Fri Apr 2 17:31:59 1993 Jim Blandy (jimb@totoro.cs.oberlin.edu) * Released version 0.12. * regex.c (regerror): If errcode is zero, that's not a valid error code, according to POSIX, but return "Success." * regex.c (regerror): Remember to actually fetch the message from re_error_msg. * regex.c (regex_compile): Don't use the trick for ".*\n" on ".+\n". Since the latter involves laying an extra choice point, the backward jump isn't adjusted properly. Thu Mar 25 21:35:18 1993 Jim Blandy (jimb@totoro.cs.oberlin.edu) * regex.c (regex_compile): In the handle_open and handle_close sections, clear pending_exact to zero. Tue Mar 9 12:03:07 1993 Jim Blandy (jimb@wookumz.gnu.ai.mit.edu) * regex.c (re_search_2): In the loop which searches forward using fastmap, don't forget to cast the character from the string to an unsigned before using it as an index into the translate map. Thu Jan 14 15:41:46 1993 David J. MacKenzie (djm@kropotkin.gnu.ai.mit.edu) * regex.h: Never define const; let the callers do it. configure.in: Don't define USING_AUTOCONF. Wed Jan 6 20:49:29 1993 Jim Blandy (jimb@geech.gnu.ai.mit.edu) * regex.c (regerror): Abort if ERRCODE is out of range. Sun Dec 20 16:19:10 1992 Jim Blandy (jimb@totoro.cs.oberlin.edu) * configure.in: Arrange to #define USING_AUTOCONF. * regex.h: If USING_AUTOCONF is #defined, don't mess with `const' at all; autoconf has taken care of it. Mon Dec 14 21:40:39 1992 David J. MacKenzie (djm@kropotkin.gnu.ai.mit.edu) * regex.h (RE_SYNTAX_AWK): Fix typo. From Arnold Robbins. Sun Dec 13 20:35:39 1992 Jim Blandy (jimb@totoro.cs.oberlin.edu) * regex.c (compile_range): Fetch the range start and end by casting the pattern pointer to an `unsigned char *' before fetching through it. Sat Dec 12 09:41:01 1992 Jim Blandy (jimb@totoro.cs.oberlin.edu) * regex.c: Undo change of 12/7/92; it's better for Emacs to #define HAVE_CONFIG_H. Fri Dec 11 22:00:34 1992 Jim Meyering (meyering@hal.gnu.ai.mit.edu) * regex.c: Define and use isascii-protected ctype.h macros. Fri Dec 11 05:10:38 1992 Jim Blandy (jimb@totoro.cs.oberlin.edu) * regex.c (re_match_2): Undo Karl's November 10th change; it keeps the group in :\(.*\) from matching :/ properly. Mon Dec 7 19:44:56 1992 Jim Blandy (jimb@wookumz.gnu.ai.mit.edu) * regex.c: #include config.h if either HAVE_CONFIG_H or emacs is #defined. Tue Dec 1 13:33:17 1992 David J. MacKenzie (djm@goldman.gnu.ai.mit.edu) * regex.c [HAVE_CONFIG_H]: Include config.h. Wed Nov 25 23:46:02 1992 David J. MacKenzie (djm@goldman.gnu.ai.mit.edu) * regex.c (regcomp): Add parens around bitwise & for clarity. Initialize preg->allocated to prevent segv. Tue Nov 24 09:22:29 1992 David J. MacKenzie (djm@goldman.gnu.ai.mit.edu) * regex.c: Use HAVE_STRING_H, not USG. * configure.in: Check for string.h, not USG. Fri Nov 20 06:33:24 1992 Karl Berry (karl@cs.umb.edu) * regex.c (SIGN_EXTEND_CHAR) [VMS]: Back out of this change, since Roland Roberts now says it was a localism. Mon Nov 16 07:01:36 1992 Karl Berry (karl@cs.umb.edu) * regex.h (const) [!HAVE_CONST]: Test another cpp symbol (from Autoconf) before zapping const. Sun Nov 15 05:36:42 1992 Jim Blandy (jimb@wookumz.gnu.ai.mit.edu) * regex.c, regex.h: Changes for VMS from Roland B Roberts . Thu Nov 12 11:31:15 1992 Karl Berry (karl@cs.umb.edu) * Makefile.in (distfiles): Include INSTALL. Tue Nov 10 09:29:23 1992 Karl Berry (karl@cs.umb.edu) * regex.c (re_match_2): At maybe_pop_jump, if at end of string and pattern, just quit the matching loop. * regex.c (LETTER_P): Rename to `WORDCHAR_P'. * regex.c (AT_STRINGS_{BEG,END}): Take `d' as an arg; change callers. * regex.c (re_match_2) [!emacs]: In wordchar and notwordchar cases, advance d. Wed Nov 4 15:43:58 1992 Karl Berry (karl@hal.gnu.ai.mit.edu) * regex.h (const) [!__STDC__]: Don't define if it's already defined. Sat Oct 17 19:28:19 1992 Karl Berry (karl@cs.umb.edu) * regex.c (bcmp, bcopy, bzero): Only #define if they are not already #defined. * configure.in: Use AC_CONST. Thu Oct 15 08:39:06 1992 Karl Berry (karl@cs.umb.edu) * regex.h (const) [!const]: Conditionalize. Fri Oct 2 13:31:42 1992 Karl Berry (karl@cs.umb.edu) * regex.h (RE_SYNTAX_ED): New definition. Sun Sep 20 12:53:39 1992 Karl Berry (karl@cs.umb.edu) * regex.[ch]: remove traces of `longest_p' -- dumb idea to put this into the pattern buffer, as it means parallelism loses. * Makefile.in (config.status): use sh to run configure --no-create. * Makefile.in (realclean): OK, don't remove configure. Sat Sep 19 09:05:08 1992 Karl Berry (karl@hayley) * regex.c (PUSH_FAILURE_POINT, POP_FAILURE_POINT) [DEBUG]: keep track of how many failure points we push and pop. (re_match_2) [DEBUG]: declare variables for that, and print results. (DEBUG_PRINT4): new macro. * regex.h (re_pattern_buffer): new field `longest_p' (to eliminate backtracking if the user doesn't need it). * regex.c (re_compile_pattern): initialize it (to 1). (re_search_2): set it to zero if register information is not needed. (re_match_2): if it's set, don't backtrack. * regex.c (re_search_2): update fastmap only after checking that the pattern is anchored. * regex.c (re_match_2): do more debugging at maybe_pop_jump. * regex.c (re_search_2): cast result of TRANSLATE for use in array subscript. Thu Sep 17 19:47:16 1992 Karl Berry (karl@geech.gnu.ai.mit.edu) * Version 0.11. Wed Sep 16 08:17:10 1992 Karl Berry (karl@hayley) * regex.c (INIT_FAIL_STACK): rewrite as statements instead of a complicated comma expr, to avoid compiler warnings (and also simplify). (re_compile_fastmap, re_match_2): change callers. * regex.c (POP_FAILURE_POINT): cast pop of regstart and regend to avoid compiler warnings. * regex.h (RE_NEWLINE_ORDINARY): remove this syntax bit, and remove uses. * regex.c (at_{beg,end}line_loc_p): go the last mile: remove the RE_NEWLINE_ORDINARY case which made the ^ in \n^ be an anchor. Tue Sep 15 09:55:29 1992 Karl Berry (karl@hayley) * regex.c (at_begline_loc_p): new fn. (at_endline_loc_p): simplify at_endline_op_p. (regex_compile): in ^/$ cases, call the above. * regex.c (POP_FAILURE_POINT): rewrite the fn as a macro again, as lord's profiling indicates the function is 20% of the time. (re_match_2): callers changed. * configure.in (AC_MEMORY_H): remove, since we never use memcpy et al. Mon Sep 14 17:49:27 1992 Karl Berry (karl@hayley) * Makefile.in (makeargs): include MFLAGS. Sun Sep 13 07:41:45 1992 Karl Berry (karl@hayley) * regex.c (regex_compile): in \1..