diff options
Diffstat (limited to 'dfa.c')
| -rw-r--r-- | dfa.c | 2832 |
1 files changed, 1563 insertions, 1269 deletions
@@ -1,213 +1,173 @@ -/* dfa.c - determinisitic extended regexp routines for GNU +/* dfa.c - deterministic extended regexp routines for GNU Copyright (C) 1988 Free Software Foundation, Inc. - Written June, 1988 by Mike Haertel - Modified July, 1988 by Arthur David Olson - to assist BMG speedups - - NO WARRANTY - - BECAUSE THIS PROGRAM IS LICENSED FREE OF CHARGE, WE PROVIDE ABSOLUTELY -NO WARRANTY, TO THE EXTENT PERMITTED BY APPLICABLE STATE LAW. EXCEPT -WHEN OTHERWISE STATED IN WRITING, FREE SOFTWARE FOUNDATION, INC, -RICHARD M. STALLMAN AND/OR OTHER PARTIES PROVIDE THIS PROGRAM "AS IS" -WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, -BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND -FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY -AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE -DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR -CORRECTION. - - IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW WILL RICHARD M. -STALLMAN, THE FREE SOFTWARE FOUNDATION, INC., AND/OR ANY OTHER PARTY -WHO MAY MODIFY AND REDISTRIBUTE THIS PROGRAM AS PERMITTED BELOW, BE -LIABLE TO YOU FOR DAMAGES, INCLUDING ANY LOST PROFITS, LOST MONIES, OR -OTHER SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE -USE OR INABILITY TO USE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR -DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY THIRD PARTIES OR -A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS) THIS -PROGRAM, EVEN IF YOU HAVE BEEN ADVISED OF THE POSSIBILITY OF SUCH -DAMAGES, OR FOR ANY CLAIM BY ANY OTHER PARTY. - - GENERAL PUBLIC LICENSE TO COPY - - 1. You may copy and distribute verbatim copies of this source file -as you receive it, in any medium, provided that you conspicuously and -appropriately publish on each copy a valid copyright notice "Copyright - (C) 1988 Free Software Foundation, Inc."; and include following the -copyright notice a verbatim copy of the above disclaimer of warranty -and of this License. You may charge a distribution fee for the -physical act of transferring a copy. - - 2. You may modify your copy or copies of this source file or -any portion of it, and copy and distribute such modifications under -the terms of Paragraph 1 above, provided that you also do the following: - - a) cause the modified files to carry prominent notices stating - that you changed the files and the date of any change; and - - b) cause the whole of any work that you distribute or publish, - that in whole or in part contains or is a derivative of this - program or any part thereof, to be licensed at no charge to all - third parties on terms identical to those contained in this - License Agreement (except that you may choose to grant more extensive - warranty protection to some or all third parties, at your option). - - c) You may charge a distribution fee for the physical act of - transferring a copy, and you may at your option offer warranty - protection in exchange for a fee. - -Mere aggregation of another unrelated program with this program (or its -derivative) on a volume of a storage or distribution medium does not bring -the other program under the scope of these terms. - - 3. You may copy and distribute this program or any portion of it in -compiled, executable or object code form under the terms of Paragraphs -1 and 2 above provided that you do the following: - - a) accompany it with the complete corresponding machine-readable - source code, which must be distributed under the terms of - Paragraphs 1 and 2 above; or, - - b) accompany it with a written offer, valid for at least three - years, to give any third party free (except for a nominal - shipping charge) a complete machine-readable copy of the - corresponding source code, to be distributed under the terms of - Paragraphs 1 and 2 above; or, - - c) accompany it with the information you received as to where the - corresponding source code may be obtained. (This alternative is - allowed only for noncommercial distribution and only if you - received the program in object code or executable form alone.) - -For an executable file, complete source code means all the source code for -all modules it contains; but, as a special exception, it need not include -source code for modules which are standard libraries that accompany the -operating system on which the executable file runs. - - 4. You may not copy, sublicense, distribute or transfer this program -except as expressly provided under this License Agreement. Any attempt -otherwise to copy, sublicense, distribute or transfer this program is void and -your rights to use the program under this License agreement shall be -automatically terminated. However, parties who have received computer -software programs from you with this License Agreement will not have -their licenses terminated so long as such parties remain in full compliance. - - 5. If you wish to incorporate parts of this program into other free -programs whose distribution conditions are different, write to the Free -Software Foundation at 675 Mass Ave, Cambridge, MA 02139. We have not yet -worked out a simple rule that can be stated here, but we will often permit -this. We will be guided by the two goals of preserving the free status of -all derivatives our free software and of promoting the sharing and reuse of -software. - - -In other words, you are welcome to use, share and improve this program. -You are forbidden to forbid anyone else to use, share and improve -what you give them. Help stamp out software-hoarding! */ - -#include "awk.h" + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2, or (at your option) + any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program; if not, write to the Free Software + Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ + +/* Written June, 1988 by Mike Haertel + Modified July, 1988 by Arthur David Olson to assist BMG speedups */ + #include <assert.h> +#include <ctype.h> +#include <stdio.h> + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#ifdef STDC_HEADERS +#include <stdlib.h> +#else +#include <sys/types.h> +extern char *calloc(), *malloc(), *realloc(); +extern void free(); +#endif -#ifdef setbit /* surprise - setbit and clrbit are macros on NeXT */ -#undef setbit +#if defined(HAVE_STRING_H) || defined(STDC_HEADERS) +#include <string.h> +#undef index +#define index strchr +#else +#include <strings.h> #endif -#ifdef clrbit -#undef clrbit + +#ifndef DEBUG /* use the same approach as regex.c */ +#undef assert +#define assert(e) +#endif /* DEBUG */ + +#ifndef isgraph +#define isgraph(C) (isprint(C) && !isspace(C)) #endif +#ifdef isascii +#define ISALPHA(C) (isascii(C) && isalpha(C)) +#define ISUPPER(C) (isascii(C) && isupper(C)) +#define ISLOWER(C) (isascii(C) && islower(C)) +#define ISDIGIT(C) (isascii(C) && isdigit(C)) +#define ISXDIGIT(C) (isascii(C) && isxdigit(C)) +#define ISSPACE(C) (isascii(C) && isspace(C)) +#define ISPUNCT(C) (isascii(C) && ispunct(C)) +#define ISALNUM(C) (isascii(C) && isalnum(C)) +#define ISPRINT(C) (isascii(C) && isprint(C)) +#define ISGRAPH(C) (isascii(C) && isgraph(C)) +#define ISCNTRL(C) (isascii(C) && iscntrl(C)) +#else +#define ISALPHA(C) isalpha(C) +#define ISUPPER(C) isupper(C) +#define ISLOWER(C) islower(C) +#define ISDIGIT(C) isdigit(C) +#define ISXDIGIT(C) isxdigit(C) +#define ISSPACE(C) isspace(C) +#define ISPUNCT(C) ispunct(C) +#define ISALNUM(C) isalnum(C) +#define ISPRINT(C) isprint(C) +#define ISGRAPH(C) isgraph(C) +#define ISCNTRL(C) iscntrl(C) +#endif + +#include "regex.h" +#include "dfa.h" + #ifdef __STDC__ typedef void *ptr_t; #else typedef char *ptr_t; #endif -typedef struct { - char ** in; - char * left; - char * right; - char * is; -} must; +static void dfamust _RE_ARGS((struct dfa *dfa)); -static ptr_t xcalloc P((MALLOC_ARG_T n, MALLOC_ARG_T s)); -static ptr_t xmalloc P((MALLOC_ARG_T n)); -static ptr_t xrealloc P((ptr_t p, MALLOC_ARG_T n)); -static int tstbit P((int b, _charset c)); -static void setbit P((int b, _charset c)); -static void clrbit P((int b, _charset c)); -static void copyset P((const _charset src, _charset dst)); -static void zeroset P((_charset s)); -static void notset P((_charset s)); -static int equal P((const _charset s1, const _charset s2)); -static int charset_index P((const _charset s)); -static _token lex P((void)); -static void addtok P((_token t)); -static void atom P((void)); -static void closure P((void)); -static void branch P((void)); -static void regexp P((void)); -static void copy P((const _position_set *src, _position_set *dst)); -static void insert P((_position p, _position_set *s)); -static void merge P((_position_set *s1, _position_set *s2, _position_set *m)); -static void delete P((_position p, _position_set *s)); -static int state_index P((struct regexp *r, _position_set *s, +static ptr_t xcalloc _RE_ARGS((size_t n, size_t s)); +static ptr_t xmalloc _RE_ARGS((size_t n)); +static ptr_t xrealloc _RE_ARGS((ptr_t p, size_t n)); +#ifdef DEBUG +static void prtok _RE_ARGS((token t)); +#endif +static int tstbit _RE_ARGS((int b, charclass c)); +static void setbit _RE_ARGS((int b, charclass c)); +static void clrbit _RE_ARGS((int b, charclass c)); +static void copyset _RE_ARGS((charclass src, charclass dst)); +static void zeroset _RE_ARGS((charclass s)); +static void notset _RE_ARGS((charclass s)); +static int equal _RE_ARGS((charclass s1, charclass s2)); +static int charclass_index _RE_ARGS((charclass s)); +static int looking_at _RE_ARGS((const char *s)); +static token lex _RE_ARGS((void)); +static void addtok _RE_ARGS((token t)); +static void atom _RE_ARGS((void)); +static int nsubtoks _RE_ARGS((int tindex)); +static void copytoks _RE_ARGS((int tindex, int ntokens)); +static void closure _RE_ARGS((void)); +static void branch _RE_ARGS((void)); +static void regexp _RE_ARGS((int toplevel)); +static void copy _RE_ARGS((position_set *src, position_set *dst)); +static void insert _RE_ARGS((position p, position_set *s)); +static void merge _RE_ARGS((position_set *s1, position_set *s2, position_set *m)); +static void delete _RE_ARGS((position p, position_set *s)); +static int state_index _RE_ARGS((struct dfa *d, position_set *s, int newline, int letter)); -static void epsclosure P((_position_set *s, struct regexp *r)); -static void build_state P((int s, struct regexp *r)); -static void build_state_zero P((struct regexp *r)); -static char *icatalloc P((char *old, const char *new)); -static char *icpyalloc P((const char *string)); -static char *istrstr P((char *lookin, char *lookfor)); -static void ifree P((char *cp)); -static void freelist P((char **cpp)); -static char **enlist P((char **cpp, char *new, size_t len)); -static char **comsubs P((char *left, char *right)); -static char **addlists P((char **old, char **new)); -static char **inboth P((char **left, char **right)); -static void resetmust P((must *mp)); -static void regmust P((struct regexp *r)); - -#undef P +static void build_state _RE_ARGS((int s, struct dfa *d)); +static void build_state_zero _RE_ARGS((struct dfa *d)); +static char *icatalloc _RE_ARGS((char *old, char *new)); +static char *icpyalloc _RE_ARGS((char *string)); +static char *istrstr _RE_ARGS((char *lookin, char *lookfor)); +static void ifree _RE_ARGS((char *cp)); +static void freelist _RE_ARGS((char **cpp)); +static char **enlist _RE_ARGS((char **cpp, char *new, size_t len)); +static char **comsubs _RE_ARGS((char *left, char *right)); +static char **addlists _RE_ARGS((char **old, char **new)); +static char **inboth _RE_ARGS((char **left, char **right)); static ptr_t xcalloc(n, s) - MALLOC_ARG_T n; - MALLOC_ARG_T s; + size_t n; + size_t s; { ptr_t r = calloc(n, s); - if (NULL == r) - reg_error("Memory exhausted"); /* reg_error does not return */ + if (!r) + dfaerror("Memory exhausted"); return r; } static ptr_t xmalloc(n) - MALLOC_ARG_T n; + size_t n; { ptr_t r = malloc(n); assert(n != 0); - if (NULL == r) - reg_error("Memory exhausted"); + if (!r) + dfaerror("Memory