Repository: garyhouston/regexp.old Branch: master Commit: c7db272ecea9 Files: 12 Total size: 51.4 KB Directory structure: gitextract_3p4x8yca/ ├── COPYRIGHT ├── Makefile ├── README ├── regerror.c ├── regexp.3 ├── regexp.c ├── regexp.h ├── regmagic.h ├── regsub.c ├── tests ├── timer.c └── try.c ================================================ FILE CONTENTS ================================================ ================================================ FILE: COPYRIGHT ================================================ Copyright (c) 1986, 1993, 1995 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 in any way, 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 (by explicit claim or omission) as being the original software. 4. This notice must not be removed or altered. ================================================ FILE: Makefile ================================================ # Things you might want to put in ENV: # -DERRAVAIL have utzoo-compatible error() function and friends ENV= # Things you might want to put in TEST: # -DDEBUG debugging hooks # -I. regexp.h from current directory, not /usr/include TEST=-I. # Things you might want to put in PROF: # -pg profiler PROF= CFLAGS=-O $(ENV) $(TEST) $(PROF) LDFLAGS=$(PROF) LIB=libregexp.a OBJ=regexp.o regsub.o regerror.o TMP=dtr.tmp default: r try: try.o $(LIB) cc $(LDFLAGS) try.o $(LIB) -o try # Making timer will probably require putting stuff in $(PROF) and then # recompiling everything; the following is just the final stage. timer: timer.o $(LIB) cc $(LDFLAGS) timer.o $(LIB) -o timer timer.o: timer.c timer.t.h timer.t.h: tests sed 's/ /","/g;s/\\/&&/g;s/.*/{"&"},/' tests >timer.t.h # Regression test. r: try tests ./try $@ ch.sml: ch $(BITS) smlize splitfigs splitfigs ch | soelim | smlize >$@ fig0 fig1 fig2: ch splitfigs splitfigs ch >/dev/null f: fig0 fig1 fig2 figs groff -Tps -s $(OPT) figs | lpr fig1.ps: fig0 fig1 ( cat fig0 ; echo ".LP" ; cat fig1 ) | groff -Tps $(OPT) >$@ fig2.ps: fig0 fig2 ( cat fig0 ; echo ".LP" ; cat fig2 ) | groff -Tps $(OPT) >$@ fp: fig1.ps fig2.ps r.1: regexp.c splitter splitter regexp.c rs.1: regsub.c splitter splitter regsub.c re.1: regerror.c splitter splitter regerror.c rm.h: regmagic.h splitter splitter regmagic.h re.h: regexp.h splitter splitter regexp.h PLAIN=COPYRIGHT README Makefile regexp.3 try.c timer.c tests FIX=regexp.h regexp.c regsub.c regerror.c regmagic.h DTR=$(PLAIN) $(FIX) dtr: r $(DTR) rm -rf $(TMP) mkdir $(TMP) cp $(PLAIN) $(TMP) for f in $(FIX) ; do normalize $$f >$(TMP)/$$f ; done ( cd $(TMP) ; makedtr $(DTR) ) >bookregexp.shar ( cd $(TMP) ; tar -cvf ../bookregexp.tar $(DTR) ) rm -rf $(TMP) ch.ps: ch Makefile $(BITS) groff -Tps $(OPT) ch >$@ copy: ch.soe ch.sml fp makedtr REMARKS ch.sml fig*.ps ch.soe >$@ go: copy dtr ================================================ FILE: README ================================================ This is a revision of my well-known regular-expression package, regexp(3). It gives C programs the ability to use egrep-style regular expressions, and does it in a much cleaner fashion than the analogous routines in SysV. It is not, alas, fully POSIX.2-compliant; that is hard. (I'm working on a full reimplementation that will do that.) This version is the one which is examined and explained in one chapter of "Software Solutions in C" (Dale Schumacher, ed.; AP Professional 1994; ISBN 0-12-632360-7), plus a couple of insignificant updates, plus one significant bug fix (done 10 Nov 1995). Although this package was inspired by the Bell V8 regexp(3), this implementation is *NOT* AT&T/Bell code, and is not derived from licensed software. Even though U of T is a V8 licensee. This software is based on a V8 manual page sent to me by Dennis Ritchie (the manual page enclosed here is a complete rewrite and hence is not covered by AT&T copyright). I admit to some familiarity with regular-expression implementations of the past, but the only one that this code traces any ancestry to is the one published in Kernighan & Plauger's "Software Tools" (from which this one draws ideas but not code). Simplistically: put this stuff into a source directory, inspect Makefile for compilation options that need changing to suit your local environment, and then do "make". This compiles the regexp(3) functions, builds a library containing them, compiles a test program, and runs a large set of regression tests. If there are no complaints, then put regexp.h into /usr/include, add regexp.o, regsub.o, and regerror.o into your C library (or put libre.a into /usr/lib), and install regexp.3 (perhaps with slight modifications) in your manual-pages directory. The files are: COPYRIGHT copyright notice README this text Makefile instructions to make everything regexp.3 manual page regexp.h header file, for /usr/include regexp.c source for regcomp() and regexec() regsub.c source for regsub() regerror.c source for default regerror() regmagic.h internal header file try.c source for test program timer.c source for timing program tests test list for try and timer This implementation uses nondeterministic automata rather than the deterministic ones found in some other implementations, which makes it simpler, smaller, and faster at compiling regular expressions, but slower at executing them. Many users have found the speed perfectly adequate, although replacing the insides of egrep with this code would be a mistake. This stuff should be pretty portable, given an ANSI C compiler and appropriate option settings. There are no "reserved" char values except for NUL, and no special significance is attached to the top bit of chars. The string(3) functions are used a fair bit, on the grounds that they are probably faster than coding the operations in line. Some attempts at code tuning have been made, but this is invariably a bit machine-specific. This distribution lives at ftp://ftp.zoo.toronto.edu/pub/bookregexp.