\9 case, make it always invalid to use \ if there is no preceding th subexpr. * regex.h (RE_NO_MISSING_BK_REF): remove this syntax bit. * regex.c (regex_compile): remove support for invalid empty groups. * regex.h (RE_NO_EMPTY_GROUPS): remove this syntax bit. * regex.c (FREE_VARIABLES) [!REGEX_MALLOC]: define as alloca (0), to reclaim memory. * regex.h (RE_SYNTAX_POSIX_SED): don't bother with this. Sat Sep 12 13:37:21 1992 Karl Berry (karl@hayley) * README: incorporate emacs.diff. * regex.h (_RE_ARGS) [!__STDC__]: define as empty parens. * configure.in: add AC_ALLOCA. * Put test files in subdir test, documentation in subdir doc. Adjust Makefile.in and configure.in accordingly. Thu Sep 10 10:29:11 1992 Karl Berry (karl@hayley) * regex.h (RE_SYNTAX_{POSIX_,}SED): new definitions. Wed Sep 9 06:27:09 1992 Karl Berry (karl@hayley) * Version 0.10. Tue Sep 8 07:32:30 1992 Karl Berry (karl@hayley) * xregex.texinfo: put the day of month into the date. * Makefile.in (realclean): remove Texinfo-generated files. (distclean): remove empty sorted index files. (clean): remove dvi files, etc. * configure.in: test for more Unix variants. * fileregex.c: new file. Makefile.in (fileregex): new target. * iregex.c (main): move variable decls to smallest scope. * regex.c (FREE_VARIABLES): free reg_{,info_}dummy. (re_match_2): check that the allocation for those two succeeded. * regex.c (FREE_VAR): replace FREE_NONNULL with this. (FREE_VARIABLES): call it. (re_match_2) [REGEX_MALLOC]: initialize all our vars to NULL. * tregress.c (do_match): generalize simple_match. (SIMPLE_NONMATCH): new macro. (SIMPLE_MATCH): change from routine. * Makefile.in (regex.texinfo): make file readonly, so we don't edit it by mistake. * many files (re_default_syntax): rename to `re_syntax_options'; call re_set_syntax instead of assigning to the variable where possible. Mon Sep 7 10:12:16 1992 Karl Berry (karl@hayley) * syntax.skel: don't use prototypes. * {configure,Makefile}.in: new files. * regex.c: include `#if USG || STDC_HEADERS'; remove obsolete test for `POSIX', and test for BSRTING. Include if we are not USG or STDC_HEADERS. Do not include . What did we ever need that for? * regex.h (RE_NO_EMPTY_ALTS): remove this. (RE_SYNTAX_AWK): remove from here, too. * regex.c (regex_compile): remove the check. * xregex.texinfo (Alternation Operator): update. * other.c (test_others): remove tests for this. * regex.h (RE_DUP_MAX): undefine if already defined. * regex.h: (RE_SYNTAX_POSIX*): redo to allow more operators, and define new syntaxes with the minimal set. * syntax.skel (main): used sscanf instead of scanf. * regex.h (RE_SYNTAX_*GREP): new definitions from mike. * regex.c (regex_compile): initialize the upper bound of intervals at the beginning of the interval, not the end. (From pclink@qld.tne.oz.au.) * regex.c (handle_bar): rename to `handle_alt', for consistency. * regex.c ({store,insert}_{op1,op2}): new routines (except the last). ({STORE,INSERT}_JUMP{,2}): macros to replace the old routines, which took arguments in different orders, and were generally weird. * regex.c (PAT_PUSH*): rename to `BUF_PUSH*' -- we're not appending info to the pattern! Sun Sep 6 11:26:49 1992 Karl Berry (karl@hayley) * regex.c (regex_compile): delete the variable `following_left_brace', since we never use it. * regex.c (print_compiled_pattern): don't print the fastmap if it's null. * regex.c (re_compile_fastmap): handle `on_failure_keep_string_jump' like `on_failure_jump'. * regex.c (re_match_2): in `charset{,_not' case, cast the bit count to unsigned, not unsigned char, in case we have a full 32-byte bit list. * tregress.c (simple_match): remove. (simple_test): rename as `simple_match'. (simple_compile): print the error string if the compile failed. * regex.c (DO_RANGE): rewrite as a function, `compile_range', so we can debug it. Change pattern characters to unsigned char *'s, and change the range variable to an unsigned. (regex_compile): change calls. Sat Sep 5 17:40:49 1992 Karl Berry (karl@hayley) * regex.h (_RE_ARGS): new macro to put in argument lists (if ANSI) or omit them (if K&R); don't declare routines twice. * many files (obscure_syntax): rename to `re_default_syntax'. Fri Sep 4 09:06:53 1992 Karl Berry (karl@hayley) * GNUmakefile (extraclean): new target. (realclean): delete the info files. Wed Sep 2 08:14:42 1992 Karl Berry (karl@hayley) * regex.h: doc fix. Sun Aug 23 06:53:15 1992 Karl Berry (karl@hayley) * regex.