exhausted"); return r; } static ptr_t xrealloc(p, n) ptr_t p; - MALLOC_ARG_T n; + size_t n; { ptr_t r = realloc(p, n); assert(n != 0); - if (NULL == r) - reg_error("Memory exhausted"); + if (!r) + dfaerror("Memory exhausted"); return r; } -#define CALLOC(p, t, n) ((p) = (t *) xcalloc((MALLOC_ARG_T)(n), sizeof (t))) -#undef MALLOC -#define MALLOC(p, t, n) ((p) = (t *) xmalloc((MALLOC_ARG_T)(n) * sizeof (t))) +#define CALLOC(p, t, n) ((p) = (t *) xcalloc((size_t)(n), sizeof (t))) +#define MALLOC(p, t, n) ((p) = (t *) xmalloc((n) * sizeof (t))) #define REALLOC(p, t, n) ((p) = (t *) xrealloc((ptr_t) (p), (n) * sizeof (t))) /* Reallocate an array of type t if nalloc is too small for index. */ @@ -218,13 +178,52 @@ xrealloc(p, n) (nalloc) *= 2; \ REALLOC(p, t, nalloc); \ } - -/* Stuff pertaining to charsets. */ + +#ifdef DEBUG + +static void +prtok(t) + token t; +{ + char *s; + + if (t < 0) + fprintf(stderr, "END"); + else if (t < NOTCHAR) + fprintf(stderr, "%c", t); + else + { + switch (t) + { + case EMPTY: s = "EMPTY"; break; + case BACKREF: s = "BACKREF"; break; + case BEGLINE: s = "BEGLINE"; break; + case ENDLINE: s = "ENDLINE"; break; + case BEGWORD: s = "BEGWORD"; break; + case ENDWORD: s = "ENDWORD"; break; + case LIMWORD: s = "LIMWORD"; break; + case NOTLIMWORD: s = "NOTLIMWORD"; break; + case QMARK: s = "QMARK"; break; + case STAR: s = "STAR"; break; + case PLUS: s = "PLUS"; break; + case CAT: s = "CAT"; break; + case OR: s = "OR"; break; + case ORTOP: s = "ORTOP"; break; + case LPAREN: s = "LPAREN"; break; + case RPAREN: s = "RPAREN"; break; + default: s = "CSET"; break; + } + fprintf(stderr, "%s", s); + } +} +#endif /* DEBUG */ + +/* Stuff pertaining to charclasses. */ static int tstbit(b, c) int b; - _charset c; + charclass c; { return c[b / INTBITS] & 1 << b % INTBITS; } @@ -232,7 +231,7 @@ tstbit(b, c) static void setbit(b, c) int b; - _charset c; + charclass c; { c[b / INTBITS] |= 1 << b % INTBITS; } @@ -240,84 +239,84 @@ setbit(b, c) static void clrbit(b, c) int b; - _charset c; + charclass c; { c[b / INTBITS] &= ~(1 << b % INTBITS); } static void copyset(src, dst) - const _charset src; - _charset dst; + charclass src; + charclass dst; { int i; - for (i = 0; i < _CHARSET_INTS; ++i) + for (i = 0; i < CHARCLASS_INTS; ++i) dst[i] = src[i]; } static void zeroset(s) - _charset s; + charclass s; { int i; - for (i = 0; i < _CHARSET_INTS; ++i) + for (i = 0; i < CHARCLASS_INTS; ++i) s[i] = 0; } static void notset(s) - _charset s; + charclass s; { int i; - for (i = 0; i < _CHARSET_INTS; ++i) + for (i = 0; i < CHARCLASS_INTS; ++i) s[i] = ~s[i]; } static int equal(s1, s2) - const _charset s1; - const _charset s2; + charclass s1; + charclass s2; { int i; - for (i = 0; i < _CHARSET_INTS; ++i) + for (i = 0; i < CHARCLASS_INTS; ++i) if (s1[i] != s2[i]) return 0; return 1; } - -/* A pointer to the current regexp is kept here during parsing. */ -static struct regexp *reg; -/* Find the index of charset s in reg->charsets, or allocate a new charset. */ +/* A pointer to the current dfa is kept here during parsing. */ +static struct dfa *dfa; + +/* Find the index of charclass s in dfa->charclasses, or allocate a new charclass. */ static int -charset_index(s) - const _charset s; +charclass_index(s) + charclass s; { int i; - for (i = 0; i < reg->cindex; ++i) - if (equal(s, reg->charsets[i])) + for (i = 0; i < dfa->cindex; ++i) + if (equal(s, dfa->charclasses[i])) return i; - REALLOC_IF_NECESSARY(reg->charsets, _charset, reg->calloc, reg->cindex); - ++reg->cindex; - copyset(s, reg->charsets[i]); + REALLOC_IF_NECESSARY(dfa->charclasses, charclass, dfa->calloc, dfa->cindex); + ++dfa->cindex; + copyset(s, dfa->charclasses[i]); return i; } /* Syntax bits controlling the behavior of the lexical analyzer. */ -static syntax_bits, syntax_bits_set; +static reg_syntax_t syntax_bits, syntax_bits_set; /* Flag for case-folding letters into sets. */ -static case_fold; +static int case_fold; /* Entry point to set syntax options. */ void -regsyntax(bits, fold) - long bits; +dfasyntax(bits, fold) + reg_syntax_t bits; int fold; { syntax_bits_set = 1; @@ -325,63 +324,136 @@ regsyntax(bits, fold) case_fold = fold; } -/* Lexical analyzer. */ -static const char *lexstart; /* Pointer to beginning of input string. */ -static const char *lexptr; /* Pointer to next input character. */ +/* Lexical analyzer. All the dross that deals with the obnoxious + GNU Regex syntax bits is located here. The poor, suffering + reader is referred to the GNU Regex documentation for the + meaning of the @#%!@#%^!@ syntax bits. */ + +static char *lexstart; /* Pointer to beginning of input string. */ +static char *lexptr; /* Pointer to next input character. */ static lexleft; /* Number of characters remaining. */ -static caret_allowed; /* True if backward context allows ^ - (meaningful only if RE_CONTEXT_INDEP_OPS - is turned off). */ -static closure_allowed; /* True if backward context allows closures - (meaningful only if RE_CONTEXT_INDEP_OPS - is turned off). */ +static token lasttok; /* Previous token returned; initially END. */ +static int laststart; /* True if we're separated from beginning or (, | + only by zero-width characters. */ +static int parens; /* Count of outstanding left parens. */ +static int minrep, maxrep; /* Repeat counts for {m,n}. */ /* Note that characters become unsigned here. */ #define FETCH(c, eoferr) \ { \ if (! lexleft) \ - if (eoferr != NULL) \ - reg_error(eoferr); \ + if (eoferr != 0) \ + dfaerror(eoferr); \ else \ - return _END; \ + return lasttok = END; \ (c) = (unsigned char) *lexptr++; \ --lexleft; \ } -static _token +#ifdef __STDC__ +#define FUNC(F, P) static int F(int c) { return P(c); } +#else +#define FUNC(F, P) static int F(c) int c; { return P(c); } +#endif + +FUNC(is_alpha, ISALPHA) +FUNC(is_upper, ISUPPER) +FUNC(is_lower, ISLOWER) +FUNC(is_digit, ISDIGIT) +FUNC(is_xdigit, ISXDIGIT) +FUNC(is_space, ISSPACE) +FUNC(is_punct, ISPUNCT) +FUNC(is_alnum, ISALNUM) +FUNC(is_print, ISPRINT) +FUNC(is_graph, ISGRAPH) +FUNC(is_cntrl, ISCNTRL) + +/* The following list maps the names of the Posix named character classes + to predicate functions that determine whether a given character is in + the class. The leading [ has already been eaten by the lexical analyzer. */ +static struct { + const char *name; + int (*pred) _RE_ARGS((int)); +} prednames[] = { + { ":alpha:]", is_alpha }, + { ":upper:]", is_upper }, + { ":lower:]", is_lower }, + { ":digit:]", is_digit }, + { ":xdigit:]", is_xdigit }, + { ":space:]", is_space }, + { ":punct:]", is_punct }, + { ":alnum:]", is_alnum }, + { ":print:]", is_print }, + { ":graph:]", is_graph }, + { ":cntrl:]", is_cntrl }, + { 0 } +}; + +static int +looking_at(s) + const char *s; +{ + size_t len; + + len = strlen(s); + if (lexleft < len) + return 0; + return strncmp(s, lexptr, len) == 0; +} + +static token lex() { - _token c, c2; - int invert; - _charset cset; + token c, c1, c2; + int backslash = 0, invert; + charclass ccl; + int i; - FETCH(c, (char *) 0); - switch (c) + /* Basic plan: We fetch a character. If it's a backslash, + we set the backslash flag and go through the loop again. + On the plus side, this avoids having a duplicate of the + main switch inside the backslash case. On the minus side, + it means that just about every case begins with + "if (backslash) ...". */ + for (i = 0; i < 2; ++i) { - case '^': - if (! (syntax_bits & RE_CONTEXT_INDEP_OPS) - && (!caret_allowed || - ((syntax_bits & RE_TIGHT_VBAR) && lexptr - 1 != lexstart))) - goto normal_char; - caret_allowed = 0; - return syntax_bits & RE_TIGHT_VBAR ? _ALLBEGLINE : _BEGLINE; - - case '$': - if (syntax_bits & RE_CONTEXT_INDEP_OPS || !lexleft - || (! (syntax_bits & RE_TIGHT_VBAR) - && ((syntax_bits & RE_NO_BK_PARENS - ? lexleft > 0 && *lexptr == ')' - : lexleft > 1 && *lexptr == '\\' && lexptr[1] == ')') - || (syntax_bits & RE_NO_BK_VBAR - ? lexleft > 0 && *lexptr == '|' - : lexleft > 1 && *lexptr == '\\' && lexptr[1] == '|')))) - return syntax_bits & RE_TIGHT_VBAR ? _ALLENDLINE : _ENDLINE; - goto normal_char; - - case '\\': - FETCH(c, "Unfinished \\ quote"); + FETCH(c, 0); switch (c) { + case '\\': + if (backslash) + goto normal_char; + if (lexleft == 0) + dfaerror("Unfinished \\ escape"); + backslash = 1; + break; + + case '^': + if (backslash) + goto normal_char; + if (syntax_bits & RE_CONTEXT_INDEP_ANCHORS + || lasttok == END + || lasttok == LPAREN + || lasttok == OR) + return lasttok = BEGLINE; + goto normal_char; + + case '$': + if (backslash) + goto normal_char; + if (syntax_bits & RE_CONTEXT_INDEP_ANCHORS + || lexleft == 0 + || (syntax_bits & RE_NO_BK_PARENS + ? lexleft > 0 && *lexptr == ')' + : lexleft > 1 && lexptr[0] == '\\' && lexptr[1] == ')') + || (syntax_bits & RE_NO_BK_VBAR + ? lexleft > 0 && *lexptr == '|' + : lexleft > 1 && lexptr[0] == '\\' && lexptr[1] == '|') + || ((syntax_bits & RE_NEWLINE_ALT) + && lexleft > 0 && *lexptr == '\n')) + return lasttok = ENDLINE; + goto normal_char; + case '1': case '2': case '3': @@ -391,236 +463,315 @@ lex() case '7': case '8': case '9': - caret_allowed = 0; - closure_allowed = 1; - return _BACKREF; + if (backslash && !(syntax_bits & RE_NO_BK_REFS)) + { + laststart = 0; + return lasttok = BACKREF; + } + goto normal_char; case '<': - caret_allowed = 0; - return _BEGWORD; + if (syntax_bits & RE_NO_GNU_OPS) + goto normal_char; + if (backslash) + return lasttok = BEGWORD; + goto normal_char; case '>': - caret_allowed = 0; - return _ENDWORD; + if (syntax_bits & RE_NO_GNU_OPS) + goto normal_char; + if (backslash) + return lasttok = ENDWORD; + goto normal_char; case 'b': - caret_allowed = 0; - return _LIMWORD; + if (syntax_bits & RE_NO_GNU_OPS) + goto normal_char; + if (backslash) + return lasttok = LIMWORD; + goto normal_char; case 'B': - caret_allowed = 0; - return _NOTLIMWORD; - - case 'w': - case 'W': - zeroset(cset); - for (c2 = 0; c2 < _NOTCHAR; ++c2) - if (ISALNUM(c2)) - setbit(c2, cset); - if (c == 'W') - notset(cset); - caret_allowed = 0; - closure_allowed = 1; - return _SET + charset_index(cset); + if (syntax_bits & RE_NO_GNU_OPS) + goto normal_char; + if (backslash) + return lasttok = NOTLIMWORD; + goto normal_char; case '?': - if (syntax_bits & RE_BK_PLUS_QM) - goto qmark; - goto normal_char; + if (syntax_bits & RE_LIMITED_OPS) + goto normal_char; + if (backslash != ((syntax_bits & RE_BK_PLUS_QM) != 0)) + goto normal_char; + if (!(syntax_bits & RE_CONTEXT_INDEP_OPS) && laststart) + goto normal_char; + return lasttok = QMARK; + + case '*': + if (backslash) + goto normal_char; + if (!(syntax_bits & RE_CONTEXT_INDEP_OPS) && laststart) + goto normal_char; + return lasttok = STAR; case '+': - if (syntax_bits & RE_BK_PLUS_QM) - goto plus; - goto normal_char; + if (syntax_bits & RE_LIMITED_OPS) + goto normal_char; + if (backslash != ((syntax_bits & RE_BK_PLUS_QM) != 0)) + goto normal_char; + if (!(syntax_bits & RE_CONTEXT_INDEP_OPS) && laststart) + goto normal_char; + return lasttok = PLUS; + + case '{': + if (!(syntax_bits & RE_INTERVALS)) + goto normal_char; + if (backslash != ((syntax_bits & RE_NO_BK_BRACES) == 0)) + goto normal_char; + minrep = maxrep = 0; + /* Cases: + {M} - exact count + {M,} - minimum count, maximum is infinity + {,M} - 0 through M + {M,N} - M through N */ + FETCH(c, "unfinished repeat count"); + if (ISDIGIT(c)) + { + minrep = c - '0'; + for (;;) + { + FETCH(c, "unfinished repeat count"); + if (!ISDIGIT(c)) + break; + minrep = 10 * minrep + c - '0'; + } + } + else if (c != ',') + dfaerror("malformed repeat count"); + if (c == ',') + for (;;) + { + FETCH(c, "unfinished repeat count"); + if (!ISDIGIT(c)) + break; + maxrep = 10 * maxrep + c - '0'; + } + else + maxrep = minrep; + if (!(syntax_bits & RE_NO_BK_BRACES)) + { + if (c != '\\') + dfaerror("malformed repeat count"); + FETCH(c, "unfinished repeat count"); + } + if (c != '}') + dfaerror("malformed repeat count"); + laststart = 0; + return lasttok = REPMN; case '|': - if (! (syntax_bits & RE_NO_BK_VBAR)) - goto or; - goto normal_char; + if (syntax_bits & RE_LIMITED_OPS) + goto normal_char; + if (backslash != ((syntax_bits & RE_NO_BK_VBAR) == 0)) + goto normal_char; + laststart = 1; + return lasttok = OR; + + case '\n': + if (syntax_bits & RE_LIMITED_OPS + || backslash + || !