{tar|shar} at present. ================================================ FILE: regerror.c ================================================ /* * regerror */ #include #include void regerror(s) char *s; { #ifdef ERRAVAIL error("regexp: %s", s); #else fprintf(stderr, "regexp(3): %s\n", s); exit(EXIT_FAILURE); #endif /* NOTREACHED */ } ================================================ FILE: regexp.3 ================================================ .TH REGEXP 3 "5 Sept 1996" .SH NAME regcomp, regexec, regsub, regerror \- regular expression handler .SH SYNOPSIS .ft B .nf #include regexp *regcomp(exp) const char *exp; int regexec(prog, string) regexp *prog; const char *string; void regsub(prog, source, dest) const regexp *prog; const char *source; char *dest; void regerror(msg) char *msg; .SH DESCRIPTION These functions implement .IR egrep (1)-style regular expressions and supporting facilities. .PP .I Regcomp compiles a regular expression into a structure of type .IR regexp , and returns a pointer to it. The space has been allocated using .IR malloc (3) and may be released by .IR free . .PP .I Regexec matches a NUL-terminated \fIstring\fR against the compiled regular expression in \fIprog\fR. It returns 1 for success and 0 for failure, and adjusts the contents of \fIprog\fR's \fIstartp\fR and \fIendp\fR (see below) accordingly. .PP The members of a .I regexp structure include at least the following (not necessarily in order): .PP .RS char *startp[NSUBEXP]; .br char *endp[NSUBEXP]; .RE .PP where .I NSUBEXP is defined (as 10) in the header file. Once a successful \fIregexec\fR has been done using the \fIregexp\fR, each \fIstartp\fR-\fIendp\fR pair describes one substring within the \fIstring\fR, with the \fIstartp\fR pointing to the first character of the substring and the \fIendp\fR pointing to the first character following the substring. The 0th substring is the substring of \fIstring\fR that matched the whole regular expression. The others are those substrings that matched parenthesized expressions within the regular expression, with parenthesized expressions numbered in left-to-right order of their opening parentheses. If a parenthesized expression does not participate in the match at all, its \fIstartp\fR and \fIendp\fR are NULL. .PP .I Regsub copies \fIsource\fR to \fIdest\fR, making substitutions according to the most recent \fIregexec\fR performed using \fIprog\fR. Each instance of `&' in \fIsource\fR is replaced by the substring indicated by \fIstartp\fR[\fI0\fR] and \fIendp\fR[\fI0\fR]. Each instance of `\e\fIn\fR', where \fIn\fR is a digit, is replaced by the substring indicated by \fIstartp\fR[\fIn\fR] and \fIendp\fR[\fIn\fR]. To get a literal `&' or `\e\fIn\fR' into \fIdest\fR, prefix it with `\e'; to get a literal `\e' preceding `&' or `\e\fIn\fR', prefix it with another `\e'. .PP .I Regerror is called whenever an error is detected in \fIregcomp\fR, \fIregexec\fR, or \fIregsub\fR. The default \fIregerror\fR writes the string \fImsg\fR, with a suitable indicator of origin, on the standard error output and invokes \fIexit\fR(2). .I Regerror can be replaced by the user if other actions are desirable. .SH "REGULAR EXPRESSION SYNTAX" A regular expression is zero or more \fIbranches\fR, separated by `|'. It matches anything that matches one of the branches. .PP A branch is zero or more \fIpieces\fR, concatenated. It matches a match for the first, followed by a match for the second, etc. .PP A piece is an \fIatom\fR possibly followed by `*', `+', or `?'. An atom followed by `*' matches a sequence of 0 or more matches of the atom. An atom followed by `+' matches a sequence of 1 or more matches of the atom. An atom followed by `?' matches a match of the atom, or the null string. .PP An atom is a regular expression in parentheses (matching a match for the regular expression), a \fIrange\fR (see below), `.' (matching any single character), `^' (matching the null string at the beginning of the input string), `$' (matching the null string at the end of the input string), a `\e' followed by a single character (matching that character), or a single character with no other significance (matching that character). .PP A \fIrange\fR is a sequence of characters enclosed in `[]'. It normally matches any single character from the sequence. If the sequence begins with `^', it matches any single character \fInot\fR from the rest of the sequence. If two characters in the sequence are separated by `\-', this is shorthand for the full list of ASCII characters between them (e.g. `[0-9]' matches any decimal digit). To include a literal `]' in the sequence, make it the first character (following a possible `^'). To include a literal `\-', make it the first or last character. .SH AMBIGUITY If a regular expression could match two different parts of the input string, it will match