[ch] (re_comp): no const in the return type (from djm). Fri Aug 14 07:25:46 1992 Karl Berry (karl@hayley) * regex.c (DO_RANGE): declare variables as unsigned chars, not signed chars (from jimb). Wed Jul 29 18:33:53 1992 Karl Berry (karl@claude.cs.umb.edu) * Version 0.9. * GNUmakefile (distclean): do not remove regex.texinfo. (realclean): remove it here. * tregress.c (simple_test): initialize buf.buffer. Sun Jul 26 08:59:38 1992 Karl Berry (karl@hayley) * regex.c (push_dummy_failure): new opcode and corresponding case in the various routines. Pushed at the end of alternatives. * regex.c (jump_past_next_alt): rename to `jump_past_alt', for brevity. (no_pop_jump): rename to `jump'. * regex.c (regex_compile) [DEBUG]: terminate printing of pattern with a newline. * NEWS: new file. * tregress.c (simple_{compile,match,test}): routines to simplify all these little tests. * tregress.c: test for matching as much as possible. Fri Jul 10 06:53:32 1992 Karl Berry (karl@hayley) * Version 0.8. Wed Jul 8 06:39:31 1992 Karl Berry (karl@hayley) * regex.c (SIGN_EXTEND_CHAR): #undef any previous definition, as ours should always work properly. Mon Jul 6 07:10:50 1992 Karl Berry (karl@hayley) * iregex.c (main) [DEBUG]: conditionalize the call to print_compiled_pattern. * iregex.c (main): initialize buf.buffer to NULL. * tregress (test_regress): likewise. * regex.c (alloca) [sparc]: #if on HAVE_ALLOCA_H instead. * tregress.c (test_regress): didn't have jla's test quite right. Sat Jul 4 09:02:12 1992 Karl Berry (karl@hayley) * regex.c (re_match_2): only REGEX_ALLOCATE all the register vectors if the pattern actually has registers. (match_end): new variable to avoid having to use best_regend[0]. * regex.c (IS_IN_FIRST_STRING): rename to FIRST_STRING_P. * regex.c: doc fixes. * tregess.c (test_regress): new fastmap test forwarded by rms. * tregress.c (test_regress): initialize the fastmap field. * tregress.c (test_regress): new test from jla that aborted in re_search_2. Fri Jul 3 09:10:05 1992 Karl Berry (karl@hayley) * tregress.c (test_regress): add tests for translating charsets, from kaoru. * GNUmakefile (common): add alloca.o. * alloca.c: new file, copied from bison. * other.c (test_others): remove var `buf', since it's no longer used. * Below changes from ro@TechFak.Uni-Bielefeld.DE. * tregress.c (test_regress): initialize buf.allocated. * regex.c (re_compile_fastmap): initialize `succeed_n_p'. * GNUmakefile (regex): depend on $(common). Wed Jul 1 07:12:46 1992 Karl Berry (karl@hayley) * Version 0.7. * regex.c: doc fixes. Mon Jun 29 08:09:47 1992 Karl Berry (karl@fosse) * regex.c (pop_failure_point): change string vars to `const char *' from `unsigned char *'. * regex.c: consolidate debugging stuff. (print_partial_compiled_pattern): avoid enum clash. Mon Jun 29 07:50:27 1992 Karl Berry (karl@hayley) * xmalloc.c: new file. * GNUmakefile (common): add it. * iregex.c (print_regs): new routine (from jimb). (main): call it. Sat Jun 27 10:50:59 1992 Jim Blandy (jimb@pogo.cs.oberlin.edu) * xregex.c (re_match_2): When we have accepted a match and restored d from best_regend[0], we need to set dend appropriately as well. Sun Jun 28 08:48:41 1992 Karl Berry (karl@hayley) * tregress.c: rename from regress.c. * regex.c (print_compiled_pattern): improve charset case to ease byte-counting. Also, don't distinguish between Emacs and non-Emacs {not,}wordchar opcodes. * regex.c (print_fastmap): move here. * test.c: from here. * regex.c (print_{{partial,}compiled_pattern,double_string}): rename from ..._printer. Change calls here and in test.c. * regex.c: create from xregex.c and regexinc.c for once and for all, and change the debug fns to be extern, instead of static. * GNUmakefile: remove traces of xregex.c. * test.c: put in externs, instead of including regexinc.c. * xregex.c: move interactive main program and scanstring to iregex.c. * iregex.c: new file. * upcase.c, printchar.c: new files. * various doc fixes and other cosmetic changes throughout. * regexinc.c (compiled_pattern_printer): change variable name, for consistency. (partial_compiled_pattern_printer): print other info about the compiled pattern, besides just the opcodes. * xregex.c (regex_compile) [DEBUG]: print the compiled pattern when we're done. * xregex.c (re_compile_fastmap): in the duplicate case, set `can_be_null' and return. Also, set `bufp->can_be_null' according to a new variable, `path_can_be_null'. Also, rewrite main while loop to not test `p != NULL', since we never set it that way. Also, eliminate special `can_be_null' value for the endline case. (re_search_2): don't test for the special value. * regex.h (struct re_pattern_buffer): remove the definition. Sat Jun 27 15:00:40 1992 Karl Berry (karl@hayley) * xregex.c (re_compile_fastmap): remove the `RE_' from `REG_RE_MATCH_NULL_AT_END'. Also, assert the fastmap in the pattern buffer is non-null. Also, reset `succeed_n_p' after we've paid attention to it, instead of every time through the loop. Also, in the `anychar' case, only clear fastmap['\n'] if the syntax says to, and don't return prematurely. Also, rearrange cases in some semblance of a rational order. * regex.h (REG_RE_MATCH_NULL_AT_END): remove the `RE_' from the name. * other.c: take bug reports from here. * regress.c: new file for them. * GNUmakefile (test): add it. * main.c (main): new possible test. * test.h (test_type): new value in enum. Thu Jun 25 17:37:43 1992 Karl Berry (karl@hayley) * xregex.c (scanstring) [test]: new function from jimb to allow some escapes. (main) [test]: call it (on the string, not the pattern). * xregex.c (main): make return type `int'. Wed Jun 24 10:43:03 1992 Karl Berry (karl@hayley) * xregex.c (pattern_offset_t): change to `int', for the benefit of patterns which compile to more than 2^15 bytes. * xregex.c (GET_BUFFER_SPACE): remove spurious braces. * xregex.texinfo (Using Registers): put in a stub to ``document'' the new function. * regex.h (re_set_registers) [!__STDC__]: declare. * xregex.c (re_set_registers): declare K&R style (also move to a different place in the file). Mon Jun 8 18:03:28 1992 Jim Blandy (jimb@pogo.cs.oberlin.edu) * regex.h (RE_NREGS): Doc fix. * xregex.c (re_set_registers): New function. * regex.h (re_set_registers): Declaration for new function. Fri Jun 5 06:55:18 1992 Karl Berry (karl@hayley) * main.c (main): `return 0' instead of `exit (0)'. (From Paul Eggert) * regexinc.c (SIGN_EXTEND_CHAR): cast to unsigned char. (extract_number, EXTRACT_NUMBER): don't bother to cast here. Tue Jun 2 07:37:53 1992 Karl Berry (karl@hayley) * Version 0.6. * Change copyrights to `1985, 89, ...'