/*
* regcomp and regexec -- regsub and regerror are elsewhere
*
* 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.
*/
/*****************************************************************************\
* ALTERATIONS *
\*****************************************************************************/
/*
* By: Colas Nahaboo (Bull Research) December 1991, for Klone distrib
* colas@sophia.inria.fr
* marked by string COLAS
* only missed one initialisation of local variable
* and some more parentheses... inspired by tcl modifications.
* supressed all extern char *strchr() declarations, replaced by DECLARE_strchr
* replaced MAGIC by REGMAGIC
* class replaced by cclass for compilation by C++ compilers
* ISMULT replaced by IS_MULT not to match "I.S.M." for my greps
* regexec, regcomp, and regdump prefixed by Kl not to conflict if Klone
* linked with other code including this one
* file renamed as klregexp.c
* macro UCHARAT, CHARBITS replaced by KLCHARBITS
*
* big enhancement: no more fixed limit to the number of parentheses
*/
#include <stdio.h>
#include "klregexp.h" /* COLAS (was: <regexp.h>) */
#include "klmachine.h" /* COLAS (for OS-specific decls) */
/* #include "regmagic.h" */ /* COLAS (included in regexp.h) */
DECLARE_strchr; /* COLAS */
/*
* 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 regcomp() 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 regexec() needs it and regcomp() 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)) /* COLAS */
/* COLAS: The previous line was:
#define NEXT(p) (((*((p)+1)&0377)<<8) + *((p)+2)&0377)
*/
#define OPERAND(p) ((p) + 3)
/*
* See regmagic.h for one further detail of program structure.
*/
/*
* Utility definitions.
*/
#ifndef KLCHARBITS
#define UCHARAT(p) ((int)*(unsigned char *)(p))
#else
#define UCHARAT(p) ((int)*(p)&KLCHARBITS)
#endif
#define FAIL(m) { Kl_regerror(m); return(NULL); }
#define IS_MULT(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. */
/*
* Global work variables for regcomp().
*/
static char *regparse; /* Input-scan pointer. */
static int regnpar; /* () count. */
static char regdummy;
static char *regcode; /* Code-emit pointer; ®dummy = don't. */
static char *regcode_start;
static long regsize; /* Code size. */
static short *o_parnums; /* list of offsets */
static int o_parnums_size;
static short *c_parnums; /* list of offsets */
static int c_parnums_size;
static int parnums; /* are there some parnums declared? */
#ifdef DEBUG
static int nsubexp;
#endif
/*
* Forward declarations for regcomp()'s friends.
*/
#ifndef STATIC
#define STATIC static
#endif
STATIC char *reg();
STATIC char *regbranch();
STATIC char *regpiece();
STATIC char *regatom();
STATIC char *regnode();
STATIC char *regnext();
STATIC void regc();
STATIC void reginsert();
STATIC void regtail();
STATIC void regoptail();
#ifdef STRCSPN
STATIC int strcspn();
#endif
/*
- regcomp - 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.
