rsalz@uunet.uu.net (Rich Salz) (02/10/89)
Submitted-by: Mike Haertel <mike@wheaties.ai.mit.edu>
Posting-number: Volume 17, Issue 101
Archive-name: gnugrep/part04
#! /bin/sh
# This is a shell archive. Remove anything before this line, then unpack
# it by saving it into a file and typing "sh file". To overwrite existing
# files, type "sh file -c". You can also feed this as standard input via
# unshar, or by typing "sh <file", e.g.. If this archive is complete, you
# will see the following message at the end:
# "End of shell archive."
# Contents: dfa.c
PATH=/bin:/usr/bin:/usr/ucb ; export PATH
if test -f 'dfa.c' -a "${1}" != "-c" ; then
echo shar: Will not clobber existing file \"'dfa.c'\"
else
echo shar: Extracting \"'dfa.c'\" \(54865 characters\)
sed "s/^X//" >'dfa.c' <<'END_OF_FILE'
X/* dfa.c - determinisitic extended regexp routines for GNU
X Copyright (C) 1988 Free Software Foundation, Inc.
X Written June, 1988 by Mike Haertel
X Modified July, 1988 by Arthur David Olson
X to assist BMG speedups
X
X NO WARRANTY
X
X BECAUSE THIS PROGRAM IS LICENSED FREE OF CHARGE, WE PROVIDE ABSOLUTELY
XNO WARRANTY, TO THE EXTENT PERMITTED BY APPLICABLE STATE LAW. EXCEPT
XWHEN OTHERWISE STATED IN WRITING, FREE SOFTWARE FOUNDATION, INC,
XRICHARD M. STALLMAN AND/OR OTHER PARTIES PROVIDE THIS PROGRAM "AS IS"
XWITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING,
XBUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
XFITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY
XAND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE
XDEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR
XCORRECTION.
X
X IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW WILL RICHARD M.
XSTALLMAN, THE FREE SOFTWARE FOUNDATION, INC., AND/OR ANY OTHER PARTY
XWHO MAY MODIFY AND REDISTRIBUTE THIS PROGRAM AS PERMITTED BELOW, BE
XLIABLE TO YOU FOR DAMAGES, INCLUDING ANY LOST PROFITS, LOST MONIES, OR
XOTHER SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
XUSE OR INABILITY TO USE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR
XDATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY THIRD PARTIES OR
XA FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS) THIS
XPROGRAM, EVEN IF YOU HAVE BEEN ADVISED OF THE POSSIBILITY OF SUCH
XDAMAGES, OR FOR ANY CLAIM BY ANY OTHER PARTY.
X
X GENERAL PUBLIC LICENSE TO COPY
X
X 1. You may copy and distribute verbatim copies of this source file
Xas you receive it, in any medium, provided that you conspicuously and
Xappropriately publish on each copy a valid copyright notice "Copyright
X (C) 1988 Free Software Foundation, Inc."; and include following the
Xcopyright notice a verbatim copy of the above disclaimer of warranty
Xand of this License. You may charge a distribution fee for the
Xphysical act of transferring a copy.
X
X 2. You may modify your copy or copies of this source file or
Xany portion of it, and copy and distribute such modifications under
Xthe terms of Paragraph 1 above, provided that you also do the following:
X
X a) cause the modified files to carry prominent notices stating
X that you changed the files and the date of any change; and
X
X b) cause the whole of any work that you distribute or publish,
X that in whole or in part contains or is a derivative of this
X program or any part thereof, to be licensed at no charge to all
X third parties on terms identical to those contained in this
X License Agreement (except that you may choose to grant more extensive
X warranty protection to some or all third parties, at your option).
X
X c) You may charge a distribution fee for the physical act of
X transferring a copy, and you may at your option offer warranty
X protection in exchange for a fee.
X
XMere aggregation of another unrelated program with this program (or its
Xderivative) on a volume of a storage or distribution medium does not bring
Xthe other program under the scope of these terms.
X
X 3. You may copy and distribute this program or any portion of it in
Xcompiled, executable or object code form under the terms of Paragraphs
X1 and 2 above provided that you do the following:
X
X a) accompany it with the complete corresponding machine-readable
X source code, which must be distributed under the terms of
X Paragraphs 1 and 2 above; or,
X
X b) accompany it with a written offer, valid for at least three
X years, to give any third party free (except for a nominal
X shipping charge) a complete machine-readable copy of the
X corresponding source code, to be distributed under the terms of
X Paragraphs 1 and 2 above; or,
X
X c) accompany it with the information you received as to where the
X corresponding source code may be obtained. (This alternative is
X allowed only for noncommercial distribution and only if you
X received the program in object code or executable form alone.)
X
XFor an executable file, complete source code means all the source code for
Xall modules it contains; but, as a special exception, it need not include
Xsource code for modules which are standard libraries that accompany the
Xoperating system on which the executable file runs.
X
X 4. You may not copy, sublicense, distribute or transfer this program
Xexcept as expressly provided under this License Agreement. Any attempt
Xotherwise to copy, sublicense, distribute or transfer this program is void and
Xyour rights to use the program under this License agreement shall be
Xautomatically terminated. However, parties who have received computer
Xsoftware programs from you with this License Agreement will not have
Xtheir licenses terminated so long as such parties remain in full compliance.
X
X 5. If you wish to incorporate parts of this program into other free
Xprograms whose distribution conditions are different, write to the Free
XSoftware Foundation at 675 Mass Ave, Cambridge, MA 02139. We have not yet
Xworked out a simple rule that can be stated here, but we will often permit
Xthis. We will be guided by the two goals of preserving the free status of
Xall derivatives our free software and of promoting the sharing and reuse of
Xsoftware.
X
X
XIn other words, you are welcome to use, share and improve this program.