(syntax_bits & RE_NEWLINE_ALT)) + goto normal_char; + laststart = 1; + return lasttok = OR; case '(': - if (! (syntax_bits & RE_NO_BK_PARENS)) - goto lparen; - goto normal_char; + if (backslash != ((syntax_bits & RE_NO_BK_PARENS) == 0)) + goto normal_char; + ++parens; + laststart = 1; + return lasttok = LPAREN; case ')': - if (! (syntax_bits & RE_NO_BK_PARENS)) - goto rparen; - goto normal_char; - - default: - goto normal_char; - } + if (backslash != ((syntax_bits & RE_NO_BK_PARENS) == 0)) + goto normal_char; + if (parens == 0 && syntax_bits & RE_UNMATCHED_RIGHT_PAREN_ORD) + goto normal_char; + --parens; + laststart = 0; + return lasttok = RPAREN; + + case '.': + if (backslash) + goto normal_char; + zeroset(ccl); + notset(ccl); + if (!(syntax_bits & RE_DOT_NEWLINE)) + clrbit('\n', ccl); + if (syntax_bits & RE_DOT_NOT_NULL) + clrbit('\0', ccl); + laststart = 0; + return lasttok = CSET + charclass_index(ccl); - case '?': - if (syntax_bits & RE_BK_PLUS_QM) - goto normal_char; - qmark: - if (! (syntax_bits & RE_CONTEXT_INDEP_OPS) && !closure_allowed) - goto normal_char; - return _QMARK; - - case '*': - if (! (syntax_bits & RE_CONTEXT_INDEP_OPS) && !closure_allowed) - goto normal_char; - return _STAR; - - case '+': - if (syntax_bits & RE_BK_PLUS_QM) - goto normal_char; - plus: - if (! (syntax_bits & RE_CONTEXT_INDEP_OPS) && !closure_allowed) - goto normal_char; - return _PLUS; - - case '|': - if (! (syntax_bits & RE_NO_BK_VBAR)) - goto normal_char; - or: - caret_allowed = 1; - closure_allowed = 0; - return _OR; - - case '\n': - if (! (syntax_bits & RE_NEWLINE_OR)) - goto normal_char; - goto or; - - case '(': - if (! (syntax_bits & RE_NO_BK_PARENS)) - goto normal_char; - lparen: - caret_allowed = 1; - closure_allowed = 0; - return _LPAREN; - - case ')': - if (! (syntax_bits & RE_NO_BK_PARENS)) - goto normal_char; - rparen: - caret_allowed = 0; - closure_allowed = 1; - return _RPAREN; - - case '.': - zeroset(cset); - notset(cset); - clrbit('\n', cset); - caret_allowed = 0; - closure_allowed = 1; - return _SET + charset_index(cset); - - case '[': - zeroset(cset); - FETCH(c, "Unbalanced ["); - if (c == '^') - { + case 'w': + case 'W': + if (!backslash || (syntax_bits & RE_NO_GNU_OPS)) + goto normal_char; + zeroset(ccl); + for (c2 = 0; c2 < NOTCHAR; ++c2) + if (ISALNUM(c2)) + setbit(c2, ccl); + if (c == 'W') + notset(ccl); + laststart = 0; + return lasttok = CSET + charclass_index(ccl); + + case '[': + if (backslash) + goto normal_char; + zeroset(ccl); FETCH(c, "Unbalanced ["); - invert = 1; - } - else - invert = 0; - do - { - FETCH(c2, "Unbalanced ["); - if ((syntax_bits & RE_AWK_CLASS_HACK) && c == '\\') + if (c == '^') { - c = c2; - FETCH(c2, "Unbalanced ["); + FETCH(c, "Unbalanced ["); + invert = 1; } - if (c2 == '-') + else + invert = 0; + do { - FETCH(c2, "Unbalanced ["); - if (c2 == ']' && (syntax_bits & RE_AWK_CLASS_HACK)) + /* Nobody ever said this had to be fast. :-) + Note that if we're looking at some other [:...:] + construct, we just treat it as a bunch of ordinary + characters. We can do this because we assume + regex has checked for syntax errors before + dfa is ever called. */ + if (c == '[' && (syntax_bits & RE_CHAR_CLASSES)) + for (c1 = 0; prednames[c1].name; ++c1) + if (looking_at(prednames[c1].name)) + { + for (c2 = 0; c2 < NOTCHAR; ++c2) + if ((*prednames[c1].pred)(c2)) + setbit(c2, ccl); + lexptr += strlen(prednames[c1].name); + lexleft -= strlen(prednames[c1].name); + FETCH(c1, "Unbalanced ["); + goto skip; + } + if (c == '\\' && (syntax_bits & RE_BACKSLASH_ESCAPE_IN_LISTS)) + FETCH(c, "Unbalanced ["); + FETCH(c1, "Unbalanced ["); + if (c1 == '-') { - setbit(c, cset); - setbit('-', cset); - break; - } + FETCH(c2, "Unbalanced ["); + if (c2 == ']') + { + /* In the case [x-], the - is an ordinary hyphen, + which is left in c1, the lookahead character. */ + --lexptr; + ++lexleft; + c2 = c; + } + else + { + if (c2 == '\\' + && (syntax_bits & RE_BACKSLASH_ESCAPE_IN_LISTS)) + FETCH(c2, "Unbalanced ["); + FETCH(c1, "Unbalanced ["); + } + } + else + c2 = c; while (c <= c2) - setbit(c++, cset); - FETCH(c, "Unbalanced ["); + { + setbit(c, ccl); + if (case_fold) + if (ISUPPER(c)) + setbit(tolower(c), ccl); + else if (ISLOWER(c)) + setbit(toupper(c), ccl); + ++c; + } + skip: + ; } - else + while ((c = c1) != ']'); + if (invert) { - setbit(c, cset); - c = c2; + notset(ccl); + if (syntax_bits & RE_HAT_LISTS_NOT_NEWLINE) + clrbit('\n', ccl); } - } - while (c != ']'); - if (invert) - notset(cset); - caret_allowed = 0; - closure_allowed = 1; - return _SET + charset_index(cset); + laststart = 0; + return lasttok = CSET + charclass_index(ccl); - default: - normal_char: - caret_allowed = 0; - closure_allowed = 1; - if (case_fold && ISALPHA(c)) - { - zeroset(cset); - if (isupper(c)) - c = tolower(c); - setbit(c, cset); - setbit(toupper(c), cset); - return _SET + charset_index(cset); + default: + normal_char: + laststart = 0; + if (case_fold && ISALPHA(c)) + { + zeroset(ccl); + setbit(c, ccl); + if (isupper(c)) + setbit(tolower(c), ccl); + else + setbit(toupper(c), ccl); + return lasttok = CSET + charclass_index(ccl); + } + return c; } - return c; } + + /* The above loop should consume at most a backslash + and some other character. */ + abort(); } - + /* Recursive descent parser for regular expressions. */ -static _token tok; /* Lookahead token. */ +static token tok; /* Lookahead token. */ static depth; /* Current depth of a hypothetical stack holding deferred productions. This is used to determine the depth that will be required of the real stack later on in - reganalyze(). */ + dfaanalyze(). */ /* Add the given token to the parse tree, maintaining the depth count and updating the maximum depth if necessary. */ static void addtok(t) - _token t; + token t; { - REALLOC_IF_NECESSARY(reg->tokens, _token, reg->talloc, reg->tindex); - reg->tokens[reg->tindex++] = t; + REALLOC_IF_NECESSARY(dfa->tokens, token, dfa->talloc, dfa->tindex); + dfa->tokens[dfa->tindex++] = t; switch (t) { - case _QMARK: - case _STAR: - case _PLUS: + case QMARK: + case STAR: + case PLUS: break; - case _CAT: - case _OR: + case CAT: + case OR: + case ORTOP: --depth; break; default: - ++reg->nleaves; - case _EMPTY: + ++dfa->nleaves; + case EMPTY: ++depth; break; } - if (depth > reg->depth) - reg->depth = depth; + if (depth > dfa->depth) + dfa->depth = depth; } /* The grammar understood by the parser is as follows. - start: - regexp - _ALLBEGLINE regexp - regexp _ALLENDLINE - _ALLBEGLINE regexp _ALLENDLINE - regexp: - regexp _OR branch + regexp OR branch branch branch: @@ -628,144 +779,187 @@ addtok(t) closure closure: - closure _QMARK - closure _STAR - closure _PLUS + closure QMARK + closure STAR + closure PLUS atom atom: <normal character> - _SET - _BACKREF - _BEGLINE - _ENDLINE - _BEGWORD - _ENDWORD - _LIMWORD - _NOTLIMWORD + CSET + BACKREF + BEGLINE + ENDLINE + BEGWORD + ENDWORD + LIMWORD + NOTLIMWORD <empty> The parser builds a parse tree in postfix form in an array of tokens. */ -#ifdef __STDC__ -static void regexp(void); -#else -static void regexp(); -#endif - static void atom() { - if (tok >= 0 && (tok < _NOTCHAR || tok >= _SET || tok == _BACKREF - || tok == _BEGLINE || tok == _ENDLINE || tok == _BEGWORD - || tok == _ENDWORD || tok == _LIMWORD || tok == _NOTLIMWORD)) + if ((tok >= 0 && tok < NOTCHAR) || tok >= CSET || tok == BACKREF + || tok == BEGLINE || tok == ENDLINE || tok == BEGWORD + || tok == ENDWORD || tok == LIMWORD || tok == NOTLIMWORD) { addtok(tok); tok = lex(); } - else if (tok == _LPAREN) + else if (tok == LPAREN) { tok = lex(); - regexp(); - if (tok != _RPAREN) - reg_error("Unbalanced ("); + regexp(0); + if (tok != RPAREN) + dfaerror("Unbalanced ("); tok = lex(); } else - addtok(_EMPTY); + addtok(EMPTY); +} + +/* Return the number of tokens in the given subexpression. */ +static int +nsubtoks(tindex) +int tindex; +{ + int ntoks1; + + switch (dfa->tokens[tindex - 1]) + { + default: + return 1; + case QMARK: + case STAR: + case PLUS: + return 1 + nsubtoks(tindex - 1); + case CAT: + case OR: + case ORTOP: + ntoks1 = nsubtoks(tindex - 1); + return 1 + ntoks1 + nsubtoks(tindex - 1 - ntoks1); + } +} + +/* Copy the given subexpression to the top of the tree. */ +static void +copytoks(tindex, ntokens) + int tindex, ntokens; +{ + int i; + + for (i = 0; i < ntokens; ++i) + addtok(dfa->tokens[tindex + i]); } static void closure() { + int tindex, ntokens, i; + atom(); - while (tok == _QMARK || tok == _STAR || tok == _PLUS) - { - addtok(tok); - tok = lex(); - } + while (tok == QMARK || tok == STAR || tok == PLUS || tok == REPMN) + if (tok == REPMN) + { + ntokens = nsubtoks(dfa->tindex); + tindex = dfa->tindex - ntokens; + if (maxrep == 0) + addtok(PLUS); + if (minrep == 0) + addtok(QMARK); + for (i = 1; i < minrep; ++i) + { + copytoks(tindex, ntokens); + addtok(CAT); + } + for (; i < maxrep; ++i) + { + copytoks(tindex, ntokens); + addtok(QMARK); + addtok(CAT); + } + tok = lex(); + } + else + { + addtok(tok); + tok = lex(); + } } static void branch() { closure(); - while (tok != _RPAREN && tok != _OR && tok != _ALLENDLINE && tok >= 0) + while (tok != RPAREN && tok != OR && tok >= 0) { closure(); - addtok(_CAT); + addtok(CAT); } } static void -regexp() +regexp(toplevel) + int toplevel; { branch(); - while (tok == _OR) + while (tok == OR) { tok = lex(); branch(); - addtok(_OR); + if (toplevel) + addtok(ORTOP); + else + addtok(OR); } } /* Main entry point for the parser. S is a string to be parsed, len is the - length of the string, so s can include NUL characters. R is a pointer to - the struct regexp to parse into. */ + length of the string, so s can include NUL characters. D is a pointer to + the struct dfa to parse into. */ void -regparse(s, len, r) - const char *s; +dfaparse(s, len, d) + char *s; size_t len; - struct regexp *r; + struct dfa *d; + { - reg = r; + dfa = d; lexstart = lexptr = s; lexleft = len; - caret_allowed = 1; - closure_allowed = 0; + lasttok = END; + laststart = 1; + parens = 0; if (! syntax_bits_set) - reg_error("No syntax specified"); + dfaerror("No syntax specified"); tok = lex(); - depth = r->depth; + depth = d->depth; - if (tok == _ALLBEGLINE) - { - addtok(_BEGLINE); - tok = lex(); - regexp(); - addtok(_CAT); - } - else - regexp(); + regexp(1); - if (tok == _ALLENDLINE) - { - addtok(_ENDLINE); - addtok(_CAT); - tok = lex(); - } - - if (tok != _END) - reg_error("Unbalanced )"); + if (tok != END) + dfaerror("Unbalanced )"); - addtok(_END - r->nregexps); - addtok(_CAT); + addtok(END - d->nregexps); + addtok(CAT); - if (r->nregexps) - addtok(_OR); + if (d->nregexps) + addtok(ORTOP); - ++r->nregexps; + ++d->nregexps; } - + /* Some primitives for operating on sets of positions. */ /* Copy one set to another; the destination must be large enough. */ static void copy(src, dst) - const _position_set *src; - _position_set *dst; + position_set *src; + position_set *dst; { int i; @@ -780,11 +974,11 @@ copy(src, dst) S->elems must point to an array large enough to hold the resulting set. */ static void insert(p, s) - _position p; - _position_set *s; + position p; + position_set *s; { int i; - _position t1, t2; + position t1, t2; for (i = 0; i < s->nelem && p.index < s->elems[i].index; ++i) ; @@ -807,9 +1001,9 @@ insert(p, s) the positions of both sets were inserted into an initially empty set. */ static void merge(s1, s2, m) - _position_set *s1; - _position_set *s2; - _position_set *m; + position_set *s1; + position_set *s2; + position_set *m; { int i = 0, j = 0; @@ -833,8 +1027,8 @@ merge(s1, s2, m) /* Delete a position from a set. */ static void delete(p, s) - _position p; - _position_set *s; + position p; + position_set *s; { int i; @@ -845,19 +1039,19 @@ delete(p, s) for (--s->nelem; i < s->nelem; ++i) s->elems[i] = s->elems[i + 1]; } - + /* Find the index of the state corresponding to the given position set with the given preceding context, or create a new state if there is no such state. Newline and letter tell whether we got here on a newline or letter, respectively. */ static int -state_index(r, s, newline, letter) - struct regexp *r; - _position_set *s; +state_index(d, s, newline, letter) + struct dfa *d; + position_set *s; int newline; int letter; { - int lhash = 0; + int hash = 0; int constraint; int i, j; @@ -865,78 +1059,80 @@ state_index(r, s, newline, letter) letter = letter ? 