
STATIC       char* regnode (char);
STATIC const char* regnext (register const char*);
STATIC       char* regnext (register char*);
STATIC void        regc (char);
STATIC void        reginsert (char, char*);
STATIC void        regtail (char*, const char*);
STATIC void        regoptail (char*, const char*);

#ifdef STRCSPN
STATIC int strcspn ();
#endif


/*
 - compile - compile a regular expression into internal code
 *
 * We can't allocate space until we know how big the compiled form will be,
 * but we can't compile it (and thus know how big it is) until we've got a
 * place to put the code.  So we cheat:  we compile it twice, once with code
 * generation turned off and size counting turned on, and once "for real".
 * This also means that we don't allocate space until we are sure that the
 * thing really will compile successfully, and we never have to move the
 * code and thus invalidate pointers into it.  (Note that it has to be in
 * one piece because free() must be able to free it all.)
 *
 * Beware that the optimization-preparation code in here knows about some
 * of the structure of the compiled regexp.
 */
void CoolRegexp::compile (const char* exp) {
    register const char* scan;
    register const char* longest;
    register int         len;
             int         flags;

    if (exp == NULL) {
      //RAISE (Error, SYM(CoolRegexp), SYM(No_Expr),
      printf ("CoolRegexp::compile(): No expression supplied.\n");
      abort ();
    }

    // First pass: determine size, legality.
    regparse = exp;
    regnpar = 1;
    regsize = 0L;
    regcode = &regdummy;
    regc(MAGIC);
    reg(0, &flags);
    this->startp[0] = this->endp[0] = this->searchstring = NULL;

    // Small enough for pointer-storage convention? 
    if (regsize >= 32767L) {    // Probably could be 65535L. 
      //RAISE (Error, SYM(CoolRegexp), SYM(Expr_Too_Big),
      printf ("CoolRegexp::compile(): Expression too big.\n");
      abort ();
    }

    // Allocate space. 
    if (this->program != NULL) delete [] this->program;  
    this->program = new char[regsize];
    this->progsize = (int) regsize;

    if (this->program == NULL) {
      //RAISE (Error, SYM(CoolRegexp), SYM(Out_Of_Memory),
      printf ("CoolRegexp::compile(): Out of memory.\n"); 
      abort ();
    }

    // Second pass: emit code.
    regparse = exp;
    regnpar = 1;
    regcode = this->program;
    regc(MAGIC);
    reg(0, &flags);

    // Dig out information for optimizations.
    this->regstart = '\0';              // Worst-case defaults.
    this->reganch = 0;
    this->regmust = NULL;
    this->regmlen = 0;
    scan = this->program + 1;   // First BRANCH.
    if (OP(regnext(scan)) == END) {     // Only one top-level choice.
        scan = OPERAND(scan);

        // Starting-point info.
        if (OP(scan) == EXACTLY)
            this->regstart = *OPERAND(scan);
        else if (OP(scan) == BOL)
            this->reganch++;

         //
         // If there's something expensive in the r.e., find the longest
         // literal string that must appear and make it the regmust.  Resolve
         // ties in favor of later strings, since the regstart check works
         // with the beginning of the r.e. and avoiding duplication
         // strengthens checking.  Not a strong reason, but sufficient in the
         // absence of others. 
         //
        if (flags & SPSTART) {
            longest = NULL;
            len = 0;
            for (; scan != NULL; scan = regnext(scan))
                if (OP(scan) == EXACTLY && strlen(OPERAND(scan)) >= len) {
                    longest = OPERAND(scan);
                    len = strlen(OPERAND(scan));
                }
            this->regmust = longest;
            this->regmlen = len;
        }
    }
}


/*
 - reg - regular expression, i.e. main body or parenthesized thing
 *
 * Caller must absorb opening parenthesis.
 *
 * Combining parenthesis handling with the base level of regular expression
 * is a trifle forced, but the need to tie the tails of the branches to what
 * follows makes it hard to avoid.
 */
static char* reg (int paren, int *flagp) {
    register char* ret;