the one which begins earliest. If both begin in the same place but match different lengths, or match the same length in different ways, life gets messier, as follows. .PP In general, the possibilities in a list of branches are considered in left-to-right order, the possibilities for `*', `+', and `?' are considered longest-first, nested constructs are considered from the outermost in, and concatenated constructs are considered leftmost-first. The match that will be chosen is the one that uses the earliest possibility in the first choice that has to be made. If there is more than one choice, the next will be made in the same manner (earliest possibility) subject to the decision on the first choice. And so forth. .PP For example, `(ab|a)b*c' could match `abc' in one of two ways. The first choice is between `ab' and `a'; since `ab' is earlier, and does lead to a successful overall match, it is chosen. Since the `b' is already spoken for, the `b*' must match its last possibility\(emthe empty string\(emsince it must respect the earlier choice. .PP In the particular case where the regular expression does not use `|' and does not apply `*', `+', or `?' to parenthesized subexpressions, the net effect is that the longest possible match will be chosen. So `ab*', presented with `xabbbby', will match `abbbb'. Note that if `ab*' is tried against `xabyabbbz', it will match `ab' just after `x', due to the begins-earliest rule. (In effect, the decision on where to start the match is the first choice to be made, hence subsequent choices must respect it even if this leads them to less-preferred alternatives.) .SH SEE ALSO egrep(1), expr(1) .SH DIAGNOSTICS \fIRegcomp\fR returns NULL for a failure (\fIregerror\fR permitting), where failures are syntax errors, exceeding implementation limits, or applying `+' or `*' to a possibly-null operand. .SH HISTORY This is a revised version. Both code and manual page were originally written by Henry Spencer at University of Toronto. They are intended to be compatible with the Bell V8 \fIregexp\fR(3), but are not derived from Bell code. .SH BUGS Empty branches and empty regular expressions are not portable to other, otherwise-similar, implementations. .PP The ban on applying `*' or `+' to a possibly-null operand is an artifact of the simplistic implementation. .PP The match-choice rules are complex. A simple ``longest match'' rule would be preferable, but is harder to implement. .PP Although there is a general similarity to POSIX.2 ``extended'' regular expressions, neither the regular-expression syntax nor the programming interface is an exact match. .PP Due to emphasis on compactness and simplicity, it's not strikingly fast. It does give some attention to handling simple cases quickly. ================================================ FILE: regexp.c ================================================ /* * regcomp and regexec -- regsub and regerror are elsewhere */ #include #include #include #include #include "regmagic.h" /* * 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 beginning of line. */ #define EOL 2 /* no Match end of line. */ #define ANY 3 /* no Match any character. */ #define ANYOF 4 /* str Match any of these. */ #define ANYBUT 5 /* str Match any but one of these. */ #define BRANCH 6 /* node Match this, or the next..\&. */ #define BACK 7 /* no "next" ptr points backward. */ #define EXACTLY 8 /* str Match this string. */ #define NOTHING 9 /* no Match empty string. */ #define STAR 10 /* node Match this 0 or more times. */ #define PLUS 11 /* node Match this 1 or more times. */ #define OPEN 20 /* no Sub-RE starts here. */ /* 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)&0177)<<8) + (*((p)+2)&0377)) #define OPERAND(p) ((p) + 3) /* * See regmagic.h for one further detail of program structure. */ /* * Utility definitions. */ #define FAIL(m) { regerror(m); return(NULL); } #define ISREPN(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. */ /* * Work-variable struct for regcomp(). */ struct comp { char *regparse; /* Input-scan pointer. */ int regnpar; /* () count. */ char *regcode; /* Code-emit pointer; ®dummy = don't. */ char regdummy[3]; /* NOTHING, 0 next ptr */ long regsize; /* Code size. */ }; #define EMITTING(cp) ((cp)->regcode != (cp)->regdummy) /* * Forward declarations for regcomp()'s friends. */ static char *reg(struct comp *cp, int paren, int *flagp); static char *regbranch(struct comp *cp, int *flagp); static char *regpiece(struct comp *cp, int *flagp); static char *regatom(struct comp *cp, int *flagp); static char *regnode(struct comp *cp, int op); static char *regnext(char *node); static void regc(struct comp *cp, int c); static void reginsert(struct comp *cp, int op, char *opnd); static void regtail(struct comp *cp, char *p, char *val); static void regoptail(struct comp *cp, char *p, char *val); /* - 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 * regcomp(exp) const char *exp; { register regexp *r; register char *scan; int flags; struct comp co; if (exp == NULL) FAIL("NULL argument to regcomp"); /* First pass: determine size, legality. */ co.regparse = (char *)exp; co.regnpar = 1; co.regsize = 0L; co.regdummy[0] = NOTHING; co.regdummy[1] = co.regdummy[2] = 0; co.regcode = co.regdummy; regc(&co, MAGIC); if (reg(&co, 0, &flags) == NULL) return(NULL); /* Small enough for pointer-storage convention? */ if (co.regsize >= 0x7fffL) /* Probably could be 0xffffL. */ FAIL("regexp too big"); /* Allocate space. */ r = (regexp *)malloc(sizeof(regexp) + (size_t)co.regsize); if (r == NULL) FAIL("out of space"); /* Second pass: emit code. */ co.regparse = (char *)exp; co.regnpar = 1; co.regcode = r->program; regc(&co, MAGIC); if (reg(&co, 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; 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 = 1; /* * 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) { register char *longest = NULL; register size_t 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 = (int)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(cp, paren, flagp) register struct comp *cp; int paren; /* Parenthesized? */ int *flagp; { register char *ret; register char *br; register char *ender; register int parno; int flags; *flagp = HASWIDTH; /* Tentatively. */ if (paren) { /* Make an OPEN node. */ if (cp->regnpar >= NSUBEXP) FAIL("too many ()"); parno = cp->regnpar; cp->regnpar++; ret = regnode(cp, OPEN+parno); } /* Pick up the branches, linking them together. */ br = regbranch(cp, &flags); if (br == NULL) return(NULL); if (paren) regtail(cp, ret, br); /* OPEN -> first. */ else ret = br; *flagp &= ~(~flags&HASWIDTH); /* Clear bit if bit 0. */ *flagp |= flags&SPSTART; while (*cp->regparse == '|') { cp->regparse++; br = regbranch(cp, &flags); if (br == NULL) return(NULL); regtail(cp, ret, br); /* BRANCH -> BRANCH. */ *flagp &= ~(~flags&HASWIDTH); *flagp |= flags&SPSTART; } /* Make a closing node, and hook it on the end. */ ender = regnode(cp, (paren) ? CLOSE+parno : END); regtail(cp, ret, ender); /* Hook the tails of the branches to the closing node. */ for (br = ret; br != NULL; br = regnext(br)) regoptail(cp, br, ender); /* Check for proper termination. */ if (paren && *cp->regparse++ != ')') { FAIL("unterminated ()"); } else if (!paren && *cp->regparse != '\0') { if (*cp->regparse == ')') { FAIL("unmatched ()"); } else FAIL("internal error: junk on end"); /* NOTREACHED */ } return(ret); } /* - regbranch - one alternative of an | operator * * Implements the concatenation operator. */ static char * regbranch(cp, flagp) register struct comp *cp; int *flagp; { register char *ret; register char *chain; register char *latest; int flags; register int c; *flagp = WORST; /* Tentatively. */ ret = regnode(cp, BRANCH); chain = NULL; while ((c = *cp->regparse) != '\0' && c != '|' && c != ')') { latest = regpiece(cp, &flags); if (latest == NULL) return(NULL); *flagp |= flags&HASWIDTH; if (chain == NULL) /* First piece. */ *flagp |= flags&SPSTART; else regtail(cp, chain, latest); chain = latest; } if (chain == NULL) /* Loop ran zero times. */ (void) regnode(cp, 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(cp, flagp) register struct comp *cp; int *flagp; { register char *ret; register char op; register char *next; int flags; ret = regatom(cp, &flags); if (ret == NULL) return(NULL); op = *cp->regparse; if (!ISREPN(op)) { *flagp = flags; return(ret); } if (!(flags&HASWIDTH) && op != '?') FAIL("*+ operand could be empty"); switch (op) { case '*': *flagp = WORST|SPSTART; break; case '+': *flagp = WORST|SPSTART|HASWIDTH; break; case '?': *flagp = WORST; break; } if (op == '*' && (flags&SIMPLE)) reginsert(cp, STAR, ret); else if (op == '*') { /* Emit x* as (x&|), where & means "self". */ reginsert(cp, BRANCH, ret); /* Either x */ regoptail(cp, ret, regnode(cp, BACK)); /* and loop */ regoptail(cp, ret, ret); /* back */ regtail(cp, ret, regnode(cp, BRANCH)); /* or */ regtail(cp, ret, regnode(cp, NOTHING)); /* null. */ } else if (op == '+' && (flags&SIMPLE)) reginsert(cp, PLUS, ret); else if (op == '+') { /* Emit x+ as x(&|), where & means "self". */ next = regnode(cp, BRANCH); /* Either */ regtail(cp, ret, next); regtail(cp, regnode(cp, BACK), ret); /* loop back */ regtail(cp, next, regnode(cp, BRANCH)); /* or */ regtail(cp, ret, regnode(cp, NOTHING)); /* null. */ } else if (op == '?') { /* Emit x? as (x|) */ reginsert(cp, BRANCH, ret); /* Either x */ regtail(cp, ret, regnode(cp, BRANCH)); /* or */ next = regnode(cp, NOTHING); /* null. */ regtail(cp, ret, next); regoptail(cp, ret, next); } cp->regparse++; if (ISREPN(*cp->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(cp, flagp) register struct comp *cp; int *flagp; { register char *ret; int flags; *flagp = WORST; /* Tentatively. */ switch (*cp->regparse++) { case '^': ret = regnode(cp, BOL); break; case '$': ret = regnode(cp, EOL); break; case '.': ret = regnode(cp, ANY); *flagp |= HASWIDTH|SIMPLE; break; case '[': { register int range; register int rangeend; register int c; if (*cp->regparse == '^') { /* Complement of range. */ ret = regnode(cp, ANYBUT); cp->regparse++; } else ret = regnode(cp, ANYOF); if ((c = *cp->regparse) == ']' || c == '-') { regc(cp, c); cp->regparse++; } while ((c = *cp->regparse++) != '\0' && c != ']') { if (c != '-') regc(cp, c); else if ((c = *cp->regparse) == ']' || c == '\0') regc(cp, '-'); else { range = (unsigned char)*(cp->regparse-2); rangeend = (unsigned char)c; if (range > rangeend) FAIL("invalid [] range"); for (range++; range <= rangeend; range++) regc(cp, range); cp->regparse++; } } regc(cp, '\0'); if (c != ']') FAIL("unmatched []"); *flagp |= HASWIDTH|SIMPLE; break; } case '(': ret = reg(cp, 1, &flags); if (ret == NULL) return(NULL); *flagp |= flags&(HASWIDTH|SPSTART); break; case '\0': case '|': case ')': /* supposed to be caught earlier */ FAIL("internal error: \\0|) unexpected"); break; case '?': case '+': case '*': FAIL("?