XYou are forbidden to forbid anyone else to use, share and improve
Xwhat you give them. Help stamp out software-hoarding! */
X
X#include <ctype.h>
X#include "dfa.h"
X
X#ifdef __STDC__
Xtypedef void *ptr_t;
X#else
Xtypedef char *ptr_t;
X#endif
X
Xstatic void regmust();
X
Xstatic ptr_t
Xxcalloc(n, s)
X int n;
X size_t s;
X{
X ptr_t r = calloc(n, s);
X
X if (r)
X return r;
X else
X regerror("Memory exhausted");
X}
X
Xstatic ptr_t
Xxmalloc(n)
X size_t n;
X{
X ptr_t r = malloc(n);
X
X if (r)
X return r;
X else
X regerror("Memory exhausted");
X}
X
Xstatic ptr_t
Xxrealloc(p, n)
X ptr_t p;
X size_t n;
X{
X ptr_t r = realloc(p, n);
X
X if (r)
X return r;
X else
X regerror("Memory exhausted");
X}
X
X#define CALLOC(p, t, n) ((p) = (t *) xcalloc((n), sizeof (t)))
X#define MALLOC(p, t, n) ((p) = (t *) xmalloc((n) * sizeof (t)))
X#define REALLOC(p, t, n) ((p) = (t *) xrealloc((ptr_t) (p), (n) * sizeof (t)))
X
X/* Reallocate an array of type t if nalloc is too small for index. */
X#define REALLOC_IF_NECESSARY(p, t, nalloc, index) \
X if ((index) >= (nalloc)) \
X { \
X while ((index) >= (nalloc)) \
X (nalloc) *= 2; \
X REALLOC(p, t, nalloc); \
X }
X
X/* Stuff pertaining to charsets. */
X
Xstatic
Xtstbit(b, c)
X int b;
X _charset c;
X{
X return c[b / INTBITS] & 1 << b % INTBITS;
X}
X
Xstatic void
Xsetbit(b, c)
X int b;
X _charset c;
X{
X c[b / INTBITS] |= 1 << b % INTBITS;
X}
X
Xstatic void
Xclrbit(b, c)
X int b;
X _charset c;
X{
X c[b / INTBITS] &= ~(1 << b % INTBITS);
X}
X
Xstatic void
Xcopyset(src, dst)
X const _charset src;
X _charset dst;
X{
X int i;
X
X for (i = 0; i < _CHARSET_INTS; ++i)
X dst[i] = src[i];
X}
X
Xstatic void
Xzeroset(s)
X _charset s;
X{
X int i;
X
X for (i = 0; i < _CHARSET_INTS; ++i)
X s[i] = 0;
X}
X
Xstatic void
Xnotset(s)
X _charset s;
X{
X int i;
X
X for (i = 0; i < _CHARSET_INTS; ++i)
X s[i] = ~s[i];
X}
X
Xstatic
Xequal(s1, s2)
X const _charset s1;
X const _charset s2;
X{
X int i;
X
X for (i = 0; i < _CHARSET_INTS; ++i)
X if (s1[i] != s2[i])
X return 0;
X return 1;
X}
X
X/* A pointer to the current regexp is kept here during parsing. */
Xstatic struct regexp *reg;
X
X/* Find the index of charset s in reg->charsets, or allocate a new charset. */
Xstatic
Xcharset_index(s)
X const _charset s;
X{
X int i;
X
X for (i = 0; i < reg->cindex; ++i)
X if (equal(s, reg->charsets[i]))
X return i;
X REALLOC_IF_NECESSARY(reg->charsets, _charset, reg->calloc, reg->cindex);
X ++reg->cindex;
X copyset(s, reg->charsets[i]);
X return i;
X}
X
X/* Syntax bits controlling the behavior of the lexical analyzer. */
Xstatic syntax_bits, syntax_bits_set;
X
X/* Flag for case-folding letters into sets. */
Xstatic case_fold;
X
X/* Entry point to set syntax options. */
Xvoid
Xregsyntax(bits, fold)
X int bits;
X int fold;
X{
X syntax_bits_set = 1;
X syntax_bits = bits;
X case_fold = fold;
X}
X
X/* Lexical analyzer. */
Xstatic const char *lexstart; /* Pointer to beginning of input string. */
Xstatic const char *lexptr; /* Pointer to next input character. */
Xstatic lexleft; /* Number of characters remaining. */
Xstatic caret_allowed; /* True if backward context allows ^
X (meaningful only if RE_CONTEXT_INDEP_OPS
X is turned off). */
Xstatic closure_allowed; /* True if backward context allows closures
X (meaningful only if RE_CONTEXT_INDEP_OPS
X is turned off). */
X
X/* Note that characters become unsigned here. */
X#define FETCH(c, eoferr) \
X { \
X if (! lexleft) \
X if (eoferr) \
X regerror(eoferr); \
X else \
X return _END; \
X (c) = (unsigned char) *lexptr++; \
X --lexleft; \
X }
X
Xstatic _token
Xlex()
X{
X _token c, c2;
X int invert;
X _charset cset;
X
X FETCH(c, (char *) 0);
X switch (c)
X {
X case '^':
X if (! (syntax_bits & RE_CONTEXT_INDEP_OPS)
X && (!caret_allowed ||
X (syntax_bits & RE_TIGHT_VBAR) && lexptr - 1 != lexstart))
X goto normal_char;
X caret_allowed = 0;
X return syntax_bits & RE_TIGHT_VBAR ? _ALLBEGLINE : _BEGLINE;
X
X case '$':
X if (syntax_bits & RE_CONTEXT_INDEP_OPS || !lexleft
X || (! (syntax_bits & RE_TIGHT_VBAR)
X && ((syntax_bits & RE_NO_BK_PARENS
X ? lexleft > 0 && *lexptr == ')'
X : lexleft > 1 && *lexptr == '\\' && lexptr[1] == ')')
X || (syntax_bits & RE_NO_BK_VBAR
X ? lexleft > 0 && *lexptr == '|'
X : lexleft > 1 && *lexptr == '\\' && lexptr[1] == '|'))))
X return syntax_bits & RE_TIGHT_VBAR ? _ALLENDLINE : _ENDLINE;
X goto normal_char;
X
X case '\\':
X FETCH(c, "Unfinished \\ quote");
X switch (c)
X {
X case '1':
X case '2':
X case '3':
X case '4':
X case '5':
X case '6':
X case '7':
X case '8':
X case '9':
X caret_allowed = 0;
X closure_allowed = 1;
X return _BACKREF;
X
X case '<':
X caret_allowed = 0;
X return _BEGWORD;
X
X case '>':
X caret_allowed = 0;
X return _ENDWORD;
X
X case 'b':
X caret_allowed = 0;
X return _LIMWORD;
X
X case 'B':
X caret_allowed = 0;
X return _NOTLIMWORD;
X
X case 'w':
X case 'W':
X zeroset(cset);
X for (c2 = 0; c2 < _NOTCHAR; ++c2)
X if (ISALNUM(c2))
X setbit(c2, cset);
X if (c == 'W')
X notset(cset);
X caret_allowed = 0;
X closure_allowed = 1;
X return _SET + charset_index(cset);
X
X case '?':
X if (syntax_bits & RE_BK_PLUS_QM)
X goto qmark;
X goto normal_char;
X
X case '+':
X if (syntax_bits & RE_BK_PLUS_QM)
X goto plus;
X goto normal_char;
X
X case '|':
X if (! (syntax_bits & RE_NO_BK_VBAR))
X goto or;
X goto normal_char;
X
X case '(':
X if (! (syntax_bits & RE_NO_BK_PARENS))
X goto lparen;
X goto normal_char;
X
X case ')':
X if (! (syntax_bits & RE_NO_BK_PARENS))
X goto rparen;
X goto normal_char;
X
X default:
X goto normal_char;
X }
X
X case '?':
X if (syntax_bits & RE_BK_PLUS_QM)
X goto normal_char;
X qmark:
X if (! (syntax_bits & RE_CONTEXT_INDEP_OPS) && !closure_allowed)
X goto normal_char;
X return _QMARK;
X
X case '*':
X if (! (syntax_bits & RE_CONTEXT_INDEP_OPS) && !closure_allowed)
X goto normal_char;
X return _STAR;
X
X case '+':
X if (syntax_bits & RE_BK_PLUS_QM)
X goto normal_char;
X plus:
X if (! (syntax_bits & RE_CONTEXT_INDEP_OPS) && !closure_allowed)
X goto normal_char;
X return _PLUS;
X
X case '|':
X if (! (syntax_bits & RE_NO_BK_VBAR))
X goto normal_char;
X or:
X caret_allowed = 1;
X closure_allowed = 0;
X return _OR;
X
X case '\n':
X if (! (syntax_bits & RE_NEWLINE_OR))
X goto normal_char;
X goto or;
X
X case '(':
X if (! (syntax_bits & RE_NO_BK_PARENS))
X goto normal_char;
X lparen:
X caret_allowed = 1;
X closure_allowed = 0;
X return _LPAREN;
X
X case ')':
X if (! (syntax_bits & RE_NO_BK_PARENS))
X goto normal_char;
X rparen:
X caret_allowed = 0;
X closure_allowed = 1;
X return _RPAREN;
X
X case '.':
X zeroset(cset);
X notset(cset);
X clrbit('\n', cset);
X caret_allowed = 0;
X closure_allowed = 1;
X return _SET + charset_index(cset);
X
X case '[':
X zeroset(cset);
X FETCH(c, "Unbalanced [");
X if (c == '^')
X {
X FETCH(c, "Unbalanced [");
X invert = 1;
X }
X else
X invert = 0;
X do
X {
X FETCH(c2, "Unbalanced [");
X if (c2 == '-')
X {
X FETCH(c2, "Unbalanced [");
X while (c <= c2)
X setbit(c++, cset);
X FETCH(c, "Unbalanced [");
X }
X else
X {
X setbit(c, cset);
X c = c2;
X }
X }
X while (c != ']');
X if (invert)
X notset(cset);
X caret_allowed = 0;
X closure_allowed = 1;
X return _SET + charset_index(cset);
X
X default:
X normal_char:
X caret_allowed = 0;
X closure_allowed = 1;
X if (case_fold && ISALPHA(c))
X {
X zeroset(cset);
X if (isupper(c))
X c = tolower(c);
X setbit(c, cset);
X setbit(toupper(c), cset);
X return _SET + charset_index(cset);
X }
X return c;
X }
X}
X
X/* Recursive descent parser for regular expressions. */
X
Xstatic _token tok; /* Lookahead token. */
Xstatic depth; /* Current depth of a hypothetical stack
X holding deferred productions. This is
X used to determine the depth that will be
X required of the real stack later on in
X reganalyze(). */
X
X/* Add the given token to the parse tree, maintaining the depth count and
X updating the maximum depth if necessary. */
Xstatic void
Xaddtok(t)
X _token t;
X{
X REALLOC_IF_NECESSARY(reg->tokens, _token, reg->talloc, reg->tindex);
X reg->tokens[reg->tindex++] = t;
X
X switch (t)
X {
X case _QMARK:
X case _STAR:
X case _PLUS:
X break;
X
X case _CAT:
X case _OR:
X --depth;
X break;
X
X default:
X ++reg->nleaves;
X case _EMPTY:
X ++depth;
X break;
X }
X if (depth > reg->depth)
X reg->depth = depth;
X}
X
X/* The grammar understood by the parser is as follows.