*/
regexp *
Klregcomp(exp)
char *exp;
{
register regexp *r;
register char *scan;
register char *longest;
register int len;
int flags;
if (exp == NULL)
FAIL("NULL argument");
/* First pass: determine size, legality. */
o_parnums = c_parnums = 0;
o_parnums_size = c_parnums_size = 0;
parnums = 0;
regparse = exp;
regnpar = 1;
regsize = 0L;
regcode_start = regcode = ®dummy;
regc(REGMAGIC);
if (reg(0, &flags) == NULL)
return(NULL);
/* Small enough for pointer-storage convention? */
if (regsize >= 32767L) /* Probably could be 65535L. */
FAIL("regexp too big");
/* Allocate space. */
r = (regexp *)malloc(sizeof(regexp) + (unsigned)regsize);
if (r == NULL)
FAIL("out of space");
r->nsubexp = regnpar; /* inits */
r->startp = (char **) malloc(sizeof(char *) * regnpar * 2);
r->endp = r->startp + regnpar; /* we save one malloc this way */
/* Second pass: emit code. */
regparse = exp;
regnpar = 1;
regcode_start = regcode = r->program;
regc(REGMAGIC);
if (reg(0, &flags) == NULL)
return(NULL);
/* Dig out information for optimizations. */
r->regstart = '\0'; /* Worst-case defaults. */
r->reganch = 0;
r->regmust = NULL;
r->regmlen = 0;
r->o_parnums = o_parnums;
r->c_parnums = c_parnums;
scan = r->program+1; /* First BRANCH. */
if (OP(regnext(scan)) == END) { /* Only one top-level choice. */
scan = OPERAND(scan);
/* Starting-point info. */
if (OP(scan) == EXACTLY)
r->regstart = *OPERAND(scan);
else if (OP(scan) == BOL)
r->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));
}
r->regmust = longest;
r->regmlen = len;
}
}
return(r);
}
/*
- 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(paren, flagp)
int paren; /* Parenthesized? */
int *flagp;
{
register char *ret;
register char *br;
register char *ender;
register int parno = 0; /* COLAS */
int flags;
*flagp = HASWIDTH; /* Tentatively. */
/* Make an OPEN node, if parenthesized. */
if (paren) {
/* COLAS
if (regnpar >= NSUBEXP)
FAIL("too many ()");
*/
parno = regnpar;
regnpar++;
if (parno < NSUBEXP_BUILTIN) {
ret = regnode(OPEN+parno);
} else {
ret = regnode(OPEN);
if (ret)
Klparnum_add(OPEN, parno, ret);
}
} else
ret = NULL;
/* Pick up the branches, linking them together. */
br = regbranch(&flags);
if (br == NULL)
return(NULL);
if (ret != NULL)
regtail(ret, br); /* OPEN -> first. */
else
ret = br;
if (!(flags&HASWIDTH))
*flagp &= ~HASWIDTH;
*flagp |= flags&SPSTART;
while (*regparse == '|') {
regparse++;
br = regbranch(&flags);
if (br == NULL)
return(NULL);
regtail(ret, br); /* BRANCH -> BRANCH. */
if (!(flags&HASWIDTH))
*flagp &= ~HASWIDTH;
*flagp |= flags&SPSTART;
}
/* Make a closing node, and hook it on the end. */
if (paren) {
if (parno < NSUBEXP_BUILTIN) {
ender = regnode(CLOSE+parno);
} else {
ender = regnode(CLOSE);
if (ender)
Klparnum_add(CLOSE, parno, ender);
}
} else {
ender = regnode(END);
}
regtail(ret, ender);
/* Hook the tails of the branches to the closing node. */
for (br = ret; br != NULL; br = regnext(br))
regoptail(br, ender);
/* Check for proper termination. */
if (paren && *regparse++ != ')') {
FAIL("unmatched ()");
} else if (!paren && *regparse != '\0') {
if (*regparse == ')') {
FAIL("unmatched ()");
} else
FAIL("junk on end"); /* "Can't happen". */
/* NOTREACHED */
}
return(ret);
}
/*
- regbranch - one alternative of an | operator
*
* Implements the concatenation operator.
*/
static char *
regbranch(flagp)
int *flagp;
{
register char *ret;
register char *chain;
register char *latest;
int flags;
*flagp = WORST; /* Tentatively. */
ret = regnode(BRANCH);
chain = NULL;
while (*regparse != '\0' && *regparse != '|' && *regparse != ')') {
latest = regpiece(&flags);
if (latest == NULL)
return(NULL);
*flagp |= flags&HASWIDTH;
if (chain == NULL) /* First piece. */
*flagp |= flags&SPSTART;
else
regtail(chain, latest);
chain = latest;
}
if (chain == NULL) /* Loop ran zero times. */
(void) regnode(NOTHING);
return(ret);
}
/*
- regpiece - something followed by possible [*+?]
*
* Note that the branching code sequences used for ? and the general cases
* of * and + are somewhat optimized: they use the same NOTHING node as
* both the endmarker for their branch list and the body of the last branch.
* It might seem that this node could be dispensed with entirely, but the
* endmarker role is not redundant.