1 : 0; for (i = 0; i < s->nelem; ++i) - lhash ^= s->elems[i].index + s->elems[i].constraint; + hash ^= s->elems[i].index + s->elems[i].constraint; /* Try to find a state that exactly matches the proposed one. */ - for (i = 0; i < r->sindex; ++i) + for (i = 0; i < d->sindex; ++i) { - if (lhash != r->states[i].hash || s->nelem != r->states[i].elems.nelem - || newline != r->states[i].newline || letter != r->states[i].letter) + if (hash != d->states[i].hash || s->nelem != d->states[i].elems.nelem + || newline != d->states[i].newline || letter != d->states[i].letter) continue; for (j = 0; j < s->nelem; ++j) if (s->elems[j].constraint - != r->states[i].elems.elems[j].constraint - || s->elems[j].index != r->states[i].elems.elems[j].index) + != d->states[i].elems.elems[j].constraint + || s->elems[j].index != d->states[i].elems.elems[j].index) break; if (j == s->nelem) return i; } /* We'll have to create a new state. */ - REALLOC_IF_NECESSARY(r->states, _dfa_state, r->salloc, r->sindex); - r->states[i].hash = lhash; - MALLOC(r->states[i].elems.elems, _position, s->nelem); - copy(s, &r->states[i].elems); - r->states[i].newline = newline; - r->states[i].letter = letter; - r->states[i].backref = 0; - r->states[i].constraint = 0; - r->states[i].first_end = 0; + REALLOC_IF_NECESSARY(d->states, dfa_state, d->salloc, d->sindex); + d->states[i].hash = hash; + MALLOC(d->states[i].elems.elems, position, s->nelem); + copy(s, &d->states[i].elems); + d->states[i].newline = newline; + d->states[i].letter = letter; + d->states[i].backref = 0; + d->states[i].constraint = 0; + d->states[i].first_end = 0; for (j = 0; j < s->nelem; ++j) - if (r->tokens[s->elems[j].index] < 0) + if (d->tokens[s->elems[j].index] < 0) { constraint = s->elems[j].constraint; - if (_SUCCEEDS_IN_CONTEXT(constraint, newline, 0, letter, 0) - || _SUCCEEDS_IN_CONTEXT(constraint, newline, 0, letter, 1) - || _SUCCEEDS_IN_CONTEXT(constraint, newline, 1, letter, 0) - || _SUCCEEDS_IN_CONTEXT(constraint, newline, 1, letter, 1)) - r->states[i].constraint |= constraint; - if (! r->states[i].first_end) - r->states[i].first_end = r->tokens[s->elems[j].index]; + if (SUCCEEDS_IN_CONTEXT(constraint, newline, 0, letter, 0) + || SUCCEEDS_IN_CONTEXT(constraint, newline, 0, letter, 1) + || SUCCEEDS_IN_CONTEXT(constraint, newline, 1, letter, 0) + || SUCCEEDS_IN_CONTEXT(constraint, newline, 1, letter, 1)) + d->states[i].constraint |= constraint; + if (! d->states[i].first_end) + d->states[i].first_end = d->tokens[s->elems[j].index]; } - else if (r->tokens[s->elems[j].index] == _BACKREF) + else if (d->tokens[s->elems[j].index] == BACKREF) { - r->states[i].constraint = _NO_CONSTRAINT; - r->states[i].backref = 1; + d->states[i].constraint = NO_CONSTRAINT; + d->states[i].backref = 1; } - ++r->sindex; + ++d->sindex; return i; } - + /* Find the epsilon closure of a set of positions. If any position of the set contains a symbol that matches the empty string in some context, replace that position with the elements of its follow labeled with an appropriate constraint. Repeat exhaustively until no funny positions are left. S->elems must be large enough to hold the result. */ +static void epsclosure _RE_ARGS((position_set *s, struct dfa *d)); + static void -epsclosure(s, r) - _position_set *s; - struct regexp *r; +epsclosure(s, d) + position_set *s; + struct dfa *d; { int i, j; int *visited; - _position p, old; + position p, old; - MALLOC(visited, int, r->tindex); - for (i = 0; i < r->tindex; ++i) + MALLOC(visited, int, d->tindex); + for (i = 0; i < d->tindex; ++i) visited[i] = 0; for (i = 0; i < s->nelem; ++i) - if (r->tokens[s->elems[i].index] >= _NOTCHAR - && r->tokens[s->elems[i].index] != _BACKREF - && r->tokens[s->elems[i].index] < _SET) + if (d->tokens[s->elems[i].index] >= NOTCHAR + && d->tokens[s->elems[i].index] != BACKREF + && d->tokens[s->elems[i].index] < CSET) { old = s->elems[i]; p.constraint = old.constraint; @@ -947,32 +1143,32 @@ epsclosure(s, r) continue; } visited[old.index] = 1; - switch (r->tokens[old.index]) + switch (d->tokens[old.index]) { - case _BEGLINE: - p.constraint &= _BEGLINE_CONSTRAINT; + case BEGLINE: + p.constraint &= BEGLINE_CONSTRAINT; break; - case _ENDLINE: - p.constraint &= _ENDLINE_CONSTRAINT; + case ENDLINE: + p.constraint &= ENDLINE_CONSTRAINT; break; - case _BEGWORD: - p.constraint &= _BEGWORD_CONSTRAINT; + case BEGWORD: + p.constraint &= BEGWORD_CONSTRAINT; break; - case _ENDWORD: - p.constraint &= _ENDWORD_CONSTRAINT; + case ENDWORD: + p.constraint &= ENDWORD_CONSTRAINT; break; - case _LIMWORD: - p.constraint &= _ENDWORD_CONSTRAINT; + case LIMWORD: + p.constraint &= LIMWORD_CONSTRAINT; break; - case _NOTLIMWORD: - p.constraint &= _NOTLIMWORD_CONSTRAINT; + case NOTLIMWORD: + p.constraint &= NOTLIMWORD_CONSTRAINT; break; default: break; } - for (j = 0; j < r->follows[old.index].nelem; ++j) + for (j = 0; j < d->follows[old.index].nelem; ++j) { - p.index = r->follows[old.index].elems[j].index; + p.index = d->follows[old.index].elems[j].index; insert(p, s); } /* Force rescan to start at the beginning. */ @@ -981,41 +1177,41 @@ epsclosure(s, r) free(visited); } - + /* Perform bottom-up analysis on the parse tree, computing various functions. Note that at this point, we're pretending constructs like \< are real characters rather than constraints on what can follow them. Nullable: A node is nullable if it is at the root of a regexp that can match the empty string. - * _EMPTY leaves are nullable. + * EMPTY leaves are nullable. * No other leaf is nullable. - * A _QMARK or _STAR node is nullable. - * A _PLUS node is nullable if its argument is nullable. - * A _CAT node is nullable if both its arguments are nullable. - * An _OR node is nullable if either argument is nullable. + * A QMARK or STAR node is nullable. + * A PLUS node is nullable if its argument is nullable. + * A CAT node is nullable if both its arguments are nullable. + * An OR node is nullable if either argument is nullable. Firstpos: The firstpos of a node is the set of positions (nonempty leaves) that could correspond to the first character of a string matching the regexp rooted at the given node. - * _EMPTY leaves have empty firstpos. + * EMPTY leaves have empty firstpos. * The firstpos of a nonempty leaf is that leaf itself. - * The firstpos of a _QMARK, _STAR, or _PLUS node is the firstpos of its + * The firstpos of a QMARK, STAR, or PLUS node is the firstpos of its argument. - * The firstpos of a _CAT node is the firstpos of the left argument, union + * The firstpos of a CAT node is the firstpos of the left argument, union the firstpos of the right if the left argument is nullable. - * The firstpos of an _OR node is the union of firstpos of each argument. + * The firstpos of an OR node is the union of firstpos of each argument. Lastpos: The lastpos of a node is the set of positions that could correspond to the last character of a string matching the regexp at the given node. - * _EMPTY leaves have empty lastpos. + * EMPTY leaves have empty lastpos. * The lastpos of a nonempty leaf is that leaf itself. - * The lastpos of a _QMARK, _STAR, or _PLUS node is the lastpos of its + * The lastpos of a QMARK, STAR, or PLUS node is the lastpos of its argument. - * The lastpos of a _CAT node is the lastpos of its right argument, union + * The lastpos of a CAT node is the lastpos of its right argument, union the lastpos of the left if the right argument is nullable. - * The lastpos of an _OR node is the union of the lastpos of each argument. + * The lastpos of an OR node is the union of the lastpos of each argument. Follow: The follow of a position is the set of positions that could correspond to the character following a character matching the node in @@ -1024,9 +1220,9 @@ epsclosure(s, r) Later, if we find that a special symbol is in a follow set, we will replace it with the elements of its follow, labeled with an appropriate constraint. - * Every node in the firstpos of the argument of a _STAR or _PLUS node is in + * Every node in the firstpos of the argument of a STAR or PLUS node is in the follow of every node in the lastpos. - * Every node in the firstpos of the second argument of a _CAT node is in + * Every node in the firstpos of the second argument of a CAT node is in the follow of every node in the lastpos of the first argument. Because of the postfix representation of the parse tree, the depth-first @@ -1035,48 +1231,61 @@ epsclosure(s, r) scheme; the number of elements in each set deeper in the stack can be used to determine the address of a particular set's array. */ void -reganalyze(r, searchflag) - struct regexp *r; +dfaanalyze(d, searchflag) + struct dfa *d; int searchflag; { int *nullable; /* Nullable stack. */ int *nfirstpos; /* Element count stack for firstpos sets. */ - _position *firstpos; /* Array where firstpos elements are stored. */ + position *firstpos; /* Array where firstpos elements are stored. */ int *nlastpos; /* Element count stack for lastpos sets. */ - _position *lastpos; /* Array where lastpos elements are stored. */ + position *lastpos; /* Array where lastpos elements are stored. */ int *nalloc; /* Sizes of arrays allocated to follow sets. */ - _position_set tmp; /* Temporary set for merging sets. */ - _position_set merged; /* Result of merging sets. */ + position_set tmp; /* Temporary set for merging sets. */ + position_set merged; /* Result of merging sets. */ int wants_newline; /* True if some position wants newline info. */ int *o_nullable; int *o_nfirst, *o_nlast; - _position *o_firstpos, *o_lastpos; + position *o_firstpos, *o_lastpos; int i, j; - _position *pos; + position *pos; - r->searchflag = searchflag; +#ifdef DEBUG + fprintf(stderr, "dfaanalyze:\n"); + for (i = 0; i < d->tindex; ++i) + { + fprintf(stderr, " %d:", i); + prtok(d->tokens[i]); + } + putc('\n', stderr); +#endif + + d->searchflag = searchflag; - MALLOC(nullable, int, r->depth); + MALLOC(nullable, int, d->depth); o_nullable = nullable; - MALLOC(nfirstpos, int, r->depth); + MALLOC(nfirstpos, int, d->depth); o_nfirst = nfirstpos; - MALLOC(firstpos, _position, r->nleaves); - o_firstpos = firstpos, firstpos += r->nleaves; - MALLOC(nlastpos, int, r->depth); + MALLOC(firstpos, position, d->nleaves); + o_firstpos = firstpos, firstpos += d->nleaves; + MALLOC(nlastpos, int, d->depth); o_nlast = nlastpos; - MALLOC(lastpos, _position, r->nleaves); - o_lastpos = lastpos, lastpos += r->nleaves; - MALLOC(nalloc, int, r->tindex); - for (i = 0; i < r->tindex; ++i) + MALLOC(lastpos, position, d->nleaves); + o_lastpos = lastpos, lastpos += d->nleaves; + MALLOC(nalloc, int, d->tindex); + for (i = 0; i < d->tindex; ++i) nalloc[i] = 0; - MALLOC(merged.elems, _position, r->nleaves); + MALLOC(merged.elems, position, d->nleaves); - CALLOC(r->follows, _position_set, r->tindex); + CALLOC(d->follows, position_set, d->tindex); - for (i = 0; i < r->tindex; ++i) - switch (r->tokens[i]) + for (i = 0; i < d->tindex; ++i) +#ifdef DEBUG + { /* Nonsyntactic #ifdef goo... */ +#endif + switch (d->tokens[i]) { - case _EMPTY: + case EMPTY: /* The empty set is nullable. */ *nullable++ = 1; @@ -1084,8 +1293,8 @@ reganalyze(r, searchflag) *nfirstpos++ = *nlastpos++ = 0; break; - case _STAR: - case _PLUS: + case STAR: + case PLUS: /* Every element in the firstpos of the argument is in the follow of every element in the lastpos. */ tmp.nelem = nfirstpos[-1]; @@ -1093,19 +1302,19 @@ reganalyze(r, searchflag) pos = lastpos; for (j = 0; j < nlastpos[-1]; ++j) { - merge(&tmp, &r->follows[pos[j].index], &merged); - REALLOC_IF_NECESSARY(r->follows[pos[j].index].elems, _position, + merge(&tmp, &d->follows[pos[j].index], &merged); + REALLOC_IF_NECESSARY(d->follows[pos[j].index].elems, position, nalloc[pos[j].index], merged.nelem - 1); - copy(&merged, &r->follows[pos[j].index]); + copy(&merged, &d->follows[pos[j].index]); } - case _QMARK: - /* A _QMARK or _STAR node is automatically nullable. */ - if (r->tokens[i] != _PLUS) + case QMARK: + /* A QMARK or STAR node is automatically nullable. */ + if (d->tokens[i] != PLUS) nullable[-1] = 1; break; - case _CAT: + case CAT: /* Every element in the firstpos of the second argument is in the follow of every element in the lastpos of the first argument. */ tmp.nelem = nfirstpos[-1]; @@ -1113,13 +1322,13 @@ reganalyze(r, searchflag) pos = lastpos + nlastpos[-1]; for (j = 0; j < nlastpos[-2]; ++j) { - merge(&tmp, &r->follows[pos[j].index], &merged); - REALLOC_IF_NECESSARY(r->follows[pos[j].index].elems, _position, + merge(&tmp, &d->follows[pos[j].index], &merged); + REALLOC_IF_NECESSARY(d->follows[pos[j].index].elems, position, nalloc[pos[j].index], merged.nelem - 1); - copy(&merged, &r->follows[pos[j].index]); + copy(&merged, &d->follows[pos[j].index]); } - /* The firstpos of a _CAT node is the firstpos of the first argument, + /* The firstpos of a CAT node is the firstpos of the first argument, union that of the second argument if the first is nullable. */ if (nullable[-2]) nfirstpos[-2] += nfirstpos[-1]; @@ -1127,7 +1336,7 @@ reganalyze(r, searchflag) firstpos += nfirstpos[-1]; --nfirstpos; - /* The lastpos of a _CAT node is the lastpos of the second argument, + /* The lastpos of a CAT node is the lastpos of the second argument, union that of the first argument if the second is nullable. */ if (nullable[-1]) nlastpos[-2] += nlastpos[-1]; @@ -1141,12 +1350,13 @@ reganalyze(r, searchflag) } --nlastpos; - /* A _CAT node is nullable if both arguments are nullable. */ + /* A CAT node is nullable if both arguments are nullable. */ nullable[-2] = nullable[-1] && nullable[-2]; --nullable; break; - case _OR: + case OR: + case ORTOP: /* The firstpos is the union of the firstpos of each argument. */ nfirstpos[-2] += nfirstpos[-1]; --nfirstpos; @@ -1155,7 +1365,7 @@ reganalyze(r, searchflag) nlastpos[-2] += nlastpos[-1]; --nlastpos; - /* An _OR node is nullable if either argument is nullable. */ + /* An OR node is nullable if either argument is nullable. */ nullable[-2] = nullable[-1] || nullable[-2]; --nullable; break; @@ -1166,31 +1376,63 @@ reganalyze(r, searchflag) an "epsilon closure" effectively makes them nullable later. Backreferences have to get a real position so we can detect transitions on them later. But they are nullable. */ - *nullable++ = r->tokens[i] == _BACKREF; + *nullable++ = d->tokens[i] == BACKREF; /* This position is in its own firstpos and lastpos. */ *nfirstpos++ = *nlastpos++ = 1; --firstpos, --lastpos; firstpos->index = lastpos->index = i; - firstpos->constraint = lastpos->constraint = _NO_CONSTRAINT; + firstpos->constraint = lastpos->constraint = NO_CONSTRAINT; /* Allocate the follow set for this position. */ nalloc[i] = 1; - MALLOC(r->follows[i].elems, _position, nalloc[i]); + MALLOC(d->follows[i].elems, position, nalloc[i]); break; } +#ifdef DEBUG + /* ... balance the above nonsyntactic #ifdef goo... */ + fprintf(stderr, "node %d:", i); + prtok(d->tokens[i]); + putc('\n', stderr); + fprintf(stderr, nullable[-1] ? " nullable: yes\n" : " nullable: no\n"); + fprintf(stderr, " firstpos:"); + for (j = nfirstpos[-1] - 1; j >= 0; --j) + { + fprintf(stderr, " %d:", firstpos[j].index); + prtok(d->tokens[firstpos[j].index]); + } + fprintf(stderr, "\n lastpos:"); + for (j = nlastpos[-1] - 1; j >= 0; --j) + { + fprintf(stderr, " %d:", lastpos[j].index); + prtok(d->tokens[lastpos[j].index]); + } + putc('\n', stderr); + } +#endif /* For each follow set that is the follow set of a real position, replace it with its epsilon closure. */ - for (i = 0; i < r->tindex; ++i) - if (r->tokens[i] < _NOTCHAR || r->tokens[i] == _BACKREF - || r->tokens[i] >= _SET) + for (i = 0; i < d->tindex; ++i) + if (d->tokens[i] < NOTCHAR || d->tokens[i] == BACKREF + || d->tokens[i] >= CSET) { - copy(&r->follows[i], &merged); - epsclosure(&merged, r); - if (r->follows[i].nelem < merged.nelem) - REALLOC(r->follows[i].elems, _position, merged.nelem); - copy(&merged, &r->follows[i]); +#ifdef DEBUG + fprintf(stderr, "follows(%d:", i); + prtok(d->tokens[i]); + fprintf(stderr, "):"); + for (j = d->follows[i].nelem - 1; j >= 0; --j) + { + fprintf(stderr, " %d:", d->follows[i].elems[j].index); + prtok(d->tokens[d->follows[i].elems[j].index]); + } + putc('\n', stderr); +#endif + copy(&d->follows[i], &merged); + epsclosure(&merged, d); + if (d->follows[i].nelem < merged.nelem) + REALLOC(d->follows[i].elems, position, merged.nelem); + copy(&merged, &d->follows[i]); } /* Get the epsilon closure of the firstpos of the regexp. The result will @@ -1198,19 +1440,19 @@ reganalyze(r, searchflag) merged.nelem = 0; for (i = 0; i < nfirstpos[-1]; ++i) insert(firstpos[i], &merged); - epsclosure(&merged, r); + epsclosure(&merged, d); /* Check if any of the positions of state 0 will want newline context. */ wants_newline = 0; for (i = 0; i < merged.nelem; ++i) - if (_PREV_NEWLINE_DEPENDENT(merged.elems[i].constraint)) + if (PREV_NEWLINE_DEPENDENT(merged.elems[i].constraint)) wants_newline = 1; /* Build the initial state. */ - r->salloc = 1; - r->sindex = 0; - MALLOC(r->states, _dfa_state, r->salloc); - state_index(r, &merged, wants_newline, 0); + d->salloc = 1; + d->sindex = 0; + MALLOC(d->states, dfa_state, d->salloc); + state_index(d, &merged, wants_newline, 0); free(o_nullable); free(o_nfirst); @@ -1220,8 +1462,8 @@ reganalyze(r, searchflag) free(nalloc); free(merged.elems); } - -/* Find, for each character, the transition out of state s of r, and store + +/* Find, for each character, the transition out of state s of d, and store it in the appropriate slot of trans. We divide the positions of s into groups (positions can appear in more @@ -1252,25 +1494,25 @@ reganalyze(r, searchflag) create a new group labeled with the characters of C and insert this position in that group. */ void -regstate(s, r, trans) +dfastate(s, d, trans) int s; - struct regexp *r; + struct dfa *d; int trans[]; { - _position_set grps[_NOTCHAR]; /* As many as will ever be needed. */ - _charset labels[_NOTCHAR]; /* Labels corresponding to the groups. */ + position_set grps[NOTCHAR]; /* As many as will ever be needed. */ + charclass labels[NOTCHAR]; /* Labels corresponding to the groups. */ int ngrps = 0; /* Number of groups actually used. */ - _position pos; /* Current position being considered. */ - _charset matches; /* Set of matching characters. */ + position pos; /* Current position being considered. */ + charclass matches; /* Set of matching characters. */ int matchesf; /* True if matches is nonempty. */ - _charset intersect; /* Intersection with some label set. */ + charclass intersect; /* Intersection with some label set. */ int intersectf; /* True if intersect is nonempty. */ - _charset leftovers; /* Stuff in the label that didn't match. */ + charclass leftovers; /* Stuff in the label that didn't match. */ int leftoversf; /* True if leftovers is nonempty. */ - static _charset letters; /* Set of characters considered letters. */ - static _charset newline; /* Set of characters that aren't newline. */ - _position_set follows; /* Union of the follows of some group. */ - _position_set tmp; /* Temporary space for merging sets. */ + static charclass letters; /* Set of characters considered letters. */ + static charclass newline; /* Set of characters that aren't newline. */ + position_set follows; /* Union of the follows of some group. */ + position_set tmp; /* Temporary space for merging sets. */ int state; /* New state. */ int wants_newline; /* New state wants to know newline context. */ int state_newline; /* New state on a newline transition. */ @@ -1283,7 +1525,7 @@ regstate(s, r, trans) if (! initialized) { initialized = 1; - for (i = 0; i < _NOTCHAR; ++i) + for (i = 0; i < NOTCHAR; ++i) if (ISALNUM(i)) setbit(i, letters); setbit('\n', newline); @@ -1291,40 +1533,40 @@ regstate(s, r, trans) zeroset(matches); - for (i = 0; i < r->states[s].elems.nelem; ++i) + for (i = 0; i < d->states[s].elems.nelem; ++i) { - pos = r->states[s].elems.elems[i]; - if (r->tokens[pos.index] >= 0 && r->tokens[pos.index] < _NOTCHAR) - setbit(r->tokens[pos.index], matches); - else if (r->tokens[pos.index] >= _SET) - copyset(r->charsets[r->tokens[pos.index] - _SET], matches); + pos = d->states[s].elems.elems[i]; + if (d->tokens[pos.index] >= 0 && d->tokens[pos.index] < NOTCHAR) + setbit(d->tokens[pos.index], matches); + else if (d->tokens[pos.index] >= CSET) + copyset(d->charclasses[d->tokens[pos.index] - CSET], matches); else continue; - /* Some characters may need to be climinated from matches because + /* Some characters may need to be eliminated from matches because they fail in the current context. */ - if (pos.constraint != 0xff) + if (pos.constraint != 0xFF) { - if (! _MATCHES_NEWLINE_CONTEXT(pos.constraint, - r->states[s].newline, 1)) + if (! MATCHES_NEWLINE_CONTEXT(pos.constraint, + d->states[s].newline, 1)) clrbit('\n', matches); - if (! _MATCHES_NEWLINE_CONTEXT(pos.constraint, - r->states[s].newline, 0)) - for (j = 0; j < _CHARSET_INTS; ++j) + if (! MATCHES_NEWLINE_CONTEXT(pos.constraint, + d->states[s].newline, 0)) + for (j = 0; j < CHARCLASS_INTS; ++j) matches[j] &= newline[j]; - if (! _MATCHES_LETTER_CONTEXT(pos.constraint, - r->states[s].letter, 1)) - for (j = 0; j < _CHARSET_INTS; ++j) + if (! MATCHES_LETTER_CONTEXT(pos.constraint, + d->states[s].letter, 1)) + for (j = 0; j < CHARCLASS_INTS; ++j) matches[j] &= ~letters[j]; - if (! _MATCHES_LETTER_CONTEXT(pos.constraint, - r->states[s].letter, 0)) - for (j = 0; j < _CHARSET_INTS; ++j) + if (! MATCHES_LETTER_CONTEXT(pos.constraint, + d->states[s].letter, 0)) + for (j = 0; j < CHARCLASS_INTS; ++j) matches[j] &= letters[j]; /* If there are no characters left, there's no point in going on. */ - for (j = 0; j < _CHARSET_INTS && !matches[j]; ++j) + for (j = 0; j < CHARCLASS_INTS && !matches[j]; ++j) ; - if (j == _CHARSET_INTS) + if (j == CHARCLASS_INTS) continue; } @@ -1333,21 +1575,21 @@ regstate(s, r, trans) /* If matches contains a single character only, and the current group's label doesn't contain that character, go on to the next group. */ - if (r->tokens[pos.index] >= 0 && r->tokens[pos.index] < _NOTCHAR - && !tstbit(r->tokens[pos.index], labels[j])) + if (d->tokens[pos.index] >= 0 && d->tokens[pos.index] < NOTCHAR + && !tstbit(d->tokens[pos.index], labels[j])) continue; /* Check if this group's label has a nonempty intersection with matches. */ intersectf = 0; - for (k = 0; k < _CHARSET_INTS; ++k) + for (k = 0; k < CHARCLASS_INTS; ++k) (intersect[k] = matches[k] & labels[j][k]) ? intersectf = 1 : 0; if (! intersectf) continue; /* It does; now find the set differences both ways. */ leftoversf = matchesf = 0; - for (k = 0; k < _CHARSET_INTS; ++k) + for (k = 0; k < CHARCLASS_INTS; ++k) { /* Even an optimizing compiler can't know this for sure. */ int match = matches[k], label = labels[j][k]; @@ -1361,7 +1603,7 @@ regstate(s, r, trans) { copyset(leftovers, labels[ngrps]); copyset(intersect, labels[j]); - MALLOC(grps[ngrps].elems, _position, r->nleaves); + MALLOC(grps[ngrps].elems, position, d->nleaves); copy(&grps[j], &grps[ngrps]); ++ngrps; } @@ -1382,46 +1624,50 @@ regstate(s, r, trans) { copyset(matches, labels[ngrps]); zeroset(matches); - MALLOC(grps[ngrps].elems, _position, r->nleaves); + MALLOC(grps[ngrps].elems, position, d->nleaves); grps[ngrps].nelem = 1; grps[ngrps].elems[0] = pos; ++ngrps; } } - MALLOC(follows.elems, _position, r->nleaves); - MALLOC(tmp.elems, _position, r->nleaves); + MALLOC(follows.elems, position, d->nleaves); + MALLOC(tmp.elems, position, d->nleaves); /* If we are a searching matcher, the default transition is to a state containing the positions of state 0, otherwise the default transition is to fail miserably. */ - if (r->searchflag) + if (d->searchflag) { wants_newline = 0; wants_letter = 0; - for (i = 0; i < r->states[0].elems.nelem; ++i) + for (i = 0; i < d->states[0].elems.nelem; ++i) { - if (_PREV_NEWLINE_DEPENDENT(r->states[0].elems.elems[i].constraint)) + if (PREV_NEWLINE_DEPENDENT(d->states[0].elems.elems[i].constraint)) wants_newline = 1; - if (_PREV_LETTER_DEPENDENT(r->states[0].elems.elems[i].constraint)) + if (PREV_LETTER_DEPENDENT(d->states[0].elems.elems[i].constraint)) wants_letter = 1; } - copy(&r->states[0].elems, &follows); - state = state_index(r, &follows, 0, 0); + copy(&d->states[0].elems, &follows); + state = state_index(d, &follows, 0, 0); if (wants_newline) - state_newline = state_index(r, &follows, 1, 0); + state_newline = state_index(d, &follows, 1, 0); else state_newline = state; if (wants_letter) - state_letter = state_index(r, &follows, 0, 1); + state_letter = state_index(d, &follows, 0, 1); else state_letter = state; - for (i = 0; i < _NOTCHAR; ++i) - trans[i] = (ISALNUM(i)) ? state_letter : state ; - trans['\n'] = state_newline; + for (i = 0; i < NOTCHAR; ++i) + if (i == '\n') + trans[i] = state_newline; + else if (ISALNUM(i)) + trans[i] = state_letter; + else + trans[i] = state; } else - for (i = 0; i < _NOTCHAR; ++i) + for (i = 0; i < NOTCHAR; ++i) trans[i] = -1; for (i = 0; i < ngrps; ++i) @@ -1431,44 +1677,44 @@ regstate(s, r, trans) /* Find the union of the follows of the positions of the group. This is a hideously inefficient loop. Fix it someday. */ for (j = 0; j < grps[i].nelem; ++j) - for (k = 0; k < r->follows[grps[i].elems[j].index].nelem; ++k) - insert(r->follows[grps[i].elems[j].index].elems[k], &follows); + for (k = 0; k < d->follows[grps[i].elems[j].index].nelem; ++k) + insert(d->follows[grps[i].elems[j].index].elems[k], &follows); /* If we are building a searching matcher, throw in the positions of state 0 as well. */ - if (r->searchflag) - for (j = 0; j < r->states[0].elems.nelem; ++j) - insert(r->states[0].elems.elems[j], &follows); + if (d->searchflag) + for (j = 0; j < d->states[0].elems.nelem; ++j) + insert(d->states[0].elems.elems[j], &follows); /* Find out if the new state will want any context information. */ wants_newline = 0; if (tstbit('\n', labels[i])) for (j = 0; j < follows.nelem; ++j) - if (_PREV_NEWLINE_DEPENDENT(follows.elems[j].constraint)) + if (PREV_NEWLINE_DEPENDENT(follows.elems[j].constraint)) wants_newline = 1; wants_letter = 0; - for (j = 0; j < _CHARSET_INTS; ++j) + for (j = 0; j < CHARCLASS_INTS; ++j) if (labels[i][j] & letters[j]) break; - if (j < _CHARSET_INTS) + if (j < CHARCLASS_INTS) for (j = 0; j < follows.nelem; ++j) - if (_PREV_LETTER_DEPENDENT(follows.elems[j].constraint)) + if (PREV_LETTER_DEPENDENT(follows.elems[j].constraint)) wants_letter = 1; /* Find the state(s) corresponding to the union of the follows. */ - state = state_index(r, &follows, 0, 0); + state = state_index(d, &follows, 0, 0); if (wants_newline) - state_newline = state_index(r, &follows, 1, 0); + state_newline = state_index(d, &follows, 1, 0); else state_newline = state; if (wants_letter) - state_letter = state_index(r, &follows, 0, 1); + state_letter = state_index(d, &follows, 0, 1); else state_letter = state; /* Set the transitions for each character in the current label. */ - for (j = 0; j < _CHARSET_INTS; ++j) + for (j = 0; j < CHARCLASS_INTS; ++j) for (k = 0; k < INTBITS; ++k) if (labels[i][j] & 1 << k) { @@ -1478,7 +1724,7 @@ regstate(s, r, trans) trans[c] = state_newline; else if (ISALNUM(c)) trans[c] = state_letter; - else if (c < _NOTCHAR) + else if (c < NOTCHAR) trans[c] = state; } } @@ -1488,18 +1734,18 @@ regstate(s, r, trans) free(follows.elems); free(tmp.elems); } - -/* Some routines for manipulating a compiled regexp's transition tables. + +/* Some routines for manipulating a compiled dfa's transition tables. Each state may or may not have a transition table; if it does, and it - is a non-accepting state, then r->trans[state] points to its table. - If it is an accepting state then r->fails[state] points to its table. - If it has no table at all, then r->trans[state] is NULL. + is a non-accepting state, then d->trans[state] points to its table. + If it is an accepting state then d->fails[state] points to its table. + If it has no table at all, then d->trans[state] is NULL. TODO: Improve this comment, get rid of the unnecessary redundancy. */ static void -build_state(s, r) +build_state(s, d) int s; - struct regexp *r; + struct dfa *d; { int *trans; /* The new transition table. */ int i; @@ -1508,87 +1754,87 @@ build_state(s, r) exist at once. 1024 is arbitrary. The idea is that the frequently used transition tables will be quickly rebuilt, whereas the ones that were only needed once or twice will be cleared away. */ - if (r->trcount >= 1024) + if (d->trcount >= 1024) { - for (i = 0; i < r->tralloc; ++i) - if (r->trans[i]) + for (i = 0; i < d->tralloc; ++i) + if (d->trans[i]) { - free((ptr_t) r->trans[i]); - r->trans[i] = NULL; + free((ptr_t) d->trans[i]); + d->trans[i] = NULL; } - else if (r->fails[i]) + else if (d->fails[i]) { - free((ptr_t) r->fails[i]); - r->fails[i] = NULL; + free((ptr_t) d->fails[i]); + d->fails[i] = NULL; } - r->trcount = 0; + d->trcount = 0; } - ++r->trcount; + ++d->trcount; /* Set up the success bits for this state. */ - r->success[s] = 0; - if (ACCEPTS_IN_CONTEXT(r->states[s].newline, 1, r->states[s].letter, 0, - s, *r)) - r->success[s] |= 4; - if (ACCEPTS_IN_CONTEXT(r->states[s].newline, 0, r->states[s].letter, 1, - s, *r)) - r->success[s] |= 2; - if (ACCEPTS_IN_CONTEXT(r->states[s].newline, 0, r->states[s].letter, 0, - s, *r)) - r->success[s] |= 1; - - MALLOC(trans, int, _NOTCHAR); - regstate(s, r, trans); + d->success[s] = 0; + if (ACCEPTS_IN_CONTEXT(d->states[s].newline, 1, d->states[s].letter, 0, + s, *d)) + d->success[s] |= 4; + if (ACCEPTS_IN_CONTEXT(d->states[s].newline, 0, d->states[s].letter, 1, + s, *d)) + d->success[s] |= 2; + if (ACCEPTS_IN_CONTEXT(d->states[s].newline, 0, d->states[s].letter, 0, + s, *d)) + d->success[s] |= 1; + + MALLOC(trans, int, NOTCHAR); + dfastate(s, d, trans); /* Now go through the new transition table, and make sure that the trans and fail arrays are allocated large enough to hold a pointer for the largest state mentioned in the table. */ - for (i = 0; i < _NOTCHAR; ++i) - if (trans[i] >= r->tralloc) + for (i = 0; i < NOTCHAR; ++i) + if (trans[i] >= d->tralloc) { - int oldalloc = r->tralloc; - - while (trans[i] >= r->tralloc) - r->tralloc *= 2; - REALLOC(r->realtrans, int *, r->tralloc + 1); - r->trans = r->realtrans + 1; - REALLOC(r->fails, int *, r->tralloc); - REALLOC(r->success, int, r->tralloc); - REALLOC(r->newlines, int, r->tralloc); - while (oldalloc < r->tralloc) + int oldalloc = d->tralloc; + + while (trans[i] >= d->tralloc) + d->tralloc *= 2; + REALLOC(d->realtrans, int *, d->tralloc + 1); + d->trans = d->realtrans + 1; + REALLOC(d->fails, int *, d->tralloc); + REALLOC(d->success, int, d->tralloc); + REALLOC(d->newlines, int, d->tralloc); + while (oldalloc < d->tralloc) { - r->trans[oldalloc] = NULL; - r->fails[oldalloc++] = NULL; + d->trans[oldalloc] = NULL; + d->fails[oldalloc++] = NULL; } } /* Keep the newline transition in a special place so we can use it as a sentinel. */ - r->newlines[s] = trans['\n']; + d->newlines[s] = trans['\n']; trans['\n'] = -1; - if (ACCEPTING(s, *r)) - r->fails[s] = trans; + if (ACCEPTING(s, *d)) + d->fails[s] = trans; else - r->trans[s] = trans; + d->trans[s] = trans; } static void -build_state_zero(r) - struct regexp *r; +build_state_zero(d) + struct dfa *d; { - r->tralloc = 1; - r->trcount = 0; - CALLOC(r->realtrans, int *, r->tralloc + 1); - r->trans = r->realtrans + 1; - CALLOC(r->fails, int *, r->tralloc); - MALLOC(r->success, int, r->tralloc); - MALLOC(r->newlines, int, r->tralloc); - build_state(0, r); + d->tralloc = 1; + d->trcount = 0; + CALLOC(d->realtrans, int *, d->tralloc + 1); + d->trans = d->realtrans + 1; + CALLOC(d->fails, int *, d->tralloc); + MALLOC(d->success, int, d->tralloc); + MALLOC(d->newlines, int, d->tralloc); + build_state(0, d); } - -/* Search through a buffer looking for a match to the given struct regexp. + +/* Search through a buffer looking for a match to the given struct dfa. Find the first occurrence of a string matching the regexp in the buffer, and the shortest possible version thereof. Return a pointer to the first character after the match, or NULL if none is found. Begin points to @@ -1602,8 +1848,8 @@ build_state_zero(r) match needs to be verified by a backtracking matcher. Otherwise we store a 0 in *backref. */ char * -regexecute(r, begin, end, newline, count, backref) - struct regexp *r; +dfaexec(d, begin, end, newline, count, backref) + struct dfa *d; char *begin; char *end; int newline; @@ -1612,9 +1858,9 @@ regexecute(r, begin, end, newline, count, backref) { register s, s1, tmp; /* Current state. */ register unsigned char *p; /* Current input character. */ - register **trans, *t; /* Copy of r->trans so it can be optimized + register **trans, *t; /* Copy of d->trans so it can be optimized into a register. */ - static sbit[_NOTCHAR]; /* Table for anding with r->success. */ + static sbit[NOTCHAR]; /* Table for anding with d->success. */ static sbit_init; if (! sbit_init) @@ -1622,41 +1868,54 @@ regexecute(r, begin, end, newline, count, backref) int i; sbit_init = 1; - for (i = 0; i < _NOTCHAR; ++i) - sbit[i] = (ISALNUM(i)) ? 2 : 1; - sbit['\n'] = 4; + for (i = 0; i < NOTCHAR; ++i) + if (i == '\n') + sbit[i] = 4; + else if (ISALNUM(i)) + sbit[i] = 2; + else + sbit[i] = 1; } - if (! r->tralloc) - build_state_zero(r); + if (! d->tralloc) + build_state_zero(d); s = s1 = 0; p = (unsigned char *) begin; - trans = r->trans; + trans = d->trans; *end = '\n'; for (;;) { - while ((t = trans[s]) != 0) { /* hand-optimized loop */ - s1 = t[*p++]; - if ((t = trans[s1]) == 0) { - tmp = s ; s = s1 ; s1 = tmp ; /* swap */ - break; - } - s = t[*p++]; - } + /* The dreaded inner loop. */ + if ((t = trans[s]) != 0) + do + { + s1 = t[*p++]; + if (! (t = trans[s1])) + goto last_was_s; + s = t[*p++]; + } + while ((t = trans[s]) != 0); + goto last_was_s1; + last_was_s: + tmp = s, s = s1, s1 = tmp; + last_was_s1: - if (s >= 0 && p <= (unsigned char *) end && r->fails[s]) + if (s >= 0 && p <= (unsigned char *) end && d->fails[s]) { - if (r->success[s] & sbit[*p]) + if (d->success[s] & sbit[*p]) { if (backref) - *backref = (r->states[s].backref != 0); + if (d->states[s].backref) + *backref = 1; + else + *backref = 0; return (char *) p; } s1 = s; - s = r->fails[s][*p++]; + s = d->fails[s][*p++]; continue; } @@ -1665,627 +1924,662 @@ regexecute(r, begin, end, newline, count, backref) ++*count; /* Check if we've run off the end of the buffer. */ - if ((char *) p >= end) + if ((char *) p > end) return NULL; if (s >= 0) { - build_state(s, r); - trans = r->trans; + build_state(s, d); + trans = d->trans; continue; } if (p[-1] == '\n' && newline) { - s = r->newlines[s1]; + s = d->newlines[s1]; continue; } s = 0; } } - -/* Initialize the components of a regexp that the other routines don't + +/* Initialize the components of a dfa that the other routines don't initialize for themselves. */ void -reginit(r) - struct regexp *r; +dfainit(d) + struct dfa *d; { - r->calloc = 1; - MALLOC(r->charsets, _charset, r->calloc); - r->cindex = 0; + d->calloc = 1; + MALLOC(d->charclasses, charclass, d->calloc); + d->cindex = 0; - r->talloc = 1; - MALLOC(r->tokens, _token, r->talloc); - r->tindex = r->depth = r->nleaves = r->nregexps = 0; + d->talloc = 1; + MALLOC(d->tokens, token, d->talloc); + d->tindex = d->depth = d->nleaves = d->nregexps = 0; - r->searchflag = 0; - r->tralloc = 0; + d->searchflag = 0; + d->tralloc = 0; + + d->musts = 0; } /* Parse and analyze a single string of the given length. */ void -regcompile(s, len, r, searchflag) - const char *s; +dfacomp(s, len, d, searchflag) + char *s; size_t len; - struct regexp *r; + struct dfa *d; int searchflag; { - if (case_fold) /* dummy folding in service of regmust() */ + if (case_fold) /* dummy folding in service of dfamust() */ { - char *regcopy; + char *lcopy; int i; - regcopy = malloc(len); - if (!regcopy) - reg_error("out of memory"); + lcopy = malloc(len); + if (!lcopy) + dfaerror("out of memory"); - /* This is a complete