X
X start:
X regexp
X _ALLBEGLINE regexp
X regexp _ALLENDLINE
X _ALLBEGLINE regexp _ALLENDLINE
X
X regexp:
X regexp _OR branch
X branch
X
X branch:
X branch closure
X closure
X
X closure:
X closure _QMARK
X closure _STAR
X closure _PLUS
X atom
X
X atom:
X <normal character>
X _SET
X _BACKREF
X _BEGLINE
X _ENDLINE
X _BEGWORD
X _ENDWORD
X _LIMWORD
X _NOTLIMWORD
X <empty>
X
X The parser builds a parse tree in postfix form in an array of tokens. */
X
X#ifdef __STDC__
Xstatic void regexp(void);
X#else
Xstatic void regexp();
X#endif
X
Xstatic void
Xatom()
X{
X if (tok >= 0 && tok < _NOTCHAR || tok >= _SET || tok == _BACKREF
X || tok == _BEGLINE || tok == _ENDLINE || tok == _BEGWORD
X || tok == _ENDWORD || tok == _LIMWORD || tok == _NOTLIMWORD)
X {
X addtok(tok);
X tok = lex();
X }
X else if (tok == _LPAREN)
X {
X tok = lex();
X regexp();
X if (tok != _RPAREN)
X regerror("Unbalanced (");
X tok = lex();
X }
X else
X addtok(_EMPTY);
X}
X
Xstatic void
Xclosure()
X{
X atom();
X while (tok == _QMARK || tok == _STAR || tok == _PLUS)
X {
X addtok(tok);
X tok = lex();
X }
X}
X
Xstatic void
Xbranch()
X{
X closure();
X while (tok != _RPAREN && tok != _OR && tok != _ALLENDLINE && tok >= 0)
X {
X closure();
X addtok(_CAT);
X }
X}
X
Xstatic void
Xregexp()
X{
X branch();
X while (tok == _OR)
X {
X tok = lex();
X branch();
X addtok(_OR);
X }
X}
X
X/* Main entry point for the parser. S is a string to be parsed, len is the
X length of the string, so s can include NUL characters. R is a pointer to
X the struct regexp to parse into. */
Xvoid
Xregparse(s, len, r)
X const char *s;
X size_t len;
X struct regexp *r;
X{
X reg = r;
X lexstart = lexptr = s;
X lexleft = len;
X caret_allowed = 1;
X closure_allowed = 0;
X
X if (! syntax_bits_set)
X regerror("No syntax specified");
X
X tok = lex();
X depth = r->depth;
X
X if (tok == _ALLBEGLINE)
X {
X addtok(_BEGLINE);
X tok = lex();
X regexp();
X addtok(_CAT);
X }
X else
X regexp();
X
X if (tok == _ALLENDLINE)
X {
X addtok(_ENDLINE);
X addtok(_CAT);
X tok = lex();
X }
X
X if (tok != _END)
X regerror("Unbalanced )");
X
X addtok(_END - r->nregexps);
X addtok(_CAT);
X
X if (r->nregexps)
X addtok(_OR);
X
X ++r->nregexps;
X}
X
X/* Some primitives for operating on sets of positions. */
X
X/* Copy one set to another; the destination must be large enough. */
Xstatic void
Xcopy(src, dst)
X const _position_set *src;
X _position_set *dst;
X{
X int i;
X
X for (i = 0; i < src->nelem; ++i)
X dst->elems[i] = src->elems[i];
X dst->nelem = src->nelem;
X}
X
X/* Insert a position in a set. Position sets are maintained in sorted
X order according to index. If position already exists in the set with
X the same index then their constraints are logically or'd together.
X S->elems must point to an array large enough to hold the resulting set. */
Xstatic void
Xinsert(p, s)
X _position p;
X _position_set *s;
X{
X int i;
X _position t1, t2;
X
X for (i = 0; i < s->nelem && p.index < s->elems[i].index; ++i)
X ;
X if (i < s->nelem && p.index == s->elems[i].index)
X s->elems[i].constraint |= p.constraint;
X else
X {
X t1 = p;
X ++s->nelem;
X while (i < s->nelem)
X {
X t2 = s->elems[i];
X s->elems[i++] = t1;
X t1 = t2;
X }
X }
X}
X
X/* Merge two sets of positions into a third. The result is exactly as if
X the positions of both sets were inserted into an initially empty set. */
Xstatic void
Xmerge(s1, s2, m)
X _position_set *s1;
X _position_set *s2;
X _position_set *m;
X{
X int i = 0, j = 0;
X
X m->nelem = 0;
X while (i < s1->nelem && j < s2->nelem)
X if (s1->elems[i].index > s2->elems[j].index)
X m->elems[m->nelem++] = s1->elems[i++];
X else if (s1->elems[i].index < s2->elems[j].index)
X m->elems[m->nelem++] = s2->elems[j++];
X else
X {
X m->elems[m->nelem] = s1->elems[i++];
X m->elems[m->nelem++].constraint |= s2->elems[j++].constraint;
X }
X while (i < s1->nelem)
X m->elems[m->nelem++] = s1->elems[i++];
X while (j < s2->nelem)
X m->elems[m->nelem++] = s2->elems[j++];
X}
X
X/* Delete a position from a set. */
Xstatic void
Xdelete(p, s)
X _position p;
X _position_set *s;
X{
X int i;
X
X for (i = 0; i < s->nelem; ++i)
X if (p.index == s->elems[i].index)
X break;
X if (i < s->nelem)
X for (--s->nelem; i < s->nelem; ++i)
X s->elems[i] = s->elems[i + 1];
X}
X
X/* Find the index of the state corresponding to the given position set with
X the given preceding context, or create a new state if there is no such
X state. Newline and letter tell whether we got here on a newline or
X letter, respectively. */
Xstatic
Xstate_index(r, s, newline, letter)
X struct regexp *r;
X _position_set *s;
X int newline;
X int letter;
X{
X int hash = 0;
X int constraint;
X int i, j;
X
X newline = newline ? 1 : 0;
X letter = letter ? 1 : 0;
X
X for (i = 0; i < s->nelem; ++i)
X hash ^= s->elems[i].index + s->elems[i].constraint;
X
X /* Try to find a state that exactly matches the proposed one. */
X for (i = 0; i < r->sindex; ++i)
X {
X if (hash != r->states[i].hash || s->nelem != r->states[i].elems.nelem
X || newline != r->states[i].newline || letter != r->states[i].letter)
X continue;
X for (j = 0; j < s->nelem; ++j)
X if (s->elems[j].constraint
X != r->states[i].elems.elems[j].constraint
X || s->elems[j].index != r->states[i].elems.elems[j].index)
X break;
X if (j == s->nelem)
X return i;
X }
X
X /* We'll have to create a new state. */
X REALLOC_IF_NECESSARY(r->states, _dfa_state, r->salloc, r->sindex);
X r->states[i].hash = hash;
X MALLOC(r->states[i].elems.elems, _position, s->nelem);
X copy(s, &r->states[i].elems);
X r->states[i].newline = newline;
X r->states[i].letter = letter;
X r->states[i].backref = 0;
X r->states[i].constraint = 0;
X r->states[i].first_end = 0;
X for (j = 0; j < s->nelem; ++j)
X if (r->tokens[s->elems[j].index] < 0)
X {
X constraint = s->elems[j].constraint;
X if (_SUCCEEDS_IN_CONTEXT(constraint, newline, 0, letter, 0)
X || _SUCCEEDS_IN_CONTEXT(constraint, newline, 0, letter, 1)
X || _SUCCEEDS_IN_CONTEXT(constraint, newline, 1, letter, 0)
X || _SUCCEEDS_IN_CONTEXT(constraint, newline, 1, letter, 1))
X r->states[i].constraint |= constraint;
X if (! r->states[i].first_end)
X r->states[i].first_end = r->tokens[s->elems[j].index];
X }
X else if (r->tokens[s->elems[j].index] == _BACKREF)
X {
X r->states[i].constraint = _NO_CONSTRAINT;
X r->states[i].backref = 1;
X }
X
X ++r->sindex;
X
X return i;
X}
X
X/* Find the epsilon closure of a set of positions. If any position of the set
X contains a symbol that matches the empty string in some context, replace
X that position with the elements of its follow labeled with an appropriate
X constraint. Repeat exhaustively until no funny positions are left.