*/
static char *
regpiece(flagp)
int *flagp;
{
register char *ret;
register char op;
register char *next;
int flags;
ret = regatom(&flags);
if (ret == NULL)
return(NULL);
op = *regparse;
if (!IS_MULT(op)) {
*flagp = flags;
return(ret);
}
if (!(flags&HASWIDTH) && op != '?')
FAIL("*+ operand could be empty");
*flagp = (op != '+') ? (WORST|SPSTART) : (WORST|HASWIDTH);
if (op == '*' && (flags&SIMPLE))
reginsert(STAR, ret);
else if (op == '*') {
/* Emit x* as (x&|), where & means "self". */
reginsert(BRANCH, ret); /* Either x */
regoptail(ret, regnode(BACK)); /* and loop */
regoptail(ret, ret); /* back */
regtail(ret, regnode(BRANCH)); /* or */
regtail(ret, regnode(NOTHING)); /* null. */
} else if (op == '+' && (flags&SIMPLE))
reginsert(PLUS, ret);
else if (op == '+') {
/* Emit x+ as x(&|), where & means "self". */
next = regnode(BRANCH); /* Either */
regtail(ret, next);
regtail(regnode(BACK), ret); /* loop back */
regtail(next, regnode(BRANCH)); /* or */
regtail(ret, regnode(NOTHING)); /* null. */
} else if (op == '?') {
/* Emit x? as (x|) */
reginsert(BRANCH, ret); /* Either x */
regtail(ret, regnode(BRANCH)); /* or */
next = regnode(NOTHING); /* null. */
regtail(ret, next);
regoptail(ret, next);
}
regparse++;
if (IS_MULT(*regparse))
FAIL("nested *?+");
return(ret);
}
/*
- regatom - the lowest level
*
* Optimization: gobbles an entire sequence of ordinary characters so that
* it can turn them into a single node, which is smaller to store and
* faster to run. Backslashed characters are exceptions, each becoming a
* separate node; the code is simpler that way and it's not worth fixing.
*/
static char *
regatom(flagp)
int *flagp;
{
register char *ret;
int flags;
*flagp = WORST; /* Tentatively. */
switch (*regparse++) {
case '^':
ret = regnode(BOL);
break;
case '$':
ret = regnode(EOL);
break;
case '.':
ret = regnode(ANY);
*flagp |= HASWIDTH|SIMPLE;
break;
case '[': {
register int cclass;
register int classend;
if (*regparse == '^') { /* Complement of range. */
ret = regnode(ANYBUT);
regparse++;
} else
ret = regnode(ANYOF);
if (*regparse == ']' || *regparse == '-')
regc(*regparse++);
while (*regparse != '\0' && *regparse != ']') {
if (*regparse == '-') {
regparse++;
if (*regparse == ']' || *regparse == '\0')
regc('-');
else {
cclass = UCHARAT(regparse-2)+1;
classend = UCHARAT(regparse);
if (cclass > classend+1)
FAIL("invalid [] range");
for (; cclass <= classend; cclass++)
regc(cclass);
regparse++;
}
} else
regc(*regparse++);
}
regc('\0');
if (*regparse != ']')
FAIL("unmatched []");
regparse++;
*flagp |= HASWIDTH|SIMPLE;
}
break;
case '(':
ret = reg(1, &flags);
if (ret == NULL)
return(NULL);
*flagp |= flags&(HASWIDTH|SPSTART);
break;
case '\0':
case '|':
case ')':
FAIL("internal urp"); /* Supposed to be caught earlier. */
break;
case '?':
case '+':
case '*':
FAIL("?+* follows nothing");
break;
case '\\':
if (*regparse == '\0')
FAIL("trailing \\");
ret = regnode(EXACTLY);
regc(*regparse++);
regc('\0');
*flagp |= HASWIDTH|SIMPLE;
break;
default: {
register int len;
register char ender;
regparse--;
len = strcspn(regparse, META);
if (len <= 0)
FAIL("internal disaster");
ender = *(regparse+len);
if (len > 1 && IS_MULT(ender))
len--; /* Back off clear of ?+* operand. */
*flagp |= HASWIDTH;
if (len == 1)
*flagp |= SIMPLE;
ret = regnode(EXACTLY);
while (len > 0) {
regc(*regparse++);
len--;
}
regc('\0');
}
break;
}
return(ret);
}
/*
- regnode - emit a node
*/
static char * /* Location. */
regnode(op)
char op;
{
register char *ret;
register char *ptr;
ret = regcode;
if (ret == ®dummy) {
regsize += 3;
return(ret);
}
ptr = ret;
*ptr++ = op;
*ptr++ = '\0'; /* Null "next" pointer. */
*ptr++ = '\0';
regcode = ptr;
return(ret);
}
/*
- regc - emit (if appropriate) a byte of code
*/
static void
regc(b)
char b;
{
if (regcode != ®dummy)
*regcode++ = b;
else
regsize++;
}
/*
- reginsert - insert an operator in front of already-emitted operand
*
* Means relocating the operand.