kludge and could potentially break - \<letter> escapes . . . */ + /* This is a kludge. */ case_fold = 0; for (i = 0; i < len; ++i) if (ISUPPER(s[i])) - regcopy[i] = tolower(s[i]); + lcopy[i] = tolower(s[i]); else - regcopy[i] = s[i]; - - reginit(r); - r->mustn = 0; - r->must[0] = '\0'; - regparse(regcopy, len, r); - free(regcopy); - regmust(r); - reganalyze(r, searchflag); + lcopy[i] = s[i]; + + dfainit(d); + dfaparse(lcopy, len, d); + free(lcopy); + dfamust(d); + d->cindex = d->tindex = d->depth = d->nleaves = d->nregexps = 0; case_fold = 1; - reginit(r); - regparse(s, len, r); - reganalyze(r, searchflag); + dfaparse(s, len, d); + dfaanalyze(d, searchflag); } else { - reginit(r); - regparse(s, len, r); - regmust(r); - reganalyze(r, searchflag); + dfainit(d); + dfaparse(s, len, d); + dfamust(d); + dfaanalyze(d, searchflag); } } -/* Free the storage held by the components of a regexp. */ +/* Free the storage held by the components of a dfa. */ void -reg_free(r) - struct regexp *r; +dfafree(d) + struct dfa *d; { int i; - - free((ptr_t) r->charsets); - free((ptr_t) r->tokens); - for (i = 0; i < r->sindex; ++i) - free((ptr_t) r->states[i].elems.elems); - free((ptr_t) r->states); - for (i = 0; i < r->tindex; ++i) - if (r->follows[i].elems) - free((ptr_t) r->follows[i].elems); - free((ptr_t) r->follows); - for (i = 0; i < r->tralloc; ++i) - if (r->trans[i]) - free((ptr_t) r->trans[i]); - else if (r->fails[i]) - free((ptr_t) r->fails[i]); - if (r->realtrans) - free((ptr_t) r->realtrans); - if (r->fails) - free((ptr_t) r->fails); - if (r->newlines) - free((ptr_t) r->newlines); + struct dfamust *dm, *ndm; + + free((ptr_t) d->charclasses); + free((ptr_t) d->tokens); + for (i = 0; i < d->sindex; ++i) + free((ptr_t) d->states[i].elems.elems); + free((ptr_t) d->states); + for (i = 0; i < d->tindex; ++i) + if (d->follows[i].elems) + free((ptr_t) d->follows[i].elems); + free((ptr_t) d->follows); + for (i = 0; i < d->tralloc; ++i) + if (d->trans[i]) + free((ptr_t) d->trans[i]); + else if (d->fails[i]) + free((ptr_t) d->fails[i]); + free((ptr_t) d->realtrans); + free((ptr_t) d->fails); + free((ptr_t) d->newlines); + for (dm = d->musts; dm; dm = ndm) + { + ndm = dm->next; + free(dm->must); + free((ptr_t) dm); + } } -/* -Having found the postfix representation of the regular expression, -try to find a long sequence of characters that must appear in any line -containing the r.e. -Finding a "longest" sequence is beyond the scope here; -we take an easy way out and hope for the best. -(Take "(ab|a)b"--please.) - -We do a bottom-up calculation of sequences of characters that must appear -in matches of r.e.'s represented by trees rooted at the nodes of the postfix -representation: +/* Having found the postfix representation of the regular expression, + try to find a long sequence of characters that must appear in any line + containing the r.e. + Finding a "longest" sequence is beyond the scope here; + we take an easy way out and hope for the best. + (Take "(ab|a)b"--please.) + + We do a bottom-up calculation of sequences of characters that must appear + in matches of r.e.'s represented by trees rooted at the nodes of the postfix + representation: sequences that must appear at the left of the match ("left") sequences that must appear at the right of the match ("right") lists of sequences that must appear somewhere in the match ("in") sequences that must constitute the match ("is") -When we get to the root of the tree, we use one of the longest of its -calculated "in" sequences as our answer. The sequence we find is returned in -r->must (where "r" is the single argument passed to "regmust"); -the length of the sequence is returned in r->mustn. - -The sequences calculated for the various types of node (in pseudo ANSI c) -are shown below. "p" is the operand of unary operators (and the left-hand -operand of binary operators); "q" is the right-hand operand of binary operators -. -"ZERO" means "a zero-length sequence" below. - -Type left right is in ----- ---- ----- -- -- -char c # c # c # c # c -SET ZERO ZERO ZERO ZERO - -STAR ZERO ZERO ZERO ZERO - -QMARK ZERO ZERO ZERO ZERO - -PLUS p->left p->right ZERO p->in - -CAT (p->is==ZERO)? (q->is==ZERO)? (p->is!=ZERO && p->in plus - p->left : q->right : q->is!=ZERO) ? q->in plus - p->is##q->left p->right##q->is p->is##q->is : p->right##q->left - ZERO - -OR longest common longest common (do p->is and substrings common to - leading trailing q->is have same p->in and q->in - (sub)sequence (sub)sequence length and - of p->left of p->right content) ? - and q->left and q->right p->is : NULL - -If there's anything else we recognize in the tree, all four sequences get set -to zero-length sequences. If there's something we don't recognize in the tree, -we just return a zero-length sequence. - -Break ties in favor of infrequent letters (choosing 'zzz' in preference to -'aaa')? - -And. . .is it here or someplace that we might ponder "optimizations" such as + When we get to the root of the tree, we use one of the longest of its + calculated "in" sequences as our answer. The sequence we find is returned in + d->must (where "d" is the single argument passed to "dfamust"); + the length of the sequence is returned in d->mustn. + + The sequences calculated for the various types of node (in pseudo ANSI c) + are shown below. "p" is the operand of unary operators (and the left-hand + operand of binary operators); "q" is the right-hand operand of binary + operators. + + "ZERO" means "a zero-length sequence" below. + + Type left right is in + ---- ---- ----- -- -- + char c # c # c # c # c + + CSET ZERO ZERO ZERO ZERO + + STAR ZERO ZERO ZERO ZERO + + QMARK ZERO ZERO ZERO ZERO + + PLUS p->left p->right ZERO p->in + + CAT (p->is==ZERO)? (q->is==ZERO)? (p->is!=ZERO && p->in plus + p->left : q->right : q->is!=ZERO) ? q->in plus + p->is##q->left p->right##q->is p->is##q->is : p->right##q->left + ZERO + + OR longest common longest common (do p->is and substrings common to + leading trailing q->is have same p->in and q->in + (sub)sequence (sub)sequence length and + of p->left of p->right content) ? + and q->left and q->right p->is : NULL + + If there's anything else we recognize in the tree, all four sequences get set + to zero-length sequences. If there's something we don't recognize in the tree, + we just return a zero-length sequence. + + Break ties in favor of infrequent letters (choosing 'zzz' in preference to + 'aaa')? + + And. . .is it here or someplace that we might ponder "optimizations" such as egrep 'psi|epsilon' -> egrep 'psi' egrep 'pepsi|epsilon' -> egrep 'epsi' (Yes, we now find "epsi" as a "string that must occur", but we might also - simplify the *entire* r.e. being sought -) + simplify the *entire* r.e. being sought) grep '[c]' -> grep 'c' grep '(ab|a)b' -> grep 'ab' grep 'ab*' -> grep 'a' grep 'a*b' -> grep 'b' -There are several issues: - Is optimization easy (enough)? - Does optimization actually accomplish anything, - or is the automaton you get from "psi|epsilon" (for example) - the same as the one you get from "psi" (for example)? + There are several issues: + + Is optimization easy (enough)? - Are optimizable r.e.'s likely to be used in real-life situations - (something like 'ab*' is probably unlikely; something like is - 'psi|epsilon' is likelier)? -*/ + Does optimization actually accomplish anything, + or is the automaton you get from "psi|epsilon" (for example) + the same as the one you get from "psi" (for example)? + + Are optimizable r.e.'s likely to be used in real-life situations + (something like 'ab*' is probably unlikely; something like is + 'psi|epsilon' is likelier)? */ static char * icatalloc(old, new) -char * old; -const char * new; + char *old; + char *new; { - register char * result; - register size_t oldsize, newsize; - - newsize = (new == NULL) ? 0 : strlen(new); - if (old == NULL) - oldsize = 0; - else if (newsize == 0) - return old; - else oldsize = strlen(old); - if (old == NULL) - result = (char *) malloc(newsize + 1); - else result = (char *) realloc((void *) old, oldsize + newsize + 1); - if (result != NULL && new != NULL) - (void) strcpy(result + oldsize, new); - return result; + char *result; + size_t oldsize, newsize; + + newsize = (new == NULL) ? 0 : strlen(new); + if (old == NULL) + oldsize = 0; + else if (newsize == 0) + return old; + else oldsize = strlen(old); + if (old == NULL) + result = (char *) malloc(newsize + 1); + else + result = (char *) realloc((void *) old, oldsize + newsize + 1); + if (result != NULL && new != NULL) + (void) strcpy(result + oldsize, new); + return result; } static char * icpyalloc(string) -const char * string; + char *string; { - return icatalloc((char *) NULL, string); + return icatalloc((char *) NULL, string); } static char * istrstr(lookin, lookfor) -char * lookin; -register char * lookfor; + char *lookin; + char *lookfor; { - register char * cp; - register size_t len; - - len = strlen(lookfor); - for (cp = lookin; *cp != '\0'; ++cp) - if (strncmp(cp, lookfor, len) == 0) - return cp; - return NULL; + char *cp; + size_t len; + + len = strlen(lookfor); + for (cp = lookin; *cp != '\0'; ++cp) + if (strncmp(cp, lookfor, len) == 0) + return cp; + return NULL; } static void ifree(cp) -char * cp; + char *cp; { - if (cp != NULL) - free(cp); + if (cp != NULL) + free(cp); } static void freelist(cpp) -register char ** cpp; + char **cpp; { - register int i; + int i; - if (cpp == NULL) - return; - for (i = 0; cpp[i] != NULL; ++i) { - free(cpp[i]); - cpp[i] = NULL; - } + if (cpp == NULL) + return; + for (i = 0; cpp[i] != NULL; ++i) + { + free(cpp[i]); + cpp[i] = NULL; + } } static char ** enlist(cpp, new, len) -register char ** cpp; -register char * new; -#ifdef __STDC__ -size_t len; -#else -int len; -#endif + char **cpp; + char *new; + size_t len; { - register int i, j; + int i, j; - if (cpp == NULL) - return NULL; - if ((new = icpyalloc(new)) == NULL) { - freelist(cpp); - return NULL; - } - new[len] = '\0'; - /* - ** Is there already something in the list that's new (or longer)? - */ - for (i = 0; cpp[i] != NULL; ++i) - if (istrstr(cpp[i], new) != NULL) { - free(new); - return cpp; - } - /* - ** Eliminate any obsoleted strings. - */ - j = 0; - while (cpp[j] != NULL) - if (istrstr(new, cpp[j]) == NULL) - ++j; - else { - free(cpp[j]); - if (--i == j) - break; - cpp[j] = cpp[i]; - } - /* - ** Add the new string. - */ - cpp = (char **) realloc((char *) cpp, (i + 2) * sizeof *cpp); - if (cpp == NULL) - return NULL; - cpp[i] = new; - cpp[i + 1] = NULL; + if (cpp == NULL) + return NULL; + if ((new = icpyalloc(new)) == NULL) + { + freelist(cpp); + return NULL; + } + new[len] = '\0'; + /* Is there already something in the list that's new (or longer)? */ + for (i = 0; cpp[i] != NULL; ++i) + if (istrstr(cpp[i], new) != NULL) + { + free(new); return cpp; + } + /* Eliminate any obsoleted strings. */ + j = 0; + while (cpp[j] != NULL) + if (istrstr(new, cpp[j]) == NULL) + ++j; + else + { + free(cpp[j]); + if (--i == j) + break; + cpp[j] = cpp[i]; + cpp[i] = NULL; + } + /* Add the new string. */ + cpp = (char **) realloc((char *) cpp, (i + 2) * sizeof *cpp); + if (cpp == NULL) + return NULL; + cpp[i] = new; + cpp[i + 1] = NULL; + return cpp; } -/* -** Given pointers to two strings, -** return a pointer to an allocated list of their distinct common substrings. -** Return NULL if something seems wild. -*/ - +/* Given pointers to two strings, return a pointer to an allocated + list of their distinct common substrings. Return NULL if something + seems wild. */ static char ** comsubs(left, right) -char * left; -char * right; + char *left; + char *right; { - register char ** cpp; - register char * lcp; - register char * rcp; - register int i, len; - - if (left == NULL || right == NULL) - return NULL; - cpp = (char **) malloc(sizeof *cpp); - if (cpp == NULL) - return NULL; - cpp[0] = NULL; - for (lcp = left; *lcp != '\0'; ++lcp) { - len = 0; - rcp = strchr(right, *lcp); - while (rcp != NULL) { - for (i = 1; lcp[i] != '\0' && lcp[i] == rcp[i]; ++i) - ; - if (i > len) - len = i; - rcp = strchr(rcp + 