X S->elems must be large enough to hold the result. */
Xepsclosure(s, r)
X _position_set *s;
X struct regexp *r;
X{
X int i, j;
X int *visited;
X _position p, old;
X
X MALLOC(visited, int, r->tindex);
X for (i = 0; i < r->tindex; ++i)
X visited[i] = 0;
X
X for (i = 0; i < s->nelem; ++i)
X if (r->tokens[s->elems[i].index] >= _NOTCHAR
X && r->tokens[s->elems[i].index] != _BACKREF
X && r->tokens[s->elems[i].index] < _SET)
X {
X old = s->elems[i];
X p.constraint = old.constraint;
X delete(s->elems[i], s);
X if (visited[old.index])
X {
X --i;
X continue;
X }
X visited[old.index] = 1;
X switch (r->tokens[old.index])
X {
X case _BEGLINE:
X p.constraint &= _BEGLINE_CONSTRAINT;
X break;
X case _ENDLINE:
X p.constraint &= _ENDLINE_CONSTRAINT;
X break;
X case _BEGWORD:
X p.constraint &= _BEGWORD_CONSTRAINT;
X break;
X case _ENDWORD:
X p.constraint &= _ENDWORD_CONSTRAINT;
X break;
X case _LIMWORD:
X p.constraint &= _ENDWORD_CONSTRAINT;
X break;
X case _NOTLIMWORD:
X p.constraint &= _NOTLIMWORD_CONSTRAINT;
X break;
X }
X for (j = 0; j < r->follows[old.index].nelem; ++j)
X {
X p.index = r->follows[old.index].elems[j].index;
X insert(p, s);
X }
X /* Force rescan to start at the beginning. */
X i = -1;
X }
X
X free(visited);
X}
X
X/* Perform bottom-up analysis on the parse tree, computing various functions.
X Note that at this point, we're pretending constructs like \< are real
X characters rather than constraints on what can follow them.
X
X Nullable: A node is nullable if it is at the root of a regexp that can
X match the empty string.
X * _EMPTY leaves are nullable.
X * No other leaf is nullable.
X * A _QMARK or _STAR node is nullable.
X * A _PLUS node is nullable if its argument is nullable.
X * A _CAT node is nullable if both its arguments are nullable.
X * An _OR node is nullable if either argument is nullable.
X
X Firstpos: The firstpos of a node is the set of positions (nonempty leaves)
X that could correspond to the first character of a string matching the
X regexp rooted at the given node.
X * _EMPTY leaves have empty firstpos.
X * The firstpos of a nonempty leaf is that leaf itself.
X * The firstpos of a _QMARK, _STAR, or _PLUS node is the firstpos of its
X argument.
X * The firstpos of a _CAT node is the firstpos of the left argument, union
X the firstpos of the right if the left argument is nullable.
X * The firstpos of an _OR node is the union of firstpos of each argument.
X
X Lastpos: The lastpos of a node is the set of positions that could
X correspond to the last character of a string matching the regexp at
X the given node.
X * _EMPTY leaves have empty lastpos.
X * The lastpos of a nonempty leaf is that leaf itself.
X * The lastpos of a _QMARK, _STAR, or _PLUS node is the lastpos of its
X argument.
X * The lastpos of a _CAT node is the lastpos of its right argument, union
X the lastpos of the left if the right argument is nullable.
X * The lastpos of an _OR node is the union of the lastpos of each argument.
X
X Follow: The follow of a position is the set of positions that could
X correspond to the character following a character matching the node in
X a string matching the regexp. At this point we consider special symbols
X that match the empty string in some context to be just normal characters.
X Later, if we find that a special symbol is in a follow set, we will
X replace it with the elements of its follow, labeled with an appropriate
X constraint.
X * Every node in the firstpos of the argument of a _STAR or _PLUS node is in
X the follow of every node in the lastpos.
X * Every node in the firstpos of the second argument of a _CAT node is in
X the follow of every node in the lastpos of the first argument.
X
X Because of the postfix representation of the parse tree, the depth-first
X analysis is conveniently done by a linear scan with the aid of a stack.
X Sets are stored as arrays of the elements, obeying a stack-like allocation
X scheme; the number of elements in each set deeper in the stack can be
X used to determine the address of a particular set's array. */
Xvoid
Xreganalyze(r, searchflag)
X struct regexp *r;
X int searchflag;
X{
X int *nullable; /* Nullable stack. */
X int *nfirstpos; /* Element count stack for firstpos sets. */
X _position *firstpos; /* Array where firstpos elements are stored. */
X int *nlastpos; /* Element count stack for lastpos sets. */
X _position *lastpos; /* Array where lastpos elements are stored. */
X int *nalloc; /* Sizes of arrays allocated to follow sets. */
X _position_set tmp; /* Temporary set for merging sets. */
X _position_set merged; /* Result of merging sets. */
X int wants_newline; /* True if some position wants newline info. */
X int *o_nullable;
X int *o_nfirst, *o_nlast;
X _position *o_firstpos, *o_lastpos;
X int i, j;
X _position *pos;
X
X r->searchflag = searchflag;
X
X MALLOC(nullable, int, r->depth);
X o_nullable = nullable;
X MALLOC(nfirstpos, int, r->depth);
X o_nfirst = nfirstpos;
X MALLOC(firstpos, _position, r->nleaves);
X o_firstpos = firstpos, firstpos += r->nleaves;
X MALLOC(nlastpos, int, r->depth);
X o_nlast = nlastpos;
X MALLOC(lastpos, _position, r->nleaves);
X o_lastpos = lastpos, lastpos += r->nleaves;
X MALLOC(nalloc, int, r->tindex);
X for (i = 0; i < r->tindex; ++i)
X nalloc[i] = 0;
X MALLOC(merged.elems, _position, r->nleaves);
X
X CALLOC(r->follows, _position_set, r->tindex);
X
X for (i = 0; i < r->tindex; ++i)
X switch (r->tokens[i])
X {
X case _EMPTY:
X /* The empty set is nullable. */
X *nullable++ = 1;
X
X /* The firstpos and lastpos of the empty leaf are both empty. */
X *nfirstpos++ = *nlastpos++ = 0;
X break;
X
X case _STAR:
X case _PLUS:
X /* Every element in the firstpos of the argument is in the follow
X of every element in the lastpos. */
X tmp.nelem = nfirstpos[-1];
X tmp.elems = firstpos;
X pos = lastpos;
X for (j = 0; j < nlastpos[-1]; ++j)
X {
X merge(&tmp, &r->follows[pos[j].index], &merged);
X REALLOC_IF_NECESSARY(r->follows[pos[j].index].elems, _position,
X nalloc[pos[j].index], merged.nelem - 1);
X copy(&merged, &r->follows[pos[j].index]);
X }
X
X case _QMARK:
X /* A _QMARK or _STAR node is automatically nullable. */
X if (r->tokens[i] != _PLUS)
X nullable[-1] = 1;
X break;
X
X case _CAT:
X /* Every element in the firstpos of the second argument is in the
X follow of every element in the lastpos of the first argument. */
X tmp.nelem = nfirstpos[-1];
X tmp.elems = firstpos;
X pos = lastpos + nlastpos[-1];
X for (j = 0; j < nlastpos[-2]; ++j)
X {
X merge(&tmp, &r->follows[pos[j].index], &merged);
X REALLOC_IF_NECESSARY(r->follows[pos[j].index].elems, _position,
X nalloc[pos[j].index], merged.nelem - 1);
X copy(&merged, &r->follows[pos[j].index]);
X }
X
X /* The firstpos of a _CAT node is the firstpos of the first argument,
X union that of the second argument if the first is nullable. */
X if (nullable[-2])
X nfirstpos[-2] += nfirstpos[-1];
X else
X firstpos += nfirstpos[-1];
X --nfirstpos;
X
X /* The lastpos of a _CAT node is the lastpos of the second argument,
X union that of the first argument if the second is nullable. */
X if (nullable[-1])
X nlastpos[-2] += nlastpos[-1];
X else
X {
X pos = lastpos + nlastpos[-2];
X for (j = nlastpos[-1] - 1; j >= 0; --j)
X pos[j] = lastpos[j];
X lastpos += nlastpos[-2];
X nlastpos[-2] = nlastpos[-1];
X }
X --nlastpos;
X
X /* A _CAT node is nullable if both arguments are nullable. */
X nullable[-2] = nullable[-1] && nullable[-2];
X --nullable;
X break;
X
X case _OR:
X /* The firstpos is the union of the firstpos of each argument. */
X nfirstpos[-2] += nfirstpos[-1];
X --nfirstpos;
X
X /* The lastpos is the union of the lastpos of each argument. */
X nlastpos[-2] += nlastpos[-1];
X --nlastpos;
X
X /* An _OR node is nullable if either argument is nullable. */
X nullable[-2] = nullable[-1] || nullable[-2];
X --nullable;
X break;
X
X default:
X /* Anything else is a nonempty position. (Note that special
X constructs like \< are treated as nonempty strings here;
X an "epsilon closure" effectively makes them nullable later.
