regexp.c

操作系统源代码

				}
				if (digit != '}')
				{
					FAIL("Bad characters after \\{");
				}
				else if (to < from || to == 0 || from >= 255)
				{
					FAIL("Invalid range for \\{ \\}");
				}
				re->minlen += from;
			}
			else
#endif
			if (peek != M_SPLAT)
			{
				re->minlen++;
			}

			/* it is okay -- make it prefix instead of postfix */
			ADD_META(build, peek);
#ifndef CRUNCH
			if (peek == M_RANGE)
			{
				*build++ = from;
				*build++ = (to < 255 ? to : 255);
			}
#endif
			

			/* take care of "needfirst" - is this the first char? */
			if (needfirst && peek == M_PLUS && !IS_META(token))
			{
				re->first = token;
			}
			needfirst = 0;

			/* we used "peek" -- need to refill it */
			peek = gettoken(&exp, re);
			if (IS_CLOSURE(peek))
			{
				FAIL("* or \\+ or \\? doubled up");
			}
		}
		else if (!IS_META(token))
		{
			/* normal char is NOT argument of closure */
			if (needfirst)
			{
				re->first = token;
				needfirst = 0;
			}
			re->minlen++;
		}
		else if (token == M_ANY || IS_CLASS(token))
		{
			/* . or [] is NOT argument of closure */
			needfirst = 0;
			re->minlen++;
		}

		/* the "token" character is not closure -- process it normally */
		if (token == M_BEGLINE)
		{
			/* set the BOL flag instead of storing M_BEGLINE */
			re->bol = 1;
		}
		else if (IS_META(token))
		{
			ADD_META(build, token);
		}
		else
		{
			*build++ = token;
		}
	}

	/* end it with a \) which MUST MATCH the opening \( */
	ADD_META(build, M_END(0));
	if (end_sp > 0)
	{
		FAIL("Not enough \\)s");
	}

	return re;
}



/*---------------------------------------------------------------------------*/


/* This function checks for a match between a character and a token which is
 * known to represent a single character.  It returns 0 if they match, or
 * 1 if they don't.
 */
int match1(re, ch, token)
	regexp		*re;
	REG char	ch;
	REG int		token;
{
	if (!ch)
	{
		/* the end of a line can't match any RE of width 1 */
		return 1;
	}
	if (token == M_ANY)
	{
		return 0;
	}
	else if (IS_CLASS(token))
	{
		if (re->program[1 + 32 * (token - M_CLASS(0)) + (ch >> 3)] & (1 << (ch & 7)))
			return 0;
	}
	else if (ch == token || *o_ignorecase && tolower(ch) == tolower(token))
	{
		return 0;
	}
	return 1;
}



/* This function checks characters up to and including the next closure, at
 * which point it does a recursive call to check the rest of it.  This function
 * returns 0 if everything matches, or 1 if something doesn't match.
 */
int match(re, str, prog, here)
	regexp		*re;	/* the regular expression */
	char		*str;	/* the string */
	REG char	*prog;	/* a portion of re->program, an compiled RE */
	REG char	*here;	/* a portion of str, the string to compare it to */
{
	REG int		token;	/* the roken pointed to by prog */
	REG int		nmatched;/* counter, used during closure matching */ 
	REG int		closure;/* the token denoting the type of closure */
	int		from;	/* minimum number of matches in closure */
	int		to;	/* maximum number of matches in closure */

	for (token = GET_META(prog); !IS_CLOSURE(token); prog++, token = GET_META(prog))
	{
		switch (token)
		{
		/*case M_BEGLINE: can't happen; re->bol is used instead */
		  case M_ENDLINE:
			if (*here)
				return 1;
			break;

		  case M_BEGWORD:
			if (here != str &&
			   (here[-1] == '_' || isalnum(here[-1])))
				return 1;
			break;

		  case M_ENDWORD:
			if (here[0] == '_' || isalnum(here[0]))
				return 1;
			break;

		  case M_START(0):
		  case M_START(1):
		  case M_START(2):
		  case M_START(3):
		  case M_START(4):
		  case M_START(5):
		  case M_START(6):
		  case M_START(7):
		  case M_START(8):
		  case M_START(9):
			re->startp[token - M_START(0)] = (char *)here;
			break;

		  case M_END(0):
		  case M_END(1):
		  case M_END(2):
		  case M_END(3):
		  case M_END(4):
		  case M_END(5):
		  case M_END(6):
		  case M_END(7):
		  case M_END(8):
		  case M_END(9):
			re->endp[token - M_END(0)] = (char *)here;
			if (token == M_END(0))
			{
				return 0;
			}
			break;

		  default: /* literal, M_CLASS(n), or M_ANY */
			if (match1(re, *here, token) != 0)
				return 1;
			here++;
		}
	}

	/* C L O S U R E */

	/* step 1: see what we have to match against, and move "prog" to point
	 * to the remainder of the compiled RE.
	 */
	closure = token;
	prog++;
	switch (closure)
	{
	  case M_SPLAT:
		from = 0;
		to = strlen(str);	/* infinity */
		break;

	  case M_PLUS:
		from = 1;
		to = strlen(str);	/* infinity */
		break;

	  case M_QMARK:
		from = 0;
		to = 1;
		break;