. * regex.h (REG_RE_MATCH_NULL_AT_END): new macro. * xregex.c (re_compile_fastmap): initialize `can_be_null' to `p==pend', instead of in the test at the top of the loop (as it was, it was always being set). Also, set `can_be_null'=1 if we would jump to the end of the pattern in the `on_failure_jump' cases. (re_search_2): check if `can_be_null' is 1, not nonzero. This was the original test in rms' regex; why did we change this? * xregex.c (re_compile_fastmap): rename `is_a_succeed_n' to `succeed_n_p'. Sat May 30 08:09:08 1992 Karl Berry (karl@hayley) * xregex.c (re_compile_pattern): declare `regnum' as `unsigned', not `regnum_t', for the benefit of those patterns with more than 255 groups. * xregex.c: rename `failure_stack' to `fail_stack', for brevity; likewise for `match_nothing' to `match_null'. * regexinc.c (REGEX_REALLOCATE): take both the new and old sizes, and copy only the old bytes. * xregex.c (DOUBLE_FAILURE_STACK): pass both old and new. * This change from Thorsten Ohl. Fri May 29 11:45:22 1992 Karl Berry (karl@hayley) * regexinc.c (SIGN_EXTEND_CHAR): define as `(signed char) c' instead of relying on __CHAR_UNSIGNED__, to work with compilers other than GCC. From Per Bothner. * main.c (main): change return type to `int'. Mon May 18 06:37:08 1992 Karl Berry (karl@hayley) * regex.h (RE_SYNTAX_AWK): typo in RE_RE_UNMATCHED... Fri May 15 10:44:46 1992 Karl Berry (karl@hayley) * Version 0.5. Sun May 3 13:54:00 1992 Karl Berry (karl@hayley) * regex.h (struct re_pattern_buffer): now it's just `regs_allocated'. (REGS_UNALLOCATED, REGS_REALLOCATE, REGS_FIXED): new constants. * xregex.c (regexec, re_compile_pattern): set the field appropriately. (re_match_2): and use it. bufp can't be const any more. Fri May 1 15:43:09 1992 Karl Berry (karl@hayley) * regexinc.c: unconditionally include , first. * regex.h (struct re_pattern_buffer): rename `caller_allocated_regs' to `regs_allocated_p'. * xregex.c (re_compile_pattern): same change here. (regexec): and here. (re_match_2): reallocate registers if necessary. Fri Apr 10 07:46:50 1992 Karl Berry (karl@hayley) * regex.h (RE_SYNTAX{_POSIX,}_AWK): new definitions from Arnold. Sun Mar 15 07:34:30 1992 Karl Berry (karl at hayley) * GNUmakefile (dist): versionize regex.{c,h,texinfo}. Tue Mar 10 07:05:38 1992 Karl Berry (karl at hayley) * Version 0.4. * xregex.c (PUSH_FAILURE_POINT): always increment the failure id. (DEBUG_STATEMENT) [DEBUG]: execute the statement even if `debug'==0. * xregex.c (pop_failure_point): if the saved string location is null, keep the current value. (re_match_2): at fail, test for a dummy failure point by checking the restored pattern value, not string value. (re_match_2): new case, `on_failure_keep_string_jump'. (regex_compile): output this opcode in the .*\n case. * regexinc.c (re_opcode_t): define the opcode. (partial_compiled_pattern_pattern): add the new case. Mon Mar 9 09:09:27 1992 Karl Berry (karl at hayley) * xregex.c (regex_compile): optimize .*\n to output an unconditional jump to the ., instead of pushing failure points each time through the loop. * xregex.c (DOUBLE_FAILURE_STACK): compute the maximum size ourselves (and correctly); change callers. Sun Mar 8 17:07:46 1992 Karl Berry (karl at hayley) * xregex.c (failure_stack_elt_t): change to `const char *', to avoid warnings. * regex.h (re_set_syntax): declare this. * xregex.c (pop_failure_point) [DEBUG]: conditionally pass the original strings and sizes; change callers. Thu Mar 5 16:35:35 1992 Karl Berry (karl at claude.cs.umb.edu) * xregex.c (regnum_t): new type for register/group numbers. (compile_stack_elt_t, regex_compile): use it. * xregex.c (regexec): declare len as `int' to match re_search. * xregex.c (re_match_2): don't declare p1 twice. * xregex.c: change `while (1)' to `for (;;)' to avoid silly compiler warnings. * regex.h [__STDC__]: use #if, not #ifdef. * regexinc.c (REGEX_REALLOCATE): cast the result of alloca to (char *), to avoid warnings. * xregex.c (regerror): declare variable as const. * xregex.c (re_compile_pattern, re_comp): define as returning a const char *. * regex.h (re_compile_pattern, re_comp): likewise. Thu Mar 5 15:57:56 1992 Karl Berry (karl@hal) * xregex.c (regcomp): declare `syntax' as unsigned. * xregex.c (re_match_2): try to avoid compiler warnings about unsigned comparisons. * GNUmakefile (test-xlc): new target. * regex.h (reg_errcode_t): remove trailing comma from definition. * regexinc.c (re_opcode_t): likewise. Thu Mar 5 06:56:07 1992 Karl Berry (karl at hayley) * GNUmakefile (dist): add version numbers automatically. (versionfiles): new variable. (regex.{c,texinfo}): don't add version numbers here. * regex.h: put in placeholder instead of the version number. Fri Feb 28 07:11:33 1992 Karl Berry (karl at hayley) * xregex.c (re_error_msg): declare const, since it is. Sun Feb 23 05:41:57 1992 Karl Berry (karl at fosse) * xregex.c (PAT_PUSH{,_2,_3}, ...): cast args to avoid warnings. (regex_compile, regexec): return REG_NOERROR, instead of 0, on success. (boolean): define as char, and #define false and true. * regexinc.c (STREQ): cast the result. Sun Feb 23 07:45:38 1992 Karl Berry (karl at hayley) * GNUmakefile (test-cc, test-hc, test-pcc): new targets. * regex.inc (extract_number, extract_number_and_incr) [DEBUG]: only define if we are debugging. * xregex.c [_AIX]: do #pragma alloca first if necessary. * regexinc.c [_AIX]: remove the #pragma from here. * regex.h (reg_syntax_t): declare as unsigned, and redo the enum as #define's again. Some compilers do stupid things with enums. Thu Feb 20 07:19:47 1992 Karl Berry (karl at hayley) * Version 0.3. * xregex.c, regex.h (newline_anchor_match_p): rename to `newline_anchor'; dumb idea to change the name. Tue Feb 18 07:09:02 1992 Karl Berry (karl at hayley) * regexinc.c: go back to original, i.e., don't include or define strchr. * xregex.c (regexec): don't bother with adding characters after newlines to the fastmap; instead, just don't use a fastmap. * xregex.c (regcomp): set the buffer and fastmap fields to zero. * xregex.texinfo (GNU r.e. compiling): have to initialize more than two fields. * regex.h (struct re_pattern_buffer): rename `newline_anchor' to `newline_anchor_match_p', as we're back to two cases. * xregex.c (regcomp, re_compile_pattern, re_comp): change accordingly. (re_match_2): at begline and endline, POSIX is not a special case anymore; just check newline_anchor_match_p. Thu Feb 13 16:29:33 1992 Karl Berry (karl at hayley) * xregex.c (*empty_string*): rename to *null_string*, for brevity. Wed Feb 12 06:36:22 1992 Karl Berry (karl at hayley) * xregex.c (re_compile_fastmap): at endline, don't set fastmap['\n']. (re_match_2): rewrite the begline/endline cases to take account of the new field newline_anchor. Tue Feb 11 14:34:55 1992 Karl Berry (karl at hayley) * regexinc.c [!USG etc.]: include and define strchr as index. * xregex.c (re_search_2): when searching backwards, declare `c' as a char and use casts when using it as an array subscript. * xregex.c (regcomp): if REG_NEWLINE, set RE_HAT_LISTS_NOT_NEWLINE. Set the `newline_anchor' field appropriately. (regex_compile): compile [^...] as matching a \n according to the syntax bit. (regexec): if doing REG_NEWLINE stuff, compile a fastmap and add characters after any \n's to the newline. * regex.h (RE_HAT_LISTS_NOT_NEWLINE): new syntax bit. (struct re_pattern_buffer): rename `posix_newline' to `newline_anchor', define constants for its values. Mon Feb 10 07:22:50 1992 Karl Berry (karl at hayley) * xregex.c (re_compile_fastmap): combine the code at the top and bottom of the loop, as it's essentially identical. Sun Feb 9 10:02:19 1992 Karl Berry (karl at hayley) * xregex.texinfo (POSIX Translate Tables): remove this, as it doesn't match the spec. * xregex.c (re_compile_fastmap): if we finish off a path, go back to the top (to set can_be_null) instead of returning immediately. * xregex.texinfo: changes from bob. Sat Feb 1 07:03:25 1992 Karl Berry (karl at hayley) * xregex.c (re_search_2): doc fix (from rms). Fri Jan 31 09:52:04 1992 Karl Berry (karl at hayley) * xregex.texinfo (GNU Searching): clarify the range arg. * xregex.c (re_match_2, at_endline_op_p): add extra parens to get rid of GCC 2's (silly, IMHO) warning about && within ||. * xregex.c (common_op_match_empty_string_p): use MATCH_NOTHING_UNSET_VALUE, not -1. Thu Jan 16 08:43:02 1992 Karl Berry (karl at hayley) * xregex.c (SET_REGS_MATCHED): only set the registers from lowest to highest. * regexinc.c (MIN): new macro. * xregex.c (re_match_2): only check min (num_regs, regs->num_regs) when we set the returned regs. * xregex.c (re_match_2): set registers after the first num_regs to -1 before we return. Tue Jan 14 16:01:42 1992 Karl Berry (karl at hayley) * xregex.c (re_match_2): initialize max (RE_NREGS, re_nsub + 1) registers (from rms). * xregex.c, regex.h: don't abbreviate `19xx' to `xx'. * regexinc.c [!emacs]: include before . (from ro@thp.Uni-Koeln.DE). Thu Jan 9 07:23:00 1992 Karl Berry (karl at hayley) * xregex.c (*unmatchable): rename to `match_empty_string_p'. (CAN_MATCH_NOTHING): rename to `REG_MATCH_EMPTY_STRING_P'. * regexinc.c (malloc, realloc): remove prototypes, as they can cause clashes (from rms). Mon Jan 6 12:43:24 1992 Karl Berry (karl at claude.cs.umb.edu) * Version 0.2. Sun Jan 5 10:50:38 1992 Karl Berry (karl at hayley) * xregex.texinfo: bring more or less up-to-date. * GNUmakefile (regex.texinfo): generate from regex.h and xregex.texinfo. * include.awk: new file. * xregex.c: change all calls to the fn extract_number_and_incr to the macro. * xregex.c (re_match_2) [emacs]: in at_dot, use PTR_CHAR_POS + 1, instead of bf_* and sl_*. Cast d to unsigned char *, to match the declaration in Emacs' buffer.h. [emacs19]: in before_dot, at_dot, and after_dot, likewise. * regexinc.c: unconditionally include . * regexinc.c (alloca) [!alloca]: Emacs config files sometimes define this, so don't define it if it's already defined. Sun Jan 5 06:06:53 1992 Karl Berry (karl at fosse) * xregex.c (re_comp): fix type conflicts with regex_compile (we haven't been compiling this). * regexinc.c (SIGN_EXTEND_CHAR): use `__CHAR_UNSIGNED__', not `CHAR_UNSIGNED'. * regexinc.c (NULL) [!NULL]: define it (as zero). * regexinc.c (extract_number): remove the temporaries. Sun Jan 5 07:50:14 1992 Karl Berry (karl at hayley) * regex.h (regerror) [!__STDC__]: return a size_t, not a size_t *. * xregex.c (PUSH_FAILURE_POINT, ...): declare `destination' as `char *' instead of `void *', to match alloca declaration. * xregex.c (regerror): use `size_t' for the intermediate values as well as the return type. * xregex.c (regexec): cast the result of malloc. * xregex.c (regexec): don't initialize `private_preg' in the declaration, as old C compilers can't do that. * xregex.c (main) [test]: declare printchar void. * xregex.c (assert) [!DEBUG]: define this to do nothing, and remove #ifdef DEBUG's from around asserts. * xregex.c (re_match_2): remove error message when not debugging. Sat Jan 4 09:45:29 1992 Karl Berry (karl at hayley) * other.c: test the bizarre duplicate case in re_compile_fastmap that I just noticed. * test.c (general_test): don't test registers beyond the end of correct_regs, as well as regs. * xregex.c (regex_compile): at handle_close, don't assign to *inner_group_loc if we didn't push a start_memory (because the group