X Backreferences have to get a real position so we can detect
X transitions on them later. But they are nullable. */
X *nullable++ = r->tokens[i] == _BACKREF;
X
X /* This position is in its own firstpos and lastpos. */
X *nfirstpos++ = *nlastpos++ = 1;
X --firstpos, --lastpos;
X firstpos->index = lastpos->index = i;
X firstpos->constraint = lastpos->constraint = _NO_CONSTRAINT;
X
X /* Allocate the follow set for this position. */
X nalloc[i] = 1;
X MALLOC(r->follows[i].elems, _position, nalloc[i]);
X break;
X }
X
X /* For each follow set that is the follow set of a real position, replace
X it with its epsilon closure. */
X for (i = 0; i < r->tindex; ++i)
X if (r->tokens[i] < _NOTCHAR || r->tokens[i] == _BACKREF
X || r->tokens[i] >= _SET)
X {
X copy(&r->follows[i], &merged);
X epsclosure(&merged, r);
X REALLOC(r->follows[i].elems, _position, merged.nelem);
X copy(&merged, &r->follows[i]);
X }
X
X /* Get the epsilon closure of the firstpos of the regexp. The result will
X be the set of positions of state 0. */
X merged.nelem = 0;
X for (i = 0; i < nfirstpos[-1]; ++i)
X insert(firstpos[i], &merged);
X epsclosure(&merged, r);
X
X /* Check if any of the positions of state 0 will want newline context. */
X wants_newline = 0;
X for (i = 0; i < merged.nelem; ++i)
X if (_PREV_NEWLINE_DEPENDENT(merged.elems[i].constraint))
X wants_newline = 1;
X
X /* Build the initial state. */
X r->salloc = 1;
X r->sindex = 0;
X MALLOC(r->states, _dfa_state, r->salloc);
X state_index(r, &merged, wants_newline, 0);
X
X free(o_nullable);
X free(o_nfirst);
X free(o_firstpos);
X free(o_nlast);
X free(o_lastpos);
X free(nalloc);
X free(merged.elems);
X}
X
X/* Find, for each character, the transition out of state s of r, and store
X it in the appropriate slot of trans.
X
X We divide the positions of s into groups (positions can appear in more
X than one group). Each group is labeled with a set of characters that
X every position in the group matches (taking into account, if necessary,
X preceding context information of s). For each group, find the union
X of the its elements' follows. This set is the set of positions of the
X new state. For each character in the group's label, set the transition
X on this character to be to a state corresponding to the set's positions,
X and its associated backward context information, if necessary.
X
X If we are building a searching matcher, we include the positions of state
X 0 in every state.
X
X The collection of groups is constructed by building an equivalence-class
X partition of the positions of s.
X
X For each position, find the set of characters C that it matches. Eliminate
X any characters from C that fail on grounds of backward context.
X
X Search through the groups, looking for a group whose label L has nonempty
X intersection with C. If L - C is nonempty, create a new group labeled
X L - C and having the same positions as the current group, and set L to
X the intersection of L and C. Insert the position in this group, set
X C = C - L, and resume scanning.
X
X If after comparing with every group there are characters remaining in C,
X create a new group labeled with the characters of C and insert this
X position in that group. */
Xvoid
Xregstate(s, r, trans)
X int s;
X struct regexp *r;
X int trans[];
X{
X _position_set grps[_NOTCHAR]; /* As many as will ever be needed. */
X _charset labels[_NOTCHAR]; /* Labels corresponding to the groups. */
X int ngrps = 0; /* Number of groups actually used. */
X _position pos; /* Current position being considered. */
X _charset matches; /* Set of matching characters. */
X int matchesf; /* True if matches is nonempty. */
X _charset intersect; /* Intersection with some label set. */
X int intersectf; /* True if intersect is nonempty. */
X _charset leftovers; /* Stuff in the label that didn't match. */
X int leftoversf; /* True if leftovers is nonempty. */
X static _charset letters; /* Set of characters considered letters. */
X static _charset newline; /* Set of characters that aren't newline. */
X _position_set follows; /* Union of the follows of some group. */
X _position_set tmp; /* Temporary space for merging sets. */
X int state; /* New state. */
X int wants_newline; /* New state wants to know newline context. */
X int state_newline; /* New state on a newline transition. */
X int wants_letter; /* New state wants to know letter context. */
X int state_letter; /* New state on a letter transition. */
X static initialized; /* Flag for static initialization. */
X int i, j, k;
X
X /* Initialize the set of letters, if necessary. */
X if (! initialized)
X {
X initialized = 1;
X for (i = 0; i < _NOTCHAR; ++i)
X if (ISALNUM(i))
X setbit(i, letters);
X setbit('\n', newline);
X }
X
X zeroset(matches);
X
X for (i = 0; i < r->states[s].elems.nelem; ++i)
X {
X pos = r->states[s].elems.elems[i];
X if (r->tokens[pos.index] >= 0 && r->tokens[pos.index] < _NOTCHAR)
X setbit(r->tokens[pos.index], matches);
X else if (r->tokens[pos.index] >= _SET)
X copyset(r->charsets[r->tokens[pos.index] - _SET], matches);
X else
X continue;
X
X /* Some characters may need to be climinated from matches because
X they fail in the current context. */
X if (pos.constraint != 0xff)
X {
X if (! _MATCHES_NEWLINE_CONTEXT(pos.constraint,
X r->states[s].newline, 1))
X clrbit('\n', matches);
X if (! _MATCHES_NEWLINE_CONTEXT(pos.constraint,
X r->states[s].newline, 0))
X for (j = 0; j < _CHARSET_INTS; ++j)
X matches[j] &= newline[j];
X if (! _MATCHES_LETTER_CONTEXT(pos.constraint,
X r->states[s].letter, 1))
X for (j = 0; j < _CHARSET_INTS; ++j)
X matches[j] &= ~letters[j];
X if (! _MATCHES_LETTER_CONTEXT(pos.constraint,
X r->states[s].letter, 0))
X for (j = 0; j < _CHARSET_INTS; ++j)
X matches[j] &= letters[j];
X
X /* If there are no characters left, there's no point in going on. */
X for (j = 0; j < _CHARSET_INTS && !matches[j]; ++j)
X ;
X if (j == _CHARSET_INTS)
X continue;
X }
X
X for (j = 0; j < ngrps; ++j)
X {
X /* If matches contains a single character only, and the current
X group's label doesn't contain that character, go on to the
X next group. */
X if (r->tokens[pos.index] >= 0 && r->tokens[pos.index] < _NOTCHAR
X && !tstbit(r->tokens[pos.index], labels[j]))
X continue;
X
X /* Check if this group's label has a nonempty intersection with
X matches. */
X intersectf = 0;
X for (k = 0; k < _CHARSET_INTS; ++k)
X (intersect[k] = matches[k] & labels[j][k]) ? intersectf = 1 : 0;
X if (! intersectf)
X continue;
X
X /* It does; now find the set differences both ways. */
X leftoversf = matchesf = 0;
X for (k = 0; k < _CHARSET_INTS; ++k)
X {
X /* Even an optimizing compiler can't know this for sure. */
X int match = matches[k], label = labels[j][k];
