scsh-0.6/scsh/rx/re1.c

/* Scheme48 interface to Henry Spencer's Posix regular expression package.
** Copyright (c) 1993, 1994, 1998 by Olin Shivers.
*/

/* Todo:
**   not_eol not_bol support on searchers
**   error code -> err msg
**   regex freeing
**   regexp-string -> regex_t caching
**   make filter_stringvec return an error code.
*/

#include <stdlib.h>
#include <sys/types.h>
#include "../regexp/regex.h"
#include "../cstuff.h"

/* Make sure our exports match up w/the implementation: */
#include "re1.h"

/*
** Compile regexp into a malloc'd struct.
** The flag sm_p is true if we want to compile for submatches.
** On success, store pointer to struct into cr and return 0.
** On failure, free the struct, store NULL into cr, 
**   and return a non-zero error code.
*/

int compile_re(s48_value re_str, int sm_p, regex_t **cr)
{
  // JMG: char *s = &STRING_REF(re_str, 0);
  char *s = s48_extract_string(re_str);
    int len = S48_STRING_LENGTH(re_str);
    int err;
    regex_t *re = Alloc(regex_t);

    if( !re ) return -1;
    
    re->re_endp = s + len;
    err = regcomp(re, s, REG_EXTENDED | REG_PEND
		                      | (sm_p ? 0 : REG_NOSUB));
    if( err ) {Free(re); *cr=0;}
    else *cr=re;

    return err;
    }

/* Do a regex search of RE through string STR, beginning at STR[START].
** - STR is passed as a Scheme value as it is allowed to contain nul bytes.
**
** - trans_vec contains the translation from the user's "virtual" submatches to
**   the actual submatches the engine will report:
**   - trans_vec[i] = #F means user submatch #i is a dead submatch.
**   - trans_vec[i] = j  means user submatch #i corresponds to paren #j in re.
**
**   Indexing fence-posts are a little complicated due to the fact that you
**   get an extra match elt back from the matcher -- match 0 is not a 
**   paren-based *sub*match, but rather the match info for the whole thing.
**
**   So, here is how it works: 
**     length(start_vec) = length(end_vec) = length(trans_vec) + 1
**   because trans_vec doesn't have a translation for submatch 0, which
**   is SRE submatch #0 => Posix submatch #0. For SRE submatch #i (1, 2, ...),
**   we want the submatch associated with Posix paren # trans_vec[i-1].
**
** - MAX_PSM is the maximum paren in which we have submatch interest -- the
**   max element in TRANS_VEC. Any parens after paren #MAX_PSM are just for
**   grouping, not for marking submatches. We only have to allocate MAX_PSM+1
**   elements in the submatch vector we pass into the search engine. If
**   MAX_PSM = -1, then we don't even want the whole-match match bounds, which
**   is really good -- the search engine can really fly in this case.
**
**   If we match, map re's submatches over to the exported start_vec and 
**   end_vec match vectors using trans_vec.
**
** Return 0 on success; #f if no match; non-zero integer error code otherwise.
*/

s48_value re_search(const regex_t *re, s48_value str, int start,
		       s48_value trans_vec, int max_psm,
		       s48_value start_vec, s48_value end_vec)
{
  // JMG: char *s = &STRING_REF(str,0);		/* Passed as a s48_value because */
  char *s = s48_extract_string(str);
  int len = S48_STRING_LENGTH(str);		/* it might contain nul bytes.      */

  int vlen = S48_VECTOR_LENGTH(start_vec);
  int retval;

  regmatch_t static_pmatch[10], *pm;

  /* If max_psm+1 > 10, we can't use static_pmatch. */
  if( max_psm < 10 ) pm = static_pmatch;
  else {
      pm = Malloc(regmatch_t, max_psm+1);/* Add 1 for the whole-match info. */
      if( !pm ) return s48_enter_fixnum(-1);
      }
  
  pm[0].rm_so = start;
  pm[0].rm_eo = len;

  retval = regexec(re, s, max_psm+1, pm, REG_STARTEND);	/* Do it. */

  /* We matched and have match-bound info, so translate it over. */
  if( !retval && max_psm >= 0 ) { 
      int i;

      //JMG:      S48_VECTOR_REF(start_vec,0) = s48_enter_fixnum(pm[0].rm_so);    /* whole-match */
      //S48_VECTOR_REF(end_vec,0)   = s48_enter_fixnum(pm[0].rm_eo);

      S48_VECTOR_SET(start_vec,0, s48_enter_fixnum(pm[0].rm_so));
      S48_VECTOR_SET(end_vec,0, s48_enter_fixnum(pm[0].rm_eo));

      for( i=vlen-1; --i >= 0; ) {			      /* submatches  */
	  s48_value j_scm = S48_VECTOR_REF(trans_vec,i);
	  if( j_scm != S48_FALSE ) {
	      int j = s48_extract_fixnum(j_scm);
	      int k = pm[j].rm_so,
		  l = pm[j].rm_eo;
	      //	JMG      S48_VECTOR_REF(start_vec,i+1) = (k != -1) ? s48_enter_fixnum(k) : S48_FALSE;
	      //S48_VECTOR_REF(end_vec,  i+1) = (l != -1) ? s48_enter_fixnum(l) : S48_FALSE;
	      S48_VECTOR_SET(start_vec,i+1, (k != -1) ? s48_enter_fixnum(k) : S48_FALSE);
	      S48_VECTOR_SET(end_vec,  i+1, (l != -1) ? s48_enter_fixnum(l) : S48_FALSE);
	  }
      }
      }

  if( max_psm >= 10 ) Free(pm);

  if( retval==REG_NOMATCH ) return S48_FALSE;
  if( ! retval ) return S48_TRUE;
  return s48_enter_fixnum(retval);
  }



/* Filter a vector of strings by regexp RE_STR.
** Stringvec is a NULL-terminated vector of strings;
** filter it in-place, copying the survivors back to compact them.
** Put the number of survivors in nummatch.
*/

int filter_stringvec(s48_value re_str, char const **stringvec)
{
  int re_len     = S48_STRING_LENGTH(re_str);/* Passed as a s48_value because */
  //JMG: char *re_chars = &STRING_REF(re_str,0);/* it might contain nul bytes.      */
  char *re_chars = s48_extract_string (re_str);/* it might contain nul bytes.      */
 
 regex_t re;

  char const **p, **q;

  /* REG_NOSUB -- We just want to know if it matches or not. */
  re.re_endp = re_chars + re_len;
  if( regcomp(&re, re_chars, REG_EXTENDED | REG_PEND | REG_NOSUB) ) {
      return 0;
      }

  for(p=q=stringvec; *p; p++) {
      char const *s = *p;
      if( ! regexec(&re, s, 0, 0, 0) ) *q++ = s;
      }

  regfree(&re);
  return q-stringvec;
  }


const char *re_errint2str(int errcode, const regex_t *re)
{
  int size = regerror(errcode, re, 0, 0);
  char *s = Malloc(char,size);
  if(s) regerror(errcode, re, s, size);
  return s;
  }


void free_re(regex_t *re)
{
  regfree(re);
  Free(re);
  }