number was too big). In fact, don't push or pop the inner_group_offset in that case. * regex.c: rename to xregex.c, since it's not the whole thing. * regex.texinfo: likewise. * GNUmakefile: change to match. * regex.c [DEBUG]: only include if debugging. * regexinc.c (SIGN_EXTEND_CHAR) [CHAR_UNSIGNED]: if it's already defined, don't redefine it. * regex.c: define _GNU_SOURCE at the beginning. * regexinc.c (isblank) [!isblank]: define it. (isgraph) [!isgraph]: change conditional to this, and remove the sequent stuff. * regex.c (regex_compile): add `blank' character class. * regex.c (regex_compile): don't use a uchar variable to loop through all characters. * regex.c (regex_compile): at '[', improve logic for checking that we have enough space for the charset. * regex.h (struct re_pattern_buffer): declare translate as char * again. We only use it as an array subscript once, I think. * regex.c (TRANSLATE): new macro to cast the data character before subscripting. (num_internal_regs): rename to `num_regs'. Fri Jan 3 07:58:01 1992 Karl Berry (karl at hayley) * regex.h (struct re_pattern_buffer): declare `allocated' and `used' as unsigned long, since these are never negative. * regex.c (compile_stack_element): rename to compile_stack_elt_t. (failure_stack_element): similarly. * regexinc.c (TALLOC, RETALLOC): new macros to simplify allocation of arrays. * regex.h (re_*) [__STDC__]: don't declare string args unsigned char *; that makes them incompatible with string constants. (struct re_pattern_buffer): declare the pattern and translate table as unsigned char *. * regex.c (most routines): use unsigned char vs. char consistently. * regex.h (re_compile_pattern): do not declare the length arg as const. * regex.c (re_compile_pattern): likewise. * regex.c (POINTER_TO_REG): rename to `POINTER_TO_OFFSET'. * regex.h (re_registers): declare `start' and `end' as `regoff_t', instead of `int'. * regex.c (regexec): if either of the malloc's for the register information fail, return failure. * regex.h (RE_NREGS): define this again, as 30 (from jla). (RE_ALLOCATE_REGISTERS): remove this. (RE_SYNTAX_*): remove it from definitions. (re_pattern_buffer): remove `return_default_num_regs', add `caller_allocated_regs'. * regex.c (re_compile_pattern): clear no_sub and caller_allocated_regs in the pattern. (regcomp): set caller_allocated_regs. (re_match_2): do all register allocation at the end of the match; implement new semantics. * regex.c (MAX_REGNUM): new macro. (regex_compile): at handle_open and handle_close, if the group number is too large, don't push the start/stop memory. Thu Jan 2 07:56:10 1992 Karl Berry (karl at hayley) * regex.c (re_match_2): if the back reference is to a group that never matched, then goto fail, not really_fail. Also, don't test if the pattern can match the empty string. Why did we ever do that? (really_fail): this label no longer needed. * regexinc.c [STDC_HEADERS]: use only this to test if we should include . * regex.c (DO_RANGE, regex_compile): translate in all cases except the single character after a \. * regex.h (RE_AWK_CLASS_HACK): rename to RE_BACKSLASH_ESCAPE_IN_LISTS. * regex.c (regex_compile): change use. * regex.c (re_compile_fastmap): do not translate the characters again; we already translated them at compilation. (From ylo@ngs.fi.) * regex.c (re_match_2): in case for at_dot, invert sense of comparison and find the character number properly. (From worley@compass.com.) (re_match_2) [emacs]: remove the cases for before_dot and after_dot, since there's no way to specify them, and the code is wrong (judging from this change). Wed Jan 1 09:13:38 1992 Karl Berry (karl at hayley) * psx-{interf,basic,extend}.c, other.c: set `t' as the first thing, so that if we run them in sucession, general_test's kludge to see if we're doing POSIX tests works. * test.h (test_type): add `all_test'. * main.c: add case for `all_test'. * regexinc.c (partial_compiled_pattern_printer, double_string_printer): don't print anything if we're passed null. * regex.c (PUSH_FAILURE_POINT): do not scan for the highest and lowest active registers. (re_match_2): compute lowest/highest active regs at start_memory and stop_memory. (NO_{LOW,HIGH}EST_ACTIVE_REG): new sentinel values. (pop_failure_point): return the lowest/highest active reg values popped; change calls. * regex.c [DEBUG]: include . (various routines) [DEBUG]: change conditionals to assertions. * regex.c (DEBUG_STATEMENT): new macro. (PUSH_FAILURE_POINT): use it to increment num_regs_pushed. (re_match_2) [DEBUG]: only declare num_regs_pushed if DEBUG. * regex.c (*can_match_nothing): rename to *unmatchable. * regex.c (re_match_2): at stop_memory, adjust argument reading. * regex.h (re_pattern_buffer): declare `can_be_null' as a 2-bit bit field. * regex.h (re_pattern_buffer): declare `buffer' unsigned char *; no, dumb idea. The pattern can have signed number. * regex.c (re_match_2): in maybe_pop_jump case, skip over the right number of args to the group operators, and don't do anything with endline if posix_newline is not set. * regex.c, regexinc.c (all the things we just changed): go back to putting the inner group count after the start_memory, because we need it in the on_failure_jump case in re_match_2. But leave it after the stop_memory also, since we need it there in re_match_2, and we don't have any way of getting back to the start_memory. * regexinc.c (partial_compiled_pattern_printer): adjust argument reading for start/stop_memory. * regex.c (re_compile_fastmap, group_can_match_nothing): likewise. Tue Dec 31 10:15:08 1991 Karl Berry (karl at hayley) * regex.c (bits list routines): remove these. (re_match_2): get the number of inner groups from the pattern, instead of keeping track of it at start and stop_memory. Put