X
X (leftovers[k] = ~match & label) ? leftoversf = 1 : 0;
X (matches[k] = match & ~label) ? matchesf = 1 : 0;
X }
X
X /* If there were leftovers, create a new group labeled with them. */
X if (leftoversf)
X {
X copyset(leftovers, labels[ngrps]);
X copyset(intersect, labels[j]);
X MALLOC(grps[ngrps].elems, _position, r->nleaves);
X copy(&grps[j], &grps[ngrps]);
X ++ngrps;
X }
X
X /* Put the position in the current group. Note that there is no
X reason to call insert() here. */
X grps[j].elems[grps[j].nelem++] = pos;
X
X /* If every character matching the current position has been
X accounted for, we're done. */
X if (! matchesf)
X break;
X }
X
X /* If we've passed the last group, and there are still characters
X unaccounted for, then we'll have to create a new group. */
X if (j == ngrps)
X {
X copyset(matches, labels[ngrps]);
X zeroset(matches);
X MALLOC(grps[ngrps].elems, _position, r->nleaves);
X grps[ngrps].nelem = 1;
X grps[ngrps].elems[0] = pos;
X ++ngrps;
X }
X }
X
X MALLOC(follows.elems, _position, r->nleaves);
X MALLOC(tmp.elems, _position, r->nleaves);
X
X /* If we are a searching matcher, the default transition is to a state
X containing the positions of state 0, otherwise the default transition
X is to fail miserably. */
X if (r->searchflag)
X {
X wants_newline = 0;
X wants_letter = 0;
X for (i = 0; i < r->states[0].elems.nelem; ++i)
X {
X if (_PREV_NEWLINE_DEPENDENT(r->states[0].elems.elems[i].constraint))
X wants_newline = 1;
X if (_PREV_LETTER_DEPENDENT(r->states[0].elems.elems[i].constraint))
X wants_letter = 1;
X }
X copy(&r->states[0].elems, &follows);
X state = state_index(r, &follows, 0, 0);
X if (wants_newline)
X state_newline = state_index(r, &follows, 1, 0);
X else
X state_newline = state;
X if (wants_letter)
X state_letter = state_index(r, &follows, 0, 1);
X else
X state_letter = state;
X for (i = 0; i < _NOTCHAR; ++i)
X if (i == '\n')
X trans[i] = state_newline;
X else if (ISALNUM(i))
X trans[i] = state_letter;
X else
X trans[i] = state;
X }
X else
X for (i = 0; i < _NOTCHAR; ++i)
X trans[i] = -1;
X
X for (i = 0; i < ngrps; ++i)
X {
X follows.nelem = 0;
X
X /* Find the union of the follows of the positions of the group.
X This is a hideously inefficient loop. Fix it someday. */
X for (j = 0; j < grps[i].nelem; ++j)
X for (k = 0; k < r->follows[grps[i].elems[j].index].nelem; ++k)
X insert(r->follows[grps[i].elems[j].index].elems[k], &follows);
X
X /* If we are building a searching matcher, throw in the positions
X of state 0 as well. */
X if (r->searchflag)
X for (j = 0; j < r->states[0].elems.nelem; ++j)
X insert(r->states[0].elems.elems[j], &follows);
X
X /* Find out if the new state will want any context information. */
X wants_newline = 0;
X if (tstbit('\n', labels[i]))
X for (j = 0; j < follows.nelem; ++j)
X if (_PREV_NEWLINE_DEPENDENT(follows.elems[j].constraint))
X wants_newline = 1;
X
X wants_letter = 0;
X for (j = 0; j < _CHARSET_INTS; ++j)
X if (labels[i][j] & letters[j])
X break;
X if (j < _CHARSET_INTS)
X for (j = 0; j < follows.nelem; ++j)
X if (_PREV_LETTER_DEPENDENT(follows.elems[j].constraint))
X wants_letter = 1;
X
X /* Find the state(s) corresponding to the union of the follows. */
X state = state_index(r, &follows, 0, 0);
X if (wants_newline)
X state_newline = state_index(r, &follows, 1, 0);
X else
X state_newline = state;
X if (wants_letter)
X state_letter = state_index(r, &follows, 0, 1);
X else
X state_letter = state;
X
X /* Set the transitions for each character in the current label. */
X for (j = 0; j < _CHARSET_INTS; ++j)
X for (k = 0; k < INTBITS; ++k)
X if (labels[i][j] & 1 << k)
X {
X int c = j * INTBITS + k;
X
X if (c == '\n')
X trans[c] = state_newline;
X else if (ISALNUM(c))
X trans[c] = state_letter;
X else if (c < _NOTCHAR)
X trans[c] = state;
X }
X }
X
X for (i = 0; i < ngrps; ++i)
X free(grps[i].elems);
X free(follows.elems);
X free(tmp.elems);
X}
X
X/* Some routines for manipulating a compiled regexp's transition tables.
X Each state may or may not have a transition table; if it does, and it
X is a non-accepting state, then r->trans[state] points to its table.
X If it is an accepting state then r->fails[state] points to its table.
X If it has no table at all, then r->trans[state] is NULL.
X TODO: Improve this comment, get rid of the unnecessary redundancy. */
X
Xstatic void
Xbuild_state(s, r)
X int s;
X struct regexp *r;
X{
X int *trans; /* The new transition table. */
X int i;
X
X /* Set an upper limit on the number of transition tables that will ever
X exist at once. 1024 is arbitrary. The idea is that the frequently
X used transition tables will be quickly rebuilt, whereas the ones that
X were only needed once or twice will be cleared away. */
X if (r->trcount >= 1024)
X {
X for (i = 0; i < r->tralloc; ++i)
X if (r->trans[i])
X {
X free((ptr_t) r->trans[i]);
X r->trans[i] = NULL;
X }
X else if (r->fails[i])
X {
X free((ptr_t) r->fails[i]);
X r->fails[i] = NULL;
X }
X r->trcount = 0;
X }
X
X ++r->trcount;
X
X /* Set up the success bits for this state. */
X r->success[s] = 0;
X if (ACCEPTS_IN_CONTEXT(r->states[s].newline, 1, r->states[s].letter, 0,
X s, *r))
X r->success[s] |= 4;
X if (ACCEPTS_IN_CONTEXT(r->states[s].newline, 0, r->states[s].letter, 1,
X s, *r))
X r->success[s] |= 2;
X if (ACCEPTS_IN_CONTEXT(r->states[s].newline, 0, r->states[s].letter, 0,
X s, *r))
X r->success[s] |= 1;
X
X MALLOC(trans, int, _NOTCHAR);
X regstate(s, r, trans);
X
X /* Now go through the new transition table, and make sure that the trans
X and fail arrays are allocated large enough to hold a pointer for the
X largest state mentioned in the table. */
X for (i = 0; i < _NOTCHAR; ++i)
X if (trans[i] >= r->tralloc)
X {
X int oldalloc = r->tralloc;
X
X while (trans[i] >= r->tralloc)
X r->tralloc *= 2;
X REALLOC(r->realtrans, int *, r->tralloc + 1);
X r->trans = r->realtrans + 1;
X REALLOC(r->fails, int *, r->tralloc);
X REALLOC(r->success, int, r->tralloc);
X REALLOC(r->newlines, int, r->tralloc);
X while (oldalloc < r->tralloc)
X {
X r->trans[oldalloc] = NULL;
X r->fails[oldalloc++] = NULL;
X }
X }
X
X /* Keep the newline transition in a special place so we can use it as
X a sentinel. */
X r->newlines[s] = trans['\n'];
X trans['\n'] = -1;
X
X if (ACCEPTING(s, *r))
X r->fails[s] = trans;
X else
X r->trans[s] = trans;
X}
X
Xstatic void
Xbuild_state_zero(r)
X struct regexp *r;
X{
X r->tralloc = 1;
X r->trcount = 0;
X CALLOC(r->realtrans, int *, r->tralloc + 1);
X r->trans = r->realtrans + 1;
X CALLOC(r->fails, int *, r->tralloc);
X MALLOC(r->success, int, r->tralloc);
X MALLOC(r->newlines, int, r->tralloc);
X build_state(0, r);
X}
X
X/* Search through a buffer looking for a match to the given struct regexp.