+* follows nothing"); break; case '\\': if (*cp->regparse == '\0') FAIL("trailing \\"); ret = regnode(cp, EXACTLY); regc(cp, *cp->regparse++); regc(cp, '\0'); *flagp |= HASWIDTH|SIMPLE; break; default: { register size_t len; register char ender; cp->regparse--; len = strcspn(cp->regparse, META); if (len == 0) FAIL("internal error: strcspn 0"); ender = *(cp->regparse+len); if (len > 1 && ISREPN(ender)) len--; /* Back off clear of ?+* operand. */ *flagp |= HASWIDTH; if (len == 1) *flagp |= SIMPLE; ret = regnode(cp, EXACTLY); for (; len > 0; len--) regc(cp, *cp->regparse++); regc(cp, '\0'); break; } } return(ret); } /* - regnode - emit a node */ static char * /* Location. */ regnode(cp, op) register struct comp *cp; char op; { register char *const ret = cp->regcode; register char *ptr; if (!EMITTING(cp)) { cp->regsize += 3; return(ret); } ptr = ret; *ptr++ = op; *ptr++ = '\0'; /* Null next pointer. */ *ptr++ = '\0'; cp->regcode = ptr; return(ret); } /* - regc - emit (if appropriate) a byte of code */ static void regc(cp, b) register struct comp *cp; char b; { if (EMITTING(cp)) *cp->regcode++ = b; else cp->regsize++; } /* - reginsert - insert an operator in front of already-emitted operand * * Means relocating the operand. */ static void reginsert(cp, op, opnd) register struct comp *cp; char op; char *opnd; { register char *place; if (!EMITTING(cp)) { cp->regsize += 3; return; } (void) memmove(opnd+3, opnd, (size_t)(cp->regcode - opnd)); cp->regcode += 3; 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(cp, p, val) register struct comp *cp; char *p; char *val; { register char *scan; register char *temp; register int offset; if (!EMITTING(cp)) return; /* Find last node. */ for (scan = p; (temp = regnext(scan)) != NULL; scan = temp) continue; offset = (OP(scan) == BACK) ? scan - val : val - scan; *(scan+1) = (offset>>8)&0177; *(scan+2) = offset&0377; } /* - regoptail - regtail on operand of first argument; nop if operandless */ static void regoptail(cp, p, val) register struct comp *cp; char *p; char *val; { /* "Operandless" and "op != BRANCH" are synonymous in practice. */ if (!EMITTING(cp) || OP(p) != BRANCH) return; regtail(cp, OPERAND(p), val); } /* * regexec and friends */ /* * Work-variable struct for regexec(). */ struct exec { char *reginput; /* String-input pointer. */ char *regbol; /* Beginning of input, for ^ check. */ char **regstartp; /* Pointer to startp array. */ char **regendp; /* Ditto for endp. */ }; /* * Forwards. */ static int regtry(struct exec *ep, regexp *rp, char *string); static int regmatch(struct exec *ep, char *prog); static size_t regrepeat(struct exec *ep, char *node); #ifdef DEBUG int regnarrate = 0; void regdump(); static char *regprop(); #endif /* - regexec - match a regexp against a string */ int regexec(prog, str) register regexp *prog; const char *str; { register char *string = (char *)str; /* avert const poisoning */ register char *s; struct exec ex; /* Be paranoid. */ if (prog == NULL || string == NULL) { regerror("NULL argument to regexec"); return(0); } /* Check validity of program. */ if ((unsigned char)*prog->program != MAGIC) { regerror("corrupted regexp"); return(0); } /* If there is a "must appear" string, look for it. */ if (prog->regmust != NULL && strstr(string, prog->regmust) == NULL) return(0); /* Mark beginning of line for ^ . */ ex.regbol = string; ex.regstartp = prog->startp; ex.regendp = prog->endp; /* Simplest case: anchored match need be tried only once. */ if (prog->reganch) return(regtry(&ex, prog, string)); /* Messy cases: unanchored match. */ if (prog->regstart != '\0') { /* We know what char it must start with. */ for (s = string; s != NULL; s = strchr(s+1, prog->regstart)) if (regtry(&ex, prog, s)) return(1); return(0); } else { /* We don't -- general case. */ for (s = string; !regtry(&ex, prog, s); s++) if (*s == '\0') return(0); return(1); } /* NOTREACHED */ } /* - regtry - try match at specific point */ static int /* 0 failure, 1 success */ regtry(ep, prog, string) register struct exec *ep; regexp *prog; char *string; { register int i; register char **stp; register char **enp; ep->reginput = string; stp = prog->startp; enp = prog->endp; for (i = NSUBEXP; i > 0; i--) { *stp++ = NULL; *enp++ = NULL; } if (regmatch(ep, prog->program + 1)) { prog->startp[0] = string; prog->endp[0] = ep->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(ep, prog) register struct exec *ep; char *prog; { register char *scan; /* Current node. */ char *next; /* Next node. */ #ifdef DEBUG if (prog != NULL && regnarrate) fprintf(stderr, "%s(\n", regprop(prog)); #endif for (scan = prog; scan != NULL; scan = next) { #ifdef DEBUG if (regnarrate) fprintf(stderr, "%s...