*/
static void
reginsert(op, opnd)
char op;
char *opnd;
{
register char *src;
register char *dst;
register char *place;
if (regcode == ®dummy) {
regsize += 3;
return;
}
src = regcode;
regcode += 3;
dst = regcode;
if (parnums) /* update parnums pointers to code */
Klparnum_shift(opnd - regcode_start - 1,
regcode - regcode_start - 1, 3);
while (src > opnd)
*--dst = *--src;
place = opnd; /* Op node, where operand used to be. */
*place++ = op;
*place++ = '\0';
*place++ = '\0';
}
/*
- regtail - set the next-pointer at the end of a node chain
*/
static void
regtail(p, val)
char *p;
char *val;
{
register char *scan;
register char *temp;
register int offset;
if (p == ®dummy)
return;
/* Find last node. */
scan = p;
for (;;) {
temp = regnext(scan);
if (temp == NULL)
break;
scan = temp;
}
if (OP(scan) == BACK)
offset = scan - val;
else
offset = val - scan;
*(scan+1) = (offset>>8)&0377;
*(scan+2) = offset&0377;
}
/*
- regoptail - regtail on operand of first argument; nop if operandless
*/
static void
regoptail(p, val)
char *p;
char *val;
{
/* "Operandless" and "op != BRANCH" are synonymous in practice. */
if (p == NULL || p == ®dummy || OP(p) != BRANCH)
return;
regtail(OPERAND(p), val);
}
/*
* regexec and friends
*/
/*
* Global work variables for regexec().
*/
static char *reginput; /* String-input pointer. */
static char *regbol; /* Beginning of input, for ^ check. */
static char **regstartp; /* Pointer to startp array. */
static char **regendp; /* Ditto for endp. */
/*
* Forwards.
*/
STATIC int regtry();
STATIC int regmatch();
STATIC int regrepeat();
#ifdef DEBUG
int regnarrate = 0;
void Klregdump();
STATIC char *regprop();
#endif
/*
- regexec - match a regexp against a string
*/
int
Klregexec(prog, string)
register regexp *prog;
register char *string;
{
register char *s;
/* Be paranoid... */
if (prog == NULL || string == NULL) {
Kl_regerror("NULL parameter");
return(0);
}
/* Check validity of program. */
if (UCHARAT(prog->program) != REGMAGIC) {
Kl_regerror("corrupted program");
return(0);
}
/* If there is a "must appear" string, look for it. */
if (prog->regmust != NULL) {
s = string;
while ((s = strchr(s, prog->regmust[0])) != NULL) {
if (strncmp(s, prog->regmust, prog->regmlen) == 0)
break; /* Found it. */
s++;
}
if (s == NULL) /* Not present. */
return(0);
}
/* Mark beginning of line for ^ . */
regbol = string;
/* Simplest case: anchored match need be tried only once. */
if (prog->reganch)
return(regtry(prog, string));
/* Messy cases: unanchored match. */
s = string;
if (prog->regstart != '\0')
/* We know what char it must start with. */
while ((s = strchr(s, prog->regstart)) != NULL) {
if (regtry(prog, s))
return(1);
s++;
}
else
/* We don't -- general case. */
do {
if (regtry(prog, s))
return(1);
} while (*s++ != '\0');
/* Failure. */
return(0);
}
/*
- regtry - try match at specific point
*/
static int /* 0 failure, 1 success */
regtry(prog, string)
regexp *prog;
char *string;
{
register int i;
register char **sp;
register char **ep;
reginput = string;
regstartp = prog->startp;
regendp = prog->endp;
sp = prog->startp;
ep = prog->endp;
for (i = prog->nsubexp; i > 0; i--) { /* COLAS: was NSUBEXP */
*sp++ = NULL;
*ep++ = NULL;
}
if (prog->o_parnums) {
regcode_start = prog->program + 1;
o_parnums = prog->o_parnums;
c_parnums = prog->c_parnums;
#ifdef DEBUG
nsubexp = prog->nsubexp;
#endif
}
if (regmatch(prog->program + 1)) {
prog->startp[0] = string;
prog->endp[0] = reginput;
return(1);
} else
return(0);
}
/*
- regmatch - main matching routine
*
* Conceptually the strategy is simple: check to see whether the current
* node matches, call self recursively to see whether the rest matches,
* and then act accordingly. In practice we make some effort to avoid
* recursion, in particular by going through "ordinary" nodes (that don't
* need to know whether the rest of the match failed) by a loop instead of
* by recursion.