X Find the first occurrence of a string matching the regexp in the buffer,
X and the shortest possible version thereof. Return a pointer to the first
X character after the match, or NULL if none is found. Begin points to
X the beginning of the buffer, and end points to the first character after
X its end. We store a newline in *end to act as a sentinel, so end had
X better point somewhere valid. Newline is a flag indicating whether to
X allow newlines to be in the matching string. If count is non-
X NULL it points to a place we're supposed to increment every time we
X see a newline. Finally, if backref is non-NULL it points to a place
X where we're supposed to store a 1 if backreferencing happened and the
X match needs to be verified by a backtracking matcher. Otherwise
X we store a 0 in *backref. */
Xchar *
Xregexecute(r, begin, end, newline, count, backref)
X struct regexp *r;
X char *begin;
X char *end;
X int newline;
X int *count;
X int *backref;
X{
X register s, s1, tmp; /* Current state. */
X register unsigned char *p; /* Current input character. */
X register **trans, *t; /* Copy of r->trans so it can be optimized
X into a register. */
X static sbit[_NOTCHAR]; /* Table for anding with r->success. */
X static sbit_init;
X
X if (! sbit_init)
X {
X int i;
X
X sbit_init = 1;
X for (i = 0; i < _NOTCHAR; ++i)
X if (i == '\n')
X sbit[i] = 4;
X else if (ISALNUM(i))
X sbit[i] = 2;
X else
X sbit[i] = 1;
X }
X
X if (! r->tralloc)
X build_state_zero(r);
X
X s = 0;
X p = (unsigned char *) begin;
X trans = r->trans;
X *end = '\n';
X
X for (;;)
X {
X /* The dreaded inner loop. */
X if (t = trans[s])
X do
X {
X s1 = t[*p++];
X if (! (t = trans[s1]))
X goto last_was_s;
X s = t[*p++];
X }
X while (t = trans[s]);
X goto last_was_s1;
X last_was_s:
X tmp = s, s = s1, s1 = tmp;
X last_was_s1:
X
X if (s >= 0 && p <= (unsigned char *) end && r->fails[s])
X {
X if (r->success[s] & sbit[*p])
X {
X if (backref)
X if (r->states[s].backref)
X *backref = 1;
X else
X *backref = 0;
X return (char *) p;
X }
X
X s1 = s;
X s = r->fails[s][*p++];
X continue;
X }
X
X /* If the previous character was a newline, count it. */
X if (count && (char *) p <= end && p[-1] == '\n')
X ++*count;
X
X /* Check if we've run off the end of the buffer. */
X if ((char *) p >= end)
X return NULL;
X
X if (s >= 0)
X {
X build_state(s, r);
X trans = r->trans;
X continue;
X }
X
X if (p[-1] == '\n' && newline)
X {
X s = r->newlines[s1];
X continue;
X }
X
X s = 0;
X }
X}
X
X/* Initialize the components of a regexp that the other routines don't
X initialize for themselves. */
Xvoid
Xreginit(r)
X struct regexp *r;
X{
X r->calloc = 1;
X MALLOC(r->charsets, _charset, r->calloc);
X r->cindex = 0;
X
X r->talloc = 1;
X MALLOC(r->tokens, _token, r->talloc);
X r->tindex = r->depth = r->nleaves = r->nregexps = 0;
X
X r->searchflag = 0;
X r->tralloc = 0;
X}
X
X/* Parse and analyze a single string of the given length. */
Xvoid
Xregcompile(s, len, r, searchflag)
X const char *s;
X size_t len;
X struct regexp *r;
X int searchflag;
X{
X if (case_fold) /* dummy folding in service of regmust() */
X {
X static char *p;
X
X case_fold = 0;
X for (p = (char *)s; *p != 0; p++)
X if (isupper((int)*p))
X *p = tolower((int) *p);
X reginit(r);
X r->mustn = 0;
X r->must[0] = '\0';
X regparse(s, len, r);
X regmust(r);
X reganalyze(r, searchflag);
X case_fold = 1;
X reginit(r);
X regparse(s, len, r);
X reganalyze(r, searchflag);
X }
X else
X {
X reginit(r);
X regparse(s, len, r);
X regmust(r);
X reganalyze(r, searchflag);
X }
X}
X
X/* Free the storage held by the components of a regexp. */
Xvoid
Xregfree(r)
X struct regexp *r;
X{
X int i;
X
X free((ptr_t) r->charsets);
X free((ptr_t) r->tokens);
X for (i = 0; i < r->sindex; ++i)
X free((ptr_t) r->states[i].elems.elems);
X free((ptr_t) r->states);
X for (i = 0; i < r->tindex; ++i)
X if (r->follows[i].elems)
X free((ptr_t) r->follows[i].elems);
X free((ptr_t) r->follows);
X for (i = 0; i < r->tralloc; ++i)
X if (r->trans[i])
X free((ptr_t) r->trans[i]);
X else if (r->fails[i])
X free((ptr_t) r->fails[i]);
X free((ptr_t) r->realtrans);
X free((ptr_t) r->fails);
X free((ptr_t) r->newlines);
X}
X
X/*
XHaving found the postfix representation of the regular expression,
Xtry to find a long sequence of characters that must appear in any line
Xcontaining the r.e.
XFinding a "longest" sequence is beyond the scope of this bagatelle;
Xwe take the easy way out and hope for the best.
X
XWe do a bottom-up calculation of several (possibly zero-length) sequences
Xof characters that must appear in matches of r.e.'s represented by trees
Xrooted at the nodes of the postfix representation:
X sequences that must appear at the left of the match ("left")
X sequences that must appear at the right of the match ("right")
X sequences that must appear somewhere in the match ("in")
X sequences that must constitute the match ("is")
XWhen we get to the root of the tree, we use its calculated "in" sequence
Xas our answer. The sequence we find is returned in r->must (where "r" is
Xthe single argument passed to "regmust"); the length of the sequence is
Xreturned in r->mustn.
X
XThe sequences calculated for the various types of node (in pseudo ANSI c)
Xare shown below. "p" is the operand of unary operators (and the left-hand
Xoperand of binary operators); "q" is the right-hand operand of binary operators.
X"ZERO" means "a zero-length sequence" below.
X
XType left right is in
X---- ---- ----- -- --
Xchar c # c # c # c # c
X
XSET ZERO ZERO ZERO ZERO
X
XSTAR ZERO ZERO ZERO ZERO
X
XQMARK ZERO ZERO ZERO ZERO
X
XPLUS p->left p->right ZERO ZERO
X
XCAT (p->is==ZERO)? (q->is==ZERO)? (p->is!=ZERO && longest of
X p->left : q->right : q->is!=ZERO) ? p->in, q->in, or
X p->is##q->left p->right##q->is p->is##q->is : p->right##q->left
X ZERO
X
XOR longest common longest common (do p->is and (do p->in and
X leading trailing q->is have same q->in have same
X (sub)sequence (sub)sequence length and length and
X of p->left of p->right content) ? content) ?