1, *lcp); - } - if (len == 0) - continue; -#ifdef __STDC__ - if ((cpp = enlist(cpp, lcp, (size_t)len)) == NULL) -#else - if ((cpp = enlist(cpp, lcp, len)) == NULL) -#endif - break; + char **cpp; + char *lcp; + char *rcp; + size_t i, len; + + if (left == NULL || right == NULL) + return NULL; + cpp = (char **) malloc(sizeof *cpp); + if (cpp == NULL) + return NULL; + cpp[0] = NULL; + for (lcp = left; *lcp != '\0'; ++lcp) + { + len = 0; + rcp = index(right, *lcp); + while (rcp != NULL) + { + for (i = 1; lcp[i] != '\0' && lcp[i] == rcp[i]; ++i) + ; + if (i > len) + len = i; + rcp = index(rcp + 1, *lcp); } - return cpp; + if (len == 0) + continue; + if ((cpp = enlist(cpp, lcp, len)) == NULL) + break; + } + return cpp; } static char ** addlists(old, new) -char ** old; -char ** new; +char **old; +char **new; { - register int i; - - if (old == NULL || new == NULL) - return NULL; - for (i = 0; new[i] != NULL; ++i) { - old = enlist(old, new[i], strlen(new[i])); - if (old == NULL) - break; - } - return old; -} + int i; -/* -** Given two lists of substrings, -** return a new list giving substrings common to both. -*/ + if (old == NULL || new == NULL) + return NULL; + for (i = 0; new[i] != NULL; ++i) + { + old = enlist(old, new[i], strlen(new[i])); + if (old == NULL) + break; + } + return old; +} +/* Given two lists of substrings, return a new list giving substrings + common to both. */ static char ** inboth(left, right) -char ** left; -char ** right; + char **left; + char **right; { - register char ** both; - register char ** temp; - register int lnum, rnum; - - if (left == NULL || right == NULL) - return NULL; - both = (char **) malloc(sizeof *both); - if (both == NULL) - return NULL; - both[0] = NULL; - for (lnum = 0; left[lnum] != NULL; ++lnum) { - for (rnum = 0; right[rnum] != NULL; ++rnum) { - temp = comsubs(left[lnum], right[rnum]); - if (temp == NULL) { - freelist(both); - return NULL; - } - both = addlists(both, temp); - freelist(temp); - if (both == NULL) - return NULL; - } + char **both; + char **temp; + int lnum, rnum; + + if (left == NULL || right == NULL) + return NULL; + both = (char **) malloc(sizeof *both); + if (both == NULL) + return NULL; + both[0] = NULL; + for (lnum = 0; left[lnum] != NULL; ++lnum) + { + for (rnum = 0; right[rnum] != NULL; ++rnum) + { + temp = comsubs(left[lnum], right[rnum]); + if (temp == NULL) + { + freelist(both); + return NULL; + } + both = addlists(both, temp); + freelist(temp); + if (both == NULL) + return NULL; } - return both; + } + return both; } -/* -typedef struct { - char ** in; - char * left; - char * right; - char * is; +typedef struct +{ + char **in; + char *left; + char *right; + char *is; } must; - */ + static void resetmust(mp) -register must * mp; +must *mp; { - mp->left[0] = mp->right[0] = mp->is[0] = '\0'; - freelist(mp->in); + mp->left[0] = mp->right[0] = mp->is[0] = '\0'; + freelist(mp->in); } static void -regmust(r) -register struct regexp * r; +dfamust(dfa) +struct dfa *dfa; { - register must * musts; - register must * mp; - register char * result = ""; - register int ri; - register int i; - register _token t; - static must must0; - - reg->mustn = 0; - reg->must[0] = '\0'; - musts = (must *) malloc((reg->tindex + 1) * sizeof *musts); - if (musts == NULL) - return; - mp = musts; - for (i = 0; i <= reg->tindex; ++i) - mp[i] = must0; - for (i = 0; i <= reg->tindex; ++i) { - mp[i].in = (char **) malloc(sizeof *mp[i].in); - mp[i].left = malloc((MALLOC_ARG_T)2); - mp[i].right = malloc((MALLOC_ARG_T)2); - mp[i].is = malloc((MALLOC_ARG_T)2); - if (mp[i].in == NULL || mp[i].left == NULL || - mp[i].right == NULL || mp[i].is == NULL) - goto done; - mp[i].left[0] = mp[i].right[0] = mp[i].is[0] = '\0'; - mp[i].in[0] = NULL; - } - for (ri = 0; ri < reg->tindex; ++ri) { - switch (t = reg->tokens[ri]) { - case _ALLBEGLINE: - case _ALLENDLINE: - case _LPAREN: - case _RPAREN: - goto done; /* "cannot happen" */ - case _EMPTY: - case _BEGLINE: - case _ENDLINE: - case _BEGWORD: - case _ENDWORD: - case _LIMWORD: - case _NOTLIMWORD: - case _BACKREF: - resetmust(mp); - break; - case _STAR: - case _QMARK: - if (mp <= musts) - goto done; /* "cannot happen" */ - --mp; - resetmust(mp); - break; - case _OR: - if (mp < &musts[2]) - goto done; /* "cannot happen" */ - { - register char ** new; - register must * lmp; - register must * rmp; - register int j, ln, rn, n; - - rmp = --mp; - lmp = --mp; - /* Guaranteed to be. Unlikely, but. . . */ - if (strcmp(lmp->is, rmp->is) != 0) - lmp->is[0] = '\0'; - /* Left side--easy */ - i = 0; - while (lmp->left[i] != '\0' && - lmp->left[i] == rmp->left[i]) - ++i; - lmp->left[i] = '\0'; - /* Right side */ - ln = strlen(lmp->right); - rn = strlen(rmp->right); - n = ln; - if (n > rn) - n = rn; - for (i = 0; i < n; ++i) - if (lmp->right[ln - i - 1] != - rmp->right[rn - i - 1]) - break; - for (j = 0; j < i; ++j) - lmp->right[j] = - lmp->right[(ln - i) + j]; - lmp->right[j] = '\0'; - new = inboth(lmp->in, rmp->in); - if (new == NULL) - goto done; - freelist(lmp->in); - free((char *) lmp->in); - lmp->in = new; - } - break; - case _PLUS: - if (mp <= musts) - goto done; /* "cannot happen" */ - --mp; - mp->is[0] = '\0'; - break; - case _END: - if (mp != &musts[1]) - goto done; /* "cannot happen" */ - for (i = 0; musts[0].in[i] != NULL; ++i) - if (strlen(musts[0].in[i]) > strlen(result)) - result = musts[0].in[i]; - goto done; - case _CAT: - if (mp < &musts[2]) - goto done; /* "cannot happen" */ - { - register must * lmp; - register must * rmp; - - rmp = --mp; - lmp = --mp; - /* - ** In. Everything in left, plus everything in - ** right, plus catenation of - ** left's right and right's left. - */ - lmp->in = addlists(lmp->in, rmp->in); - if (lmp->in == NULL) - goto done; - if (lmp->right[0] != '\0' && - rmp->left[0] != '\0') { - register char * tp; - - tp = icpyalloc(lmp->right); - if (tp == NULL) - goto done; - tp = icatalloc(tp, rmp->left); - if (tp == NULL) - goto done; - lmp->in = enlist(lmp->in, tp, - strlen(tp)); - free(tp); - if (lmp->in == NULL) - goto done; - } - /* Left-hand */ - if (lmp->is[0] != '\0') { - lmp->left = icatalloc(lmp->left, - rmp->left); - if (lmp->left == NULL) - goto done; - } - /* Right-hand */ - if (rmp->is[0] == '\0') - lmp->right[0] = '\0'; - lmp->right = icatalloc(lmp->right, rmp->right); - if (lmp->right == NULL) - goto done; - /* Guaranteed to be */ - if (lmp->is[0] != '\0' && rmp->is[0] != '\0') { - lmp->is = icatalloc(lmp->is, rmp->is); - if (lmp->is == NULL) - goto done; - } - } - break; - default: - if (t < _END) { - /* "cannot happen" */ - goto done; - } else if (t == '\0') { - /* not on *my* shift */ - goto done; - } else if (t >= _SET) { - /* easy enough */ - resetmust(mp); - } else { - /* plain character */ - resetmust(mp); - mp->is[0] = mp->left[0] = mp->right[0] = t; - mp->is[1] = mp->left[1] = mp->right[1] = '\0'; - mp->in = enlist(mp->in, mp->is, 1); - if (mp->in == NULL) - goto done; - } - break; - } - ++mp; - } -done: - (void) strncpy(reg->must, result, MUST_MAX - 1); - reg->must[MUST_MAX - 1] = '\0'; - reg->mustn = strlen(reg->must); - mp = musts; - for (i = 0; i <= reg->tindex; ++i) { - freelist(mp[i].in); - ifree((char *) mp[i].in); - ifree(mp[i].left); - ifree(mp[i].right); - ifree(mp[i].is); + must *musts; + must *mp; + char *result; + int ri; + int i; + int exact; + token t; + static must must0; + struct dfamust *dm; + static char empty_string[] = ""; + + result = empty_string; + exact = 0; + musts = (must *) malloc((dfa->tindex + 1) * sizeof *musts); + if (musts == NULL) + return; + mp = musts; + for (i = 0; i <= dfa->tindex; ++i) + mp[i] = must0; + for (i = 0; i <= dfa->tindex; ++i) + { + mp[i].in = (char **) malloc(sizeof *mp[i].in); + mp[i].left = malloc(2); + mp[i].right = malloc(2); + mp[i].is = malloc(2); + if (mp[i].in == NULL || mp[i].left == NULL || + mp[i].right == NULL || mp[i].is == NULL) + goto done; + mp[i].left[0] = mp[i].right[0] = mp[i].is[0] = '\0'; + mp[i].in[0] = NULL; + } +#ifdef DEBUG + fprintf(stderr, "dfamust:\n"); + for (i = 0; i < dfa->tindex; ++i) + { + fprintf(stderr, " %d:", i); + prtok(dfa->tokens[i]); + } + putc('\n', stderr); +#endif + for (ri = 0; ri < dfa->tindex; ++ri) + { + switch (t = dfa->tokens[ri]) + { + case LPAREN: + case RPAREN: + goto done; /* "cannot happen" */ + case EMPTY: + case BEGLINE: + case ENDLINE: + case BEGWORD: + case ENDWORD: + case LIMWORD: + case NOTLIMWORD: + case BACKREF: + resetmust(mp); + break; + case STAR: + case QMARK: + if (mp <= musts) + goto done; /* "cannot happen" */ + --mp; + resetmust(mp); + break; + case OR: + case ORTOP: + if (mp < &musts[2]) + goto done; /* "cannot happen" */ + { + char **new; + must *lmp; + must *rmp; + int j, ln, rn, n; + + rmp = --mp; + lmp = --mp; + /* Guaranteed to be. Unlikely, but. . . */ + if (strcmp(lmp->is, rmp->is) != 0) + lmp->is[0] = '\0'; + /* Left side--easy */ + i = 0; + while (lmp->left[i] != '\0' && lmp->left[i] == rmp->left[i]) + ++i; + lmp->left[i] = '\0'; + /* Right side */ + ln = strlen(lmp->right); + rn = strlen(rmp->right); + n = ln; + if (n > rn) + n = rn; + for (i = 0; i < n; ++i) + if (lmp->right[ln - i - 1] != rmp->right[rn - i - 1]) + break; + for (j = 0; j < i; ++j) + lmp->right[j] = lmp->right[(ln - i) + j]; + lmp->right[j] = '\0'; + new = inboth(lmp->in, rmp->in); + if (new == NULL) + goto done; + freelist(lmp->in); + free((char *) lmp->in); + lmp->in = new; + } + break; + case PLUS: + if (mp <= musts) + goto done; /* "cannot happen" */ + --mp; + mp->is[0] = '\0'; + break; + case END: + if (mp != &musts[1]) + goto done; /* "cannot happen" */ + for (i = 0; musts[0].in[i] != NULL; ++i) + if (strlen(musts[0].in[i]) > strlen(result)) + result = musts[0].in[i]; + if (strcmp(result, musts[0].is) == 0) + exact = 1; + goto done; + case CAT: + if (mp < &musts[2]) + goto done; /* "cannot happen" */ + { + must *lmp; + must *rmp; + + rmp = --mp; + lmp = --mp; + /* In. Everything in left, plus everything in + right, plus catenation of + left's right and right's left. */ + lmp->in = addlists(lmp->in, rmp->in); + if (lmp->in == NULL) + goto done; + if (lmp->right[0] != '\0' && + rmp->left[0] != '\0') + { + char *tp; + + tp = icpyalloc(lmp->right); + if (tp == NULL) + goto done; + tp = icatalloc(tp, rmp->left); + if (tp == NULL) + goto done; + lmp->in = enlist(lmp->in, tp, + strlen(tp)); + free(tp); + if (lmp->in == NULL) + goto done; + } + /* Left-hand */ + if (lmp->is[0] != '\0') + { + lmp->left = icatalloc(lmp->left, + rmp->left); + if (lmp->left == NULL) + goto done; + } + /* Right-hand */ + if (rmp->is[0] == '\0') + lmp->right[0] = '\0'; + lmp->right = icatalloc(lmp->right, rmp->right); + if (lmp->right == NULL) + goto done; + /* Guaranteed to be */ + if (lmp->is[0] != '\0' && rmp->is[0] != '\0') + { + lmp->is = icatalloc(lmp->is, rmp->is); + if (lmp->is == NULL) + goto done; + } + else + lmp->is[0] = '\0'; + } + break; + default: + if (t < END) + { + /* "cannot happen" */ + goto done; + } + else if (t == '\0') + { + /* not on *my* shift */ + goto done; + } + else if (t >= CSET) + { + /* easy enough */ + resetmust(mp); + } + else + { + /* plain character */ + resetmust(mp); + mp->is[0] = mp->left[0] = mp->right[0] = t; + mp->is[1] = mp->left[1] = mp->right[1] = '\0'; + mp->in = enlist(mp->in, mp->is, (size_t)1); + if (mp->in == NULL) + goto done; + } + break; } - free((char *) mp); +#ifdef DEBUG + fprintf(stderr, " node: %d:", ri); + prtok(dfa->tokens[ri]); + fprintf(stderr, "\n in:"); + for (i = 0; mp->in[i]; ++i) + fprintf(stderr, " \"%s\"", mp->in[i]); + fprintf(stderr, "\n is: \"%s\"\n", mp->is); + fprintf(stderr, " left: \"%s\"\n", mp->left); + fprintf(stderr, " right: \"%s\"\n", mp->right); +#endif + ++mp; + } + done: + if (strlen(result)) + { + dm = (struct dfamust *) malloc(sizeof (struct dfamust)); + dm->exact = exact; + dm->must = malloc(strlen(result) + 1); + strcpy(dm->must, result); + dm->next = dfa->musts; + dfa->musts = dm; + } + mp = musts; + for (i = 0; i <= dfa->tindex; ++i) + { + freelist(mp[i].in); + ifree((char *) mp[i].in); + ifree(mp[i].left); + ifree(mp[i].right); + ifree(mp[i].is); + } + free((char *) mp); } |