*/
static int /* 0 failure, 1 success */
regmatch(prog)
char *prog;
{
register char *scan; /* Current node. */
char *next; /* Next node. */
scan = prog;
#ifdef DEBUG
if (scan != NULL && regnarrate)
fprintf(stderr, "%s(\n", regprop(scan));
#endif
while (scan != NULL) {
#ifdef DEBUG
if (regnarrate)
fprintf(stderr, "%s...\n", regprop(scan));
#endif
next = regnext(scan);
switch (OP(scan)) {
case BOL:
if (reginput != regbol)
return(0);
break;
case EOL:
if (*reginput != '\0')
return(0);
break;
case ANY:
if (*reginput == '\0')
return(0);
reginput++;
break;
case EXACTLY: {
register int len;
register char *opnd;
opnd = OPERAND(scan);
/* Inline the first character, for speed. */
if (*opnd != *reginput)
return(0);
len = strlen(opnd);
if (len > 1 && strncmp(opnd, reginput, len) != 0)
return(0);
reginput += len;
}
break;
case ANYOF:
if (*reginput == '\0' || strchr(OPERAND(scan), *reginput) == NULL)
return(0);
reginput++;
break;
case ANYBUT:
if (*reginput == '\0' || strchr(OPERAND(scan), *reginput) != NULL)
return(0);
reginput++;
break;
case NOTHING:
break;
case BACK:
break;
case OPEN:
case OPEN+1:
case OPEN+2:
case OPEN+3:
case OPEN+4:
case OPEN+5:
case OPEN+6:
case OPEN+7:
case OPEN+8:
case OPEN+9: {
register int no;
register char *save;
if (!(no = OP(scan) - OPEN))
no = Klparnum_find_o(scan - regcode_start);
save = reginput;
if (regmatch(next)) {
/*
* Don't set startp if some later
* invocation of the same parentheses
* already has.
*/
if (regstartp[no] == NULL)
regstartp[no] = save;
return(1);
} else
return(0);
}
break;
case CLOSE:
case CLOSE+1:
case CLOSE+2:
case CLOSE+3:
case CLOSE+4:
case CLOSE+5:
case CLOSE+6:
case CLOSE+7:
case CLOSE+8:
case CLOSE+9: {
register int no;
register char *save;
no = OP(scan) - CLOSE;
if (!(no = OP(scan) - CLOSE))
no = Klparnum_find_c(scan - regcode_start);
save = reginput;
if (regmatch(next)) {
/*
* Don't set endp if some later
* invocation of the same parentheses
* already has.
*/
if (regendp[no] == NULL)
regendp[no] = save;
return(1);
} else
return(0);
}
break;
case BRANCH: {
register char *save;
if (OP(next) != BRANCH) /* No choice. */
next = OPERAND(scan); /* Avoid recursion. */
else {
do {
save = reginput;
if (regmatch(OPERAND(scan)))
return(1);
reginput = save;
scan = regnext(scan);
} while (scan != NULL && OP(scan) == BRANCH);
return(0);
/* NOTREACHED */
}
}
break;
case STAR:
case PLUS: {
register char nextch;
register int no;
register char *save;
register int min;
/*
* Lookahead to avoid useless match attempts
* when we know what character comes next.