#ifndef CRUNCH
	  case M_RANGE:
		from = UCHAR(*prog++);
		to = UCHAR(*prog++);
		if (to == 255)
		{
			to = strlen(str); /* infinity */
		}
		break;
#endif
	}
	token = GET_META(prog);
	prog++;

	/* step 2: see how many times we can match that token against the string */
	for (nmatched = 0;
	     nmatched < to && *here && match1(re, *here, token) == 0;
	     nmatched++, here++)
	{
	}

	/* step 3: try to match the remainder, and back off if it doesn't */
	while (nmatched >= from && match(re, str, prog, here) != 0)
	{
		nmatched--;
		here--;
	}

	/* so how did it work out? */
	if (nmatched >= from)
		return 0;
	return 1;
}



/* This function searches through a string for text that matches an RE. */
int regexec(re, str, bol)
	regexp	*re;	/* the compiled regexp to search for */
	char	*str;	/* the string to search through */
	int	bol;	/* boolean: does str start at the beginning of a line? */
{
	char	*prog;	/* the entry point of re->program */
	int	len;	/* length of the string */
	REG char	*here;

	/* if must start at the beginning of a line, and this isn't, then fail */
	if (re->bol && !bol)
	{
		return 0;
	}

	len = strlen(str);
	prog = re->program + 1 + 32 * re->program[0];

	/* search for the RE in the string */
	if (re->bol)
	{
		/* must occur at BOL */
		if ((re->first
			&& match1(re, *(char *)str, re->first))/* wrong first letter? */
		 || len < re->minlen			/* not long enough? */
		 || match(re, (char *)str, prog, str))	/* doesn't match? */
			return 0;			/* THEN FAIL! */
	}
#ifndef CRUNCH
	else if (!*o_ignorecase)
	{
		/* can occur anywhere in the line, noignorecase */
		for (here = (char *)str;
		     (re->first && re->first != *here)
			|| match(re, (char *)str, prog, here);
		     here++, len--)
		{
			if (len < re->minlen)
				return 0;
		}
	}
#endif
	else
	{
		/* can occur anywhere in the line, ignorecase */
		for (here = (char *)str;
		     (re->first && match1(re, *here, (int)re->first))
			|| match(re, (char *)str, prog, here);
		     here++, len--)
		{
			if (len < re->minlen)
				return 0;
		}
	}

	/* if we didn't fail, then we must have succeeded */
	return 1;
}

/*============================================================================*/
#else /* NO_MAGIC */

regexp *regcomp(exp)
	char	*exp;
{
	char	*src;
	char	*dest;
	regexp	*re;
	int	i;

	/* allocate a big enough regexp structure */
#ifdef lint
	re = (regexp *)0;
#else
	re = (regexp *)malloc((unsigned)(strlen(exp) + 1 + sizeof(struct regexp)));
#endif
	if (!re)
	{
		regerror("Could not malloc a regexp structure");
		return (regexp *)0;
	}

	/* initialize all fields of the structure */
	for (i = 0; i < NSUBEXP; i++)
	{
		re->startp[i] = re->endp[i] = (char *)0;
	}
	re->minlen = 0;
	re->first = 0;
	re->bol = 0;

	/* copy the string into it, translating ^ and $ as needed */
	for (src = exp, dest = re->program + 1; *src; src++)
	{
		switch (*src)
		{
		  case '^':
			if (src == exp)
			{
				re->bol += 1;
			}
			else
			{
				*dest++ = '^';
				re->minlen++;
			}
			break;

		  case '$':
			if (!src[1])
			{
				re->bol += 2;
			}
			else
			{
				*dest++ = '$';
				re->minlen++;
			}
			break;

		  case '\\':
			if (src[1])
			{
				*dest++ = *++src;
				re->minlen++;
			}
			else
			{
				regerror("extra \\ at end of regular expression");
			}
			break;

		  default:
			*dest++ = *src;
			re->minlen++;
		}
	}
	*dest = '\0';

	return re;
}


/* This "helper" function checks for a match at a given location.  It returns
 * 1 if it matches, 0 if it doesn't match here but might match later on in the
 * string, or -1 if it could not possibly match
 */
static int reghelp(prog, string, bolflag)
	struct regexp	*prog;
	char		*string;
	int		bolflag;
{
	char		*scan;
	char		*str;

	/* if ^, then require bolflag */
	if ((prog->bol & 1) && !bolflag)
	{
		return -1;
	}

	/* if it matches, then it will start here */
	prog->startp[0] = string;

	/* compare, possibly ignoring case */
	if (*o_ignorecase)
	{
		for (scan = &prog->program[1]; *scan; scan++, string++)
			if (tolower(*scan) != tolower(*string))
				return *string ? 0 : -1;
	}
	else
	{
		for (scan = &prog->program[1]; *scan; scan++, string++)
			if (*scan != *string)
				return *string ? 0 : -1;
	}

	/* if $, then require string to end here, too */
	if ((prog->bol & 2) && *string)
	{
		return 0;
	}

	/* if we get to here, it matches */
	prog->endp[0] = string;
	return 1;
}



int regexec(prog, string, bolflag)
	struct regexp	*prog;
	char		*string;
	int		bolflag;
{
	int		rc;

	/* keep trying to match it */
	for (rc = reghelp(prog, string, bolflag); rc == 0; rc = reghelp(prog, string, 0))
	{
		string++;
	}

	/* did we match? */
	return rc == 1;
}
#endif