X and q->left and q->right p->is : NULL p->in : NULL
X
XIf there's anything else we recognize in the tree, all four sequences get set
Xto zero-length sequences. If there's something we don't recognize in the tree,
Xwe just return a zero-length sequence.
X
XAfter the above calculations are performed, three additional steps are taken:
X
X1. If there's a non-zero-length "is" sequence, it replaces the
X "left", "right", and "in" sequences.
X2. If the "left" sequence is longer than the "in" sequence, it replaces
X the "in" sequence.
X3. If the "right" sequence is longer than the "in" sequence, it replaces
X the "in" sequence.
X
XPossibilities:
X1. Find the longest common (sub)sequence of p->in and q->in when doing
X an OR node's "in" sequence? Possibly effective, as in
X egrep 'pepsi|epsilon'
X but is it cheap and easy enough?
X2. In replacing "in" sequences with "left" and "right" sequences, how
X should ties be broken?
X3. Switch to allocated memory, rather than relying on a defined MUST_MAX?
X*/
X
X#define TRUE 1
X#define FALSE 0
X
Xtypedef struct {
X int n;
X char p[MUST_MAX];
X} counted;
X
X#define initcounted(cp) ((cp)->n = 0)
X
Xstatic void
Xcntcpy(top, fromp)
Xcounted * top;
Xcounted * fromp;
X{
X register char * fp;
X register char * tp;
X register int n;
X
X fp = fromp->p;
X tp = top->p;
X n = fromp->n;
X top->n = n;
X while (n-- > 0)
X *tp++ = *fp++;
X}
X
Xstatic void
Xcntcat(top, fromp)
Xcounted * top;
Xcounted * fromp;
X{
X register char * fp;
X register char * tp;
X register int n;
X
X fp = fromp->p;
X tp = top->p + top->n;
X n = fromp->n;
X top->n += n;
X while (n-- > 0)
X *tp++ = *fp++;
X}
X
Xstatic int
Xcntsame(acp, bcp)
Xcounted * acp;
Xcounted * bcp;
X{
X register int i;
X
X if (acp->n != bcp->n)
X return FALSE;
X for (i = 0; i < acp->n; ++i)
X if (acp->p[i] != bcp->p[i])
X return FALSE;
X return TRUE;
X}
X
X
Xtypedef struct {
X counted left;
X counted right;
X counted in;
X counted is;
X} must;
X
Xstatic void
Xinitmust(mp)
Xmust * mp;
X{
X initcounted(&mp->left);
X initcounted(&mp->right);
X initcounted(&mp->in);
X initcounted(&mp->is);
X}
X
Xstatic void
Xregmust(r)
Xregister struct regexp * r;
X{
X must musts[MUST_MAX];
X register must * mp;
X counted result;
X register int ri;
X register int i;
X register _token t;
X
X reg->mustn = 0;
X reg->must[0] = '\0';
X if (reg->tindex > MUST_MAX)
X return;
X mp = musts;
X initcounted(&result);
X for (ri = 0; ri < reg->tindex; ++ri) {
X switch (t = reg->tokens[ri]) {
X case _ALLBEGLINE:
X case _ALLENDLINE:
X case _LPAREN:
X case _RPAREN:
X goto done; /* "cannot happen" */
X case _EMPTY:
X case _BEGLINE:
X case _ENDLINE:
X case _BEGWORD:
X case _ENDWORD:
X case _LIMWORD:
X case _NOTLIMWORD:
X case _BACKREF:
X initmust(mp);
X break;
X case _STAR:
X case _QMARK:
X if (mp <= musts)
X goto done; /* "cannot happen" */
X --mp;
X initmust(mp);
X break;
X case _OR:
X if (mp < &musts[2])
X goto done; /* "cannot happen" */
X {
X register must * lmp;
X register must * rmp;
X register int j, n;
X
X rmp = --mp;
X lmp = --mp;
X /* Guaranteed to be. Unlikely, but. . . */
X if (!cntsame(&lmp->is, &rmp->is))
X initcounted(&lmp->is);
X /* Left side--easy */
X n = lmp->left.n;
X if (n > rmp->left.n)
X n = rmp->left.n;
X for (i = 0; i < n; ++i)
X if (lmp->left.p[i] != rmp->left.p[i])
X break;
X lmp->left.n = i;
X /* Right side */
X n = lmp->right.n;
X if (n > rmp->right.n)
X n = rmp->right.n;
X for (i = 0; i < n; ++i)
X if (lmp->right.p[lmp->right.n-i-1] !=
X rmp->right.p[rmp->right.n-i-1])
X break;
X for (j = 0; j < i; ++j)
X lmp->right.p[j] =
X lmp->right.p[(lmp->right.n -
X i) + j];
X lmp->right.n = i;
X /* Includes. Unlikely, but. . . */
X if (!cntsame(&lmp->in, &rmp->in))
X initcounted(&lmp->in);
X }
X break;
X case _PLUS:
X if (mp <= musts)
X goto done; /* "cannot happen" */
X --mp;
X initcounted(&mp->is);
X break;
X case _END:
X if (mp != &musts[1])
X goto done; /* "cannot happen" */
X result = musts[0].in;
X goto done;
X case _CAT:
X if (mp < &musts[2])
X goto done; /* "cannot happen" */
X {
X must * lmp;
X must * rmp;
X must new;
X must * nmp;
X int a, b, c;
X
X rmp = --mp;
X lmp = --mp;
X nmp = &new;
X initmust(nmp);
X /* Left-hand */
X cntcat(&nmp->left, &lmp->left);
X if (lmp->is.n != 0)
X cntcat(&nmp->left, &rmp->left);
X /* Right-hand */
X if (rmp->is.n != 0)
X cntcat(&nmp->right, &lmp->right);
X cntcat(&nmp->right, &rmp->right);
X /* Guaranteed to be */
X if (lmp->is.n != 0 && rmp->is.n != 0) {
X cntcat(&nmp->is, &lmp->is);
X cntcat(&nmp->is, &rmp->is);
X }
X /* Interior */
X a = lmp->in.n;
X b = rmp->in.n;
X c = lmp->right.n + rmp->left.n;
X if (a == 0 && b == 0 && c == 0) {
X /* nothing */
X ;
X } else if (c > a && c > b) {
X cntcat(&nmp->in, &lmp->right);
X cntcat(&nmp->in, &rmp->left);
X } else if (a > b) {
X cntcat(&nmp->in, &lmp->in);
X } else {
X cntcat(&nmp->in, &rmp->in);
X }
X *mp = new;
X }
X break;
X default:
X if (t < _END) {
X /* "cannot happen" */
X goto done;
X } else if (t >= _SET) {
X /* easy enough */
X initmust(mp);
X } else {
X /* plain character */
X mp->left.p[0] = mp->right.p[0] =
X mp->in.p[0] = mp->is.p[0] = t;
X mp->left.n = mp->right.n =
X mp->in.n = mp->is.n = 1;
X break;
X }
X break;
X }
X /*
X ** "Additional steps"
X */
X if (mp->is.n > 0) {
X cntcpy(&mp->left, &mp->is);
X cntcpy(&mp->right, &mp->is);
X cntcpy(&mp->in, &mp->is);
X }
X if (mp->left.n > mp->in.n)
X cntcpy(&mp->in, &mp->left);
X if (mp->right.n > mp->in.n)
X cntcpy(&mp->in, &mp->right);
X ++mp;
X }
Xdone:
X reg->mustn = result.n;
X for (i = 0; i < result.n; ++i)
X reg->must[i] = result.p[i];
X}
END_OF_FILE
if test 54865 -ne `wc -c <'dfa.c'`; then
echo shar: \"'dfa.c'\" unpacked with wrong size!
fi
# end of 'dfa.c'
fi
echo shar: End of shell archive.
exit 0
--
Please send comp.sources.unix-related mail to rsalz@uunet.uu.net.