*/
nextch = '\0';
if (OP(next) == EXACTLY)
nextch = *OPERAND(next);
min = (OP(scan) == STAR) ? 0 : 1;
save = reginput;
no = regrepeat(OPERAND(scan));
while (no >= min) {
/* If it could work, try it. */
if (nextch == '\0' || *reginput == nextch)
if (regmatch(next))
return(1);
/* Couldn't or didn't -- back up. */
no--;
reginput = save + no;
}
return(0);
}
break;
case END:
return(1); /* Success! */
break;
default:
Kl_regerror("memory corruption");
return(0);
break;
}
scan = next;
}
/*
* We get here only if there's trouble -- normally "case END" is
* the terminating point.
*/
Kl_regerror("corrupted pointers");
return(0);
}
/*
- regrepeat - repeatedly match something simple, report how many
*/
static int
regrepeat(p)
char *p;
{
register int count = 0;
register char *scan;
register char *opnd;
scan = reginput;
opnd = OPERAND(p);
switch (OP(p)) {
case ANY:
count = strlen(scan);
scan += count;
break;
case EXACTLY:
while (*opnd == *scan) {
count++;
scan++;
}
break;
case ANYOF:
while (*scan != '\0' && strchr(opnd, *scan) != NULL) {
count++;
scan++;
}
break;
case ANYBUT:
while (*scan != '\0' && strchr(opnd, *scan) == NULL) {
count++;
scan++;
}
break;
default: /* Oh dear. Called inappropriately. */
Kl_regerror("internal foulup");
count = 0; /* Best compromise. */
break;
}
reginput = scan;
return(count);
}
/*
- regnext - dig the "next" pointer out of a node
*/
static char *
regnext(p)
register char *p;
{
register int offset;
if (p == ®dummy)
return(NULL);
offset = NEXT(p);
if (offset == 0)
return(NULL);
if (OP(p) == BACK)
return(p-offset);
else
return(p+offset);
}
#ifdef DEBUG
STATIC char *regprop();
/*
- regdump - dump a regexp onto stdout in vaguely comprehensible form
*/
void
Klregdump(r)
regexp *r;
{
register char *s;
register char op = EXACTLY; /* Arbitrary non-END op. */
register char *next;
s = r->program + 1;
while (op != END) { /* While that wasn't END last time... */
op = OP(s);
printf("%2d%s", s-r->program, regprop(s)); /* Where, what. */
next = regnext(s);
if (next == NULL) /* Next ptr. */
printf("(0)");
else
printf("(%d)", (s-r->program)+(next-s));
s += 3;
if (op == ANYOF || op == ANYBUT || op == EXACTLY) {
/* Literal string, where present. */
while (*s != '\0') {
putchar(*s);
s++;
}
s++;
}
putchar('\n');
}
/* Header fields of interest. */
if (r->regstart != '\0')
printf("start `%c' ", r->regstart);
if (r->reganch)
printf("anchored ");
if (r->regmust != NULL)
printf("must have \"%s\"", r->regmust);
printf("\n");
}
/*
- regprop - printable representation of opcode
*/
static char *
regprop(op)
char *op;
{
register char *p;
static char buf[50];
(void) strcpy(buf, ":");
switch (OP(op)) {
case BOL:
p = "BOL";
break;
case EOL:
p = "EOL";
break;
case ANY:
p = "ANY";
break;
case ANYOF:
p = "ANYOF";
break;
case ANYBUT:
p = "ANYBUT";
break;
case BRANCH:
p = "BRANCH";
break;
case EXACTLY:
p = "EXACTLY";
break;
case NOTHING:
p = "NOTHING";
break;
case BACK:
p = "BACK";
break;
case END:
p = "END";
break;
case OPEN:
case OPEN+1:
case OPEN+2:
case OPEN+3:
case OPEN+4:
case OPEN+5:
case OPEN+6:
case OPEN+7:
case OPEN+8:
case OPEN+9:
{
int no = OP(op)-OPEN;
if (!no) no = Klparnum_find_o(op - regcode_start);
sprintf(buf+strlen(buf), "OPEN%d", no);
p = NULL;
break;
}
case CLOSE:
case CLOSE+1:
case CLOSE+2:
case CLOSE+3:
case CLOSE+4:
case CLOSE+5:
case CLOSE+6:
case CLOSE+7:
case CLOSE+8:
case CLOSE+9:
{
int no = OP(op)-CLOSE;
if (!no) no = Klparnum_find_c(op - regcode_start);
sprintf(buf+strlen(buf), "CLOSE%d", no);
p = NULL;
break;
}
case STAR:
p = "STAR";
break;
case PLUS:
p = "PLUS";
break;
default:
Kl_regerror("corrupted opcode");
break;
}
if (p != NULL)
(void) strcat(buf, p);
return(buf);
}
#endif
/*
* The following is provided for those people who do not have strcspn() in
* their C libraries. They should get off their butts and do something
* about it; at least one public-domain implementation of those (highly
* useful) string routines has been published on Usenet.