\n", regprop(scan)); #endif next = regnext(scan); switch (OP(scan)) { case BOL: if (ep->reginput != ep->regbol) return(0); break; case EOL: if (*ep->reginput != '\0') return(0); break; case ANY: if (*ep->reginput == '\0') return(0); ep->reginput++; break; case EXACTLY: { register size_t len; register char *const opnd = OPERAND(scan); /* Inline the first character, for speed. */ if (*opnd != *ep->reginput) return(0); len = strlen(opnd); if (len > 1 && strncmp(opnd, ep->reginput, len) != 0) return(0); ep->reginput += len; break; } case ANYOF: if (*ep->reginput == '\0' || strchr(OPERAND(scan), *ep->reginput) == NULL) return(0); ep->reginput++; break; case ANYBUT: if (*ep->reginput == '\0' || strchr(OPERAND(scan), *ep->reginput) != NULL) return(0); ep->reginput++; break; case NOTHING: break; case BACK: break; 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 const int no = OP(scan) - OPEN; register char *const input = ep->reginput; if (regmatch(ep, next)) { /* * Don't set startp if some later * invocation of the same parentheses * already has. */ if (ep->regstartp[no] == NULL) ep->regstartp[no] = input; return(1); } else return(0); break; } 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 const int no = OP(scan) - CLOSE; register char *const input = ep->reginput; if (regmatch(ep, next)) { /* * Don't set endp if some later * invocation of the same parentheses * already has. */ if (ep->regendp[no] == NULL) ep->regendp[no] = input; return(1); } else return(0); break; } case BRANCH: { register char *const save = ep->reginput; if (OP(next) != BRANCH) /* No choice. */ next = OPERAND(scan); /* Avoid recursion. */ else { while (OP(scan) == BRANCH) { if (regmatch(ep, OPERAND(scan))) return(1); ep->reginput = save; scan = regnext(scan); } return(0); /* NOTREACHED */ } break; } case STAR: case PLUS: { register const char nextch = (OP(next) == EXACTLY) ? *OPERAND(next) : '\0'; register size_t no; register char *const save = ep->reginput; register const size_t min = (OP(scan) == STAR) ? 0 : 1; for (no = regrepeat(ep, OPERAND(scan)) + 1; no > min; no--) { ep->reginput = save + no - 1; /* If it could work, try it. */ if (nextch == '\0' || *ep->reginput == nextch) if (regmatch(ep, next)) return(1); } return(0); break; } case END: return(1); /* Success! */ break; default: regerror("regexp corruption"); return(0); break; } } /* * We get here only if there's trouble -- normally "case END" is * the terminating point. */ regerror("corrupted pointers"); return(0); } /* - regrepeat - report how many times something simple would match */ static size_t regrepeat(ep, node) register struct exec *ep; char *node; { register size_t count; register char *scan; register char ch; switch (OP(node)) { case ANY: return(strlen(ep->reginput)); break; case EXACTLY: ch = *OPERAND(node); count = 0; for (scan = ep->reginput; *scan == ch; scan++) count++; return(count); break; case ANYOF: return(strspn(ep->reginput, OPERAND(node))); break; case ANYBUT: return(strcspn(ep->reginput, OPERAND(node))); break; default: /* Oh dear. Called inappropriately. */ regerror("internal error: bad call of regrepeat"); return(0); /* Best compromise. */ break; } /* NOTREACHED */ } /* - regnext - dig the "next" pointer out of a node */ static char * regnext(p) register char *p; { register const int offset = NEXT(p); if (offset == 0) return(NULL); return((OP(p) == BACK) ? p-offset : p+offset); } #ifdef DEBUG static char *regprop(); /* - regdump - dump a regexp onto stdout in vaguely comprehensible form */ void regdump(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+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: sprintf(buf+strlen(buf), "OPEN%d", OP(op)-OPEN); p = NULL; break; 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: sprintf(buf+strlen(buf), "CLOSE%d", OP(op)-CLOSE); p = NULL; break; case STAR: p = "STAR"; break; case PLUS: p = "PLUS"; break; default: regerror("corrupted opcode"); break; } if (p != NULL) (void) strcat(buf, p); return(buf); } #endif ================================================ FILE: regexp.h ================================================ /* * Definitions etc. for regexp(3) routines. * * Caveat: this is V8 regexp(3) [actually, a reimplementation thereof], * not the System V one. */ #define NSUBEXP 10 typedef struct regexp { char *startp[NSUBEXP]; char *endp[NSUBEXP]; char regstart; /* Internal use only. */ char reganch; /* Internal use only. */ char *regmust; /* Internal use only. */ int regmlen; /* Internal use only. */ char program[1]; /* Unwarranted chumminess with compiler. */ } regexp; extern regexp *regcomp(const char *re); extern int regexec(regexp *rp, const char *s); extern void regsub(const regexp *rp, const char *src, char *dst); extern void regerror(char *message); ================================================ FILE: regmagic.h ================================================ /* * The first byte of the regexp internal "program" is actually this magic * number; the start node begins in the second byte. */ #define MAGIC 0234 ================================================ FILE: regsub.c ================================================ /* * regsub */ #include #include #include #include #include #include "regmagic.h" /* - regsub - perform substitutions after a regexp match */ void regsub(rp, source, dest) const regexp *rp; const char *source; char *dest; { register regexp * const prog = (regexp *)rp; register char *src = (char *)source; register char *dst = dest; register char c; register int no; register size_t len; if (prog == NULL || source == NULL || dest == NULL) { regerror("NULL parameter to regsub"); return; } if ((unsigned char)*(prog->program) != MAGIC) { regerror("damaged regexp"); return; } while ((c = *src++) != '\0') { if (c == '&') no = 0; else if (c == '\\' && isdigit(*src)) no = *src++ - '0'; else no = -1; if (no < 0) { /* Ordinary character. */ if (c == '\\' && (*src == '\\' || *src == '&')) c = *src++; *dst++ = c; } else if (prog->startp[no] != NULL && prog->endp[no] != NULL && prog->endp[no] > prog->startp[no]) { len = prog->endp[no] - prog->startp[no]; (void) strncpy(dst, prog->startp[no], len); dst += len; if (*(dst-1) == '\0') { /* strncpy hit NUL. */ regerror("damaged match string"); return; } } } *dst++ = '\0'; } ================================================ FILE: tests ================================================ abc abc y & abc abc xbc n - - abc axc n - - abc abx n - - abc xabcy y & abc abc ababc y & abc ab*c abc y & abc ab*bc abc y & abc ab*bc abbc y & abbc ab*bc abbbbc y & abbbbc ab+bc abbc y & abbc ab+bc abc n - - ab+bc abq n - - ab+bc abbbbc y & abbbbc ab?bc abbc y & abbc ab?bc abc y & abc ab?bc abbbbc n - - ab?c abc y & abc ^abc$ abc y & abc ^abc$ abcc n - - ^abc abcc y & abc ^abc$ aabc n - - abc$ aabc y & abc ^ abc y & $ abc y & a.c abc y & abc a.c axc y & axc a.*c axyzc y & axyzc a.*c axyzd n - - a[bc]d abc n - - a[bc]d abd y & abd a[b-d]e abd n - - a[b-d]e ace y & ace a[b-d] aac y & ac a[-b] a- y & a- a[b-] a- y & a- [k] ab n - - a[b-a] - c - - a[]b - c - - a[ - c - - a] a] y & a] a[]]b a]b y & a]b a[^bc]d aed y & aed a[^bc]d abd n - - a[^-b]c adc y & adc a[^-b]c a-c n - - a[^]b]c a]c n - - a[^]b]c adc y & adc ab|cd abc y & ab ab|cd abcd y & ab ()ef def y &-\1 ef- ()* - c - - *a - c - - ^* - c - - $* - c - - (*)b - c - - $b b n - - a\ - c - - a\(b a(b y &-\1 a(b- a\(*b ab y & ab a\(*b a((b y & a((b a\\b a\b y & a\b abc) - c - - (abc - c - - ((a)) abc y &-\1-\2 a-a-a (a)b(c) abc y &-\1-\2 abc-a-c a+b+c aabbabc y & abc a** - c - - a*? - c - - (a*)* - c - - (a*)+ - c - - (a|)* - c - - (a*|b)* - c - - (a+|b)* ab y &-\1 ab-b (a+|b)+ ab y &-\1 ab-b (a+|b)? ab y &-\1 a-a [^ab]* cde y & cde (^)* - c - - (ab|)* - c - - )( - c - - abc y & abc n - - a* y & abcd abcd y &-\&-\\& abcd-&-\abcd a(bc)d abcd y \1-\\1-\\\1 bc-\1-\bc ([abc])*d abbbcd y &-\1 abbbcd-c ([abc])*bcd abcd y &-\1 abcd-a a|b|c|d|e e y & e (a|b|c|d|e)f ef y &-\1 ef-e ((a*|b))* - c - - abcd*efg abcdefg y & abcdefg ab* xabyabbbz y & ab ab* xayabbbz y & a (ab|cd)e abcde y &-\1 cde-cd [abhgefdc]ij hij y & hij ^(ab|cd)e abcde n x\1y xy (abc|)ef abcdef y &-\1 ef- (a|b)c*d abcd y &-\1 bcd-b (ab|ab*)bc abc y &-\1 abc-a a([bc]*)c* abc y &-\1 abc-bc a([bc]*)(c*d) abcd y &-\1-\2 abcd-bc-d a([bc]+)(c*d) abcd y &-\1-\2 abcd-bc-d a([bc]*)(c+d) abcd y &-\1-\2 abcd-b-cd a[bcd]*dcdcde adcdcde y & adcdcde a[bcd]+dcdcde adcdcde n - - (ab|a)b*c abc y &-\1 abc-ab ((a)(b)c)(d) abcd y \1-\2-\3-\4 abc-a-b-d [ -~]* abc y & abc [ -~ -~]* abc y & abc [ -~ -~ -~]* abc y & abc [ -~ -~ -~ -~]* abc y & abc [ -~ -~ -~ -~ -~]* abc y & abc [ -~ -~ -~ -~ -~ -~]* abc y & abc [ -~ -~ -~ -~ -~ -~ -~]* abc y & abc [a-zA-Z_][a-zA-Z0-9_]* alpha y & alpha ^a(bc+|b[eh])g|.h$ abh y &-\1 bh- (bc+d$|ef*g.|h?i(j|k)) effgz y &-\1-\2 effgz-effgz- (bc+d$|ef*g.|h?i(j|k)) ij y &-\1-\2 ij-ij-j (bc+d$|ef*g.|h?i(j|k)) effg n - - (bc+d$|ef*g.|h?i(j|k)) bcdd n - - (bc+d$|ef*g.|h?i(j|k)) reffgz y &-\1-\2 effgz-effgz- ((((((((((a)))))))))) - c - - (((((((((a))))))))) a y & a multiple words of text uh-uh n - - multiple words multiple words, yeah y & multiple words (.*)c(.*) abcde y &-\1-\2 abcde-ab-de \((.*), (.*)\) (a, b) y (\2, \1) (b, a) ================================================ FILE: timer.c ================================================ /* * Simple timing program for regcomp(). * Usage: timer ncomp nexec nsub * or * timer ncomp nexec nsub regexp string [ answer [ sub ] ] * * The second form is for timing repetitions of a single test case. * The first form's test data is a compiled-in copy of the "tests" file. * Ncomp, nexec, nsub are how many times to do each regcomp, regexec, * and regsub. The way to time an operation individually is to do something * like "timer 1 50 1". */ #include struct try { char *re, *str, *ans, *src, *dst; } tests[] = { #include "timer.t.h" { NULL, NULL, NULL, NULL, NULL } }; #include int errreport = 0; /* Report errors via errseen? */ char *errseen = NULL; /* Error message. */ char *progname; /* ARGSUSED */ main(argc, argv) int argc; char *argv[]; { int ncomp, nexec, nsub; struct try one; char dummy[512]; if (argc < 4) { ncomp = 1; nexec = 1; nsub = 1; } else { ncomp = atoi(argv[1]); nexec = atoi(argv[2]); nsub = atoi(argv[3]); } progname = argv[0]; if (argc > 5) { one.re = argv[4]; one.str = argv[5]; if (argc > 6) one.ans = argv[6]; else one.ans = "y"; if (argc > 7) { one.src = argv[7]; one.dst = "xxx"; } else { one.src = "x"; one.dst = "x"; } errreport = 1; try(one, ncomp, nexec, nsub); } else multiple(ncomp, nexec, nsub); exit(0); } void regerror(s) char *s; { if (errreport) errseen = s; else error(s, ""); } #ifndef ERRAVAIL error(s1, s2) char *s1; char *s2; { fprintf(stderr, "regexp: "); fprintf(stderr, s1, s2); fprintf(stderr, "\n"); exit(1); } #endif int lineno = 0; multiple(ncomp, nexec, nsub) int ncomp, nexec, nsub; { register int i; extern char *strchr(); errreport = 1; for (i = 0; tests[i].re != NULL; i++) { lineno++; try(tests[i], ncomp, nexec, nsub); } } try(fields, ncomp, nexec, nsub) struct try fields; int ncomp, nexec, nsub; { regexp *r; char dbuf[BUFSIZ]; register int i; errseen = NULL; r = regcomp(fields.re); if (r == NULL) { if (*fields.ans != 'c') complain("regcomp failure in `%s'", fields.re); return; } if (*fields.ans == 'c') { complain("unexpected