the count after the stop_memory, not after the start_memory. (compile_stack_element): remove `fixup_inner_group' member, since we now put it in when we can compute it. (regex_compile): at handle_open, don't push the inner group offset, and at handle_close, don't pop it. * regex.c (level routines): remove these, and their uses in regex_compile. This was another manifestation of having to find $'s that were endlines. * regex.c (regexec): this does searching, not matching (a well-disguised part of the standard). So rewrite to use `re_search' instead of `re_match'. * psx-interf.c (test_regexec): add tests to, uh, match. * regex.h (RE_TIGHT_ALT): remove this; nobody uses it. * regex.c: remove the code that was supposed to implement it. * other.c (test_others): ^ and $ never match newline characters; RE_CONTEXT_INVALID_OPS doesn't affect anchors. * psx-interf.c (test_regerror): update for new error messages. * psx-extend.c: it's now ok to have an alternative be just a $, so remove all the tests which supposed that was invalid. Wed Dec 25 09:00:05 1991 Karl Berry (karl at hayley) * regex.c (regex_compile): in handle_open, don't skip over ^ and $ when checking for an empty group. POSIX has changed the grammar. * psx-extend.c (test_posix_extended): thus, move (^$) tests to valid section. * regexinc.c (boolean): move from here to test.h and regex.c. * test files: declare verbose, omit_register_tests, and test_should_match as boolean. * psx-interf.c (test_posix_c_interface): remove the `c_'. * main.c: likewise. * psx-basic.c (test_posix_basic): ^ ($) is an anchor after (before) an open (close) group. * regex.c (re_match_2): in endline, correct precedence of posix_newline condition. Tue Dec 24 06:45:11 1991 Karl Berry (karl at hayley) * test.h: incorporate private-tst.h. * test files: include test.h, not private-tst.h. * test.c (general_test): set posix_newline to zero if we are doing POSIX tests (unfortunately, it's difficult to call regcomp in this case, which is what we should really be doing). * regex.h (reg_syntax_t): make this an enumeration type which defines the syntax bits; renames re_syntax_t. * regex.c (at_endline_op_p): don't preincrement p; then if it's not an empty string op, we lose. * regex.h (reg_errcode_t): new enumeration type of the error codes. * regex.c (regex_compile): return that type. * regex.c (regex_compile): in [, initialize just_had_a_char_class to false; somehow I had changed this to true. * regex.h (RE_NO_CONSECUTIVE_REPEATS): remove this, since we don't use it, and POSIX doesn't require this behavior anymore. * regex.c (regex_compile): remove it from here. * regex.c (regex_compile): remove the no_op insertions for verify_and_adjust_endlines, since that doesn't exist anymore. * regex.c (regex_compile) [DEBUG]: use printchar to print the pattern, so unprintable bytes will print properly. * regex.c: move re_error_msg back. * test.c (general_test): print the compile error if the pattern was invalid. Mon Dec 23 08:54:53 1991 Karl Berry (karl at hayley) * regexinc.c: move re_error_msg here. * regex.c (re_error_msg): the ``message'' for success must be NULL, to keep the interface to re_compile_pattern the same. (regerror): if the msg is null, use "Success". * rename most test files for consistency. Change Makefile correspondingly. * test.c (most routines): add casts to (unsigned char *) when we call re_{match,search}{,_2}. Sun Dec 22 09:26:06 1991 Karl Berry (karl at hayley) * regex.c (re_match_2): declare string args as unsigned char * again; don't declare non-pointer args const; declare the pattern buffer const. (re_match): likewise. (re_search_2, re_search): likewise, except don't declare the pattern const, since we make a fastmap. * regex.h [__STDC__]: change prototypes. * regex.c (regex_compile): return an error code, not a string. (re_err_list): new table to map from error codes to string. (re_compile_pattern): return an element of re_err_list. (regcomp): don't test all the strings. (regerror): just use the list. (put_in_buffer): remove this. * regex.c (equivalent_failure_points): remove this. * regex.c (re_match_2): don't copy the string arguments into non-const pointers. We never alter the data. * regex.c (re_match_2): move assignment to `is_a_jump_n' out of the main loop. Just initialize it right before we do something with it. * regex.[ch] (re_match_2): don't declare the int parameters const. Sat Dec 21 08:52:20 1991 Karl Berry (karl at hayley) * regex.h (re_syntax_t): new type; declare to be unsigned (previously we used int, but since we do bit operations on this, unsigned is better, according to H&S). (obscure_syntax, re_pattern_buffer): use that type. * regex.c (re_set_syntax, regex_compile): likewise. * regex.h (re_pattern_buffer): new field `posix_newline'. * regex.c (re_comp, re_compile_pattern): set to zero. (regcomp): set to REG_NEWLINE. * regex.h (RE_HAT_LISTS_NOT_NEWLINE): remove this (we can just check `posix_newline' instead.) * regex.c (op_list_type, op_list, add_op): remove these. (verify_and_adjust_endlines): remove this. (pattern_offset_list_type, *pattern_offset* routines): and these. These things all implemented the nonleading/nontrailing position code, which was very long, had a few remaining problems, and is no longer needed. So... * regexinc.c (STREQ): new macro to abbreviate strcmp(,)==0, for brevity. Change various places in regex.c to use it. * regex{,inc}.c (enum regexpcode): change to a typedef re_opcode_t, for brevity. * regex.h (re_syntax_table) [SYNTAX_TABLE]: remove this; it should only be in regex.c, I think, since we don't define it in this case. Maybe it should be conditional on !SYNTAX_TABLE? * regexinc.c (partial_compiled_pattern_printer): simplify and distinguish the emacs/not-emacs (not)wordchar