*/
#ifdef STRCSPN
/*
* strcspn - find length of initial segment of s1 consisting entirely
* of characters not from s2
*/
static int
strcspn(s1, s2)
char *s1;
char *s2;
{
register char *scan1;
register char *scan2;
register int count;
count = 0;
for (scan1 = s1; *scan1 != '\0'; scan1++) {
for (scan2 = s2; *scan2 != '\0';) /* ++ moved down. */
if (*scan1 == *scan2++)
return(count);
count++;
}
return(count);
}
#endif
/* COLAS: management of paren levels > 10
* we maintain a p-list of parenleven, offset in program
*/
Klparnum_add(type, parno, ptr)
int type; /* OPEN or CLOSE */
int parno;
char * ptr;
{
if (ptr == ®dummy) /* we are in 1st pass, do nothing */
return;
parno -= NSUBEXP_BUILTIN;
parnums = 1;
if (type == OPEN) {
if (!o_parnums_size) {
o_parnums_size = parno+1;
o_parnums = (short *) malloc(sizeof(short) * o_parnums_size);
} else if (parno >= o_parnums_size) {
o_parnums_size = parno+1;
o_parnums = (short *) realloc(o_parnums,
sizeof(short) * o_parnums_size);
}
o_parnums[parno] = ptr - regcode_start - 1;
} else {
if (!c_parnums_size) {
c_parnums_size = parno+1;
c_parnums = (short *) malloc(sizeof(short) * c_parnums_size);
} else if (parno >= c_parnums_size) {
c_parnums_size = parno+1;
c_parnums = (short *) realloc(c_parnums,
sizeof(short) * c_parnums_size);
}
c_parnums[parno] = ptr - regcode_start - 1;
}
}
int
Klparnum_find_o(offset)
int offset;
{
short *p = o_parnums;
for (; *p != offset; p++) {
#ifdef DEBUG
if (p >= (o_parnums + nsubexp))
stop_if_in_dbx("garbled o_parnums!");
#endif
}
return p - o_parnums + NSUBEXP_BUILTIN;
}
int
Klparnum_find_c(offset)
int offset;
{
short *p = c_parnums;
for (;*p != offset; p++) {
#ifdef DEBUG
if (p >= (c_parnums + nsubexp))
stop_if_in_dbx("garbled c_parnums!");
#endif
}
return p - c_parnums + NSUBEXP_BUILTIN;
}
/* shift parnums pointing to space [start, end] by n bytes */
Klparnum_shift(start, end, n)
int start;
int end;
int n;
{
int parno;
for (parno = 0; parno < o_parnums_size; parno++) {
if (o_parnums[parno] >= start && o_parnums[parno] < end)
o_parnums[parno] += n;
}
for (parno = 0; parno < c_parnums_size; parno++) {
if (c_parnums[parno] >= start && c_parnums[parno] < end)
c_parnums[parno] += n;
}
}
syntax highlighted by Code2HTML, v. 0.9.1