regcomp success in `%s'", fields.re); free((char *)r); return; } for (i = ncomp-1; i > 0; i--) { free((char *)r); r = regcomp(fields.re); } if (!regexec(r, fields.str)) { if (*fields.ans != 'n') complain("regexec failure in `%s'", ""); free((char *)r); return; } if (*fields.ans == 'n') { complain("unexpected regexec success", ""); free((char *)r); return; } for (i = nexec-1; i > 0; i--) (void) regexec(r, fields.str); errseen = NULL; for (i = nsub; i > 0; i--) regsub(r, fields.src, dbuf); if (errseen != NULL) { complain("regsub complaint", ""); free((char *)r); return; } if (strcmp(dbuf, fields.dst) != 0) complain("regsub result `%s' wrong", dbuf); free((char *)r); } complain(s1, s2) char *s1; char *s2; { fprintf(stderr, "try: %d: ", lineno); fprintf(stderr, s1, s2); fprintf(stderr, " (%s)\n", (errseen != NULL) ? errseen : ""); } ================================================ FILE: try.c ================================================ /* * Simple test program for regexp(3) stuff. Knows about debugging hooks. * Usage: try re [string [output [-]]] * The re is compiled and dumped, regexeced against the string, the result * is applied to output using regsub(). The - triggers a running narrative * from regexec(). Dumping and narrative don't happen unless DEBUG. * * If there are no arguments, stdin is assumed to be a stream of lines with * five fields: a r.e., a string to match it against, a result code, a * source string for regsub, and the proper result. Result codes are 'c' * for compile failure, 'y' for match success, 'n' for match failure. * Field separator is tab. */ #include #include #ifdef ERRAVAIL char *progname; extern char *mkprogname(); #endif #ifdef DEBUG extern int regnarrate; #endif char buf[BUFSIZ]; int errreport = 0; /* Report errors via errseen? */ char *errseen = NULL; /* Error message. */ int status = 0; /* Exit status. */ /* ARGSUSED */ main(argc, argv) int argc; char *argv[]; { regexp *r; int i; #ifdef ERRAVAIL progname = mkprogname(argv[0]); #endif if (argc == 1) { multiple(); exit(status); } r = regcomp(argv[1]); if (r == NULL) error("regcomp failure", ""); #ifdef DEBUG regdump(r); if (argc > 4) regnarrate++; #endif if (argc > 2) { i = regexec(r, argv[2]); printf("%d", i); for (i = 1; i < NSUBEXP; i++) if (r->startp[i] != NULL && r->endp[i] != NULL) printf(" \\%d", i); printf("\n"); } if (argc > 3) { regsub(r, argv[3], buf); printf("%s\n", buf); } exit(status); } void regerror(s) char *s; { if (errreport) errseen = s; else error(s, ""); } #ifndef ERRAVAIL error(s1, s2) char *s1; char *s2; { fprintf(stderr, "regexp: "); fprintf(stderr, s1, s2); fprintf(stderr, "\n"); exit(1); } #endif int lineno; regexp badregexp; /* Implicit init to 0. */ multiple() { char rbuf[BUFSIZ]; char *field[5]; char *scan; int i; regexp *r; extern char *strchr(); errreport = 1; lineno = 0; while (fgets(rbuf, sizeof(rbuf), stdin) != NULL) { rbuf[strlen(rbuf)-1] = '\0'; /* Dispense with \n. */ lineno++; scan = rbuf; for (i = 0; i < 5; i++) { field[i] = scan; if (field[i] == NULL) { complain("bad testfile format", ""); exit(1); } scan = strchr(scan, '\t'); if (scan != NULL) *scan++ = '\0'; } try(field); } /* And finish up with some internal testing... */ lineno = 9990; errseen = NULL; if (regcomp((char *)NULL) != NULL || errseen == NULL) complain("regcomp(NULL) doesn't complain", ""); lineno = 9991; errseen = NULL; if (regexec((regexp *)NULL, "foo") || errseen == NULL) complain("regexec(NULL, ...) doesn't complain", ""); lineno = 9992; r = regcomp("foo"); if (r == NULL) { complain("regcomp(\"foo\") fails", ""); return; } lineno = 9993; errseen = NULL; if (regexec(r, (char *)NULL) || errseen == NULL) complain("regexec(..., NULL) doesn't complain", ""); lineno = 9994; errseen = NULL; regsub((regexp *)NULL, "foo", rbuf); if (errseen == NULL) complain("regsub(NULL, ..., ...) doesn't complain", ""); lineno = 9995; errseen = NULL; regsub(r, (char *)NULL, rbuf); if (errseen == NULL) complain("regsub(..., NULL, ...) doesn't complain", ""); lineno = 9996; errseen = NULL; regsub(r, "foo", (char *)NULL); if (errseen == NULL) complain("regsub(..., ..., NULL) doesn't complain", ""); lineno = 9997; errseen = NULL; if (regexec(&badregexp, "foo") || errseen == NULL) complain("regexec(nonsense, ...) doesn't complain", ""); lineno = 9998; errseen = NULL; regsub(&badregexp, "foo", rbuf); if (errseen == NULL) complain("regsub(nonsense, ..., ...) doesn't complain", ""); } try(fields) char **fields; { regexp *r; char dbuf[BUFSIZ]; errseen = NULL; r = regcomp(fields[0]); if (r == NULL) { if (*fields[2] != 'c') complain("regcomp failure in `%s'", fields[0]); return; } if (*fields[2] == 'c') { complain("unexpected regcomp success in `%s'", fields[0]); free((char *)r); return; } if (!regexec(r, fields[1])) { if (*fields[2] != 'n') complain("regexec failure in `%s'", fields[0]); free((char *)r); return; } if (*fields[2] == 'n') { complain("unexpected regexec success", ""); free((char *)r); return; } errseen = NULL; regsub(r, fields[3], dbuf); if (errseen != NULL) { complain("regsub complaint", ""); free((char *)r); return; } if (strcmp(dbuf, fields[4]) != 0) complain("regsub result `%s' wrong", dbuf); free((char *)r); } complain(s1, s2) char *s1; char *s2; { fprintf(stderr, "try: %d: ", lineno); fprintf(stderr, s1, s2); fprintf(stderr, " (%s)\n", (errseen != NULL) ? errseen : ""); status = 1; }