cases. Fri Dec 20 08:11:38 1991 Karl Berry (karl at hayley) * regexinc.c (regexpcode) [emacs]: only define the Emacs opcodes if we are ifdef emacs. * regex.c (BUF_PUSH*): rename to PAT_PUSH*. * regex.c (regex_compile): in $ case, go back to essentially the original code for deciding endline op vs. normal char. (at_endline_op_p): new routine. * regex.h (RE_ANCHORS_ONLY_AT_ENDS, RE_CONTEXT_INVALID_ANCHORS, RE_REPEATED_ANCHORS_AWAY, RE_NO_ANCHOR_AT_NEWLINE): remove these. POSIX has simplified the rules for anchors in draft 11.2. (RE_NEWLINE_ORDINARY): new syntax bit. (RE_CONTEXT_INDEP_ANCHORS): change description to be compatible with POSIX. * regex.texinfo (Syntax Bits): remove the descriptions. Mon Dec 16 08:12:40 1991 Karl Berry (karl at hayley) * regex.c (re_match_2): in jump_past_next_alt, unconditionally goto no_pop. The only register we were finding was one which enclosed the whole alternative expression, not one around an individual alternative. So we were never doing what we thought we were doing, and this way makes (|a) against the empty string fail. * regex.c (regex_compile): remove `highest_ever_regnum', and don't restore regnum from the stack; just put it into a temporary to put into the stop_memory. Otherwise, groups aren't numbered consecutively. * regex.c (is_in_compile_stack): rename to `group_in_compile_stack'; remove unnecessary test for the stack being empty. * regex.c (re_match_2): in on_failure_jump, skip no_op's before checking for the start_memory, in case we were called from succeed_n. Sun Dec 15 16:20:48 1991 Karl Berry (karl at hayley) * regex.c (regex_compile): in duplicate case, use highest_ever_regnum instead of regnum, since the latter is reverted at stop_memory. * regex.c (re_match_2): in on_failure_jump, if the * applied to a group, save the information for that group and all inner groups (by making it active), even though we're not inside it yet. Sat Dec 14 09:50:59 1991 Karl Berry (karl at hayley) * regex.c (PUSH_FAILURE_ITEM, POP_FAILURE_ITEM): new macros. Use them instead of copying the stack manipulating a zillion times. * regex.c (PUSH_FAILURE_POINT, pop_failure_point) [DEBUG]: save and restore a unique identification value for each failure point. * regexinc.c (partial_compiled_pattern_printer): don't print an extra / after duplicate commands. * regex.c (regex_compile): in back-reference case, allow a back reference to register `regnum'. Otherwise, even `\(\)\1' fails, since regnum is 1 at the back-reference. * regex.c (re_match_2): in fail, don't examine the pattern if we restored to pend. * test_private.h: rename to private_tst.h. Change includes. * regex.c (extend_bits_list): compute existing size for realloc in bytes, not blocks. * regex.c (re_match_2): in jump_past_next_alt, the for loop was missing its (empty) statement. Even so, some register tests still fail, although in a different way than in the previous change. Fri Dec 13 15:55:08 1991 Karl Berry (karl at hayley) * regex.c (re_match_2): in jump_past_next_alt, unconditionally goto no_pop, since we weren't properly detecting if the alternative matched something anyway. No, we need to not jump to keep the register values correct; just change to not look at register zero and not test RE_NO_EMPTY_ALTS (which is a compile-time thing). * regex.c (SET_REGS_MATCHED): start the loop at 1, since we never care about register zero until the very end. (I think.) * regex.c (PUSH_FAILURE_POINT, pop_failure_point): go back to pushing and popping the active registers, instead of only doing the registers before a group: (fooq|fo|o)*qbar against fooqbar fails, since we restore back into the middle of group 1, yet it isn't active, because the previous restore clobbered the active flag. Thu Dec 12 17:25:36 1991 Karl Berry (karl at hayley) * regex.c (PUSH_FAILURE_POINT): do not call `equivalent_failure_points' after all; it causes the registers to be ``wrong'' (according to POSIX), and an infinite loop on `((a*)*)*' against `ab'. * regex.c (re_compile_fastmap): don't push `pend' on the failure stack. Tue Dec 10 10:30:03 1991 Karl Berry (karl at hayley) * regex.c (PUSH_FAILURE_POINT): if pushing same failure point that is on the top of the stack, fail. (equivalent_failure_points): new routine. * regex.c (re_match_2): add debug statements for every opcode we execute. * regex.c (regex_compile/handle_close): restore `fixup_inner_group_count' and `regnum' from the stack. Mon Dec 9 13:51:15 1991 Karl Berry (karl at hayley) * regex.c (PUSH_FAILURE_POINT): declare `this_reg' as int, so unsigned arithmetic doesn't happen when we don't want to save the registers. Tue Dec 3 08:11:10 1991 Karl Berry (karl at hayley) * regex.c (extend_bits_list): divide size by bits/block. * regex.c (init_bits_list): remove redundant assignmen to `bits_list_ptr'. * regexinc.c (partial_compiled_pattern_printer): don't do *p++ twice in the same expr. * regex.c (re_match_2): at on_failure_jump, use the correct pattern positions for getting the stuff following the start_memory. * regex.c (struct register_info): remove the bits_list for the inner groups; make that a separate variable. Mon Dec 2 10:42:07 1991 Karl Berry (karl at hayley) * regex.c (PUSH_FAILURE_POINT): don't pass `failure_stack' as an arg; change callers. * regex.c (PUSH_FAILURE_POINT): print items in order they are pushed. (pop_failure_point): likewise. * regex.c (main): prompt for the pattern and string. * regex.c (FREE_VARIABLES) [!REGEX_MALLOC]: declare as nothing; remove #ifdefs from around calls. * regex.c (extract_number, extract_number_and_incr): declare static. * regex.c: remove the canned main program. * main.c: new file. * Makefile (COMMON): add main.o. Tue Sep 24 06:26:51 1991 Kathy Hargreaves (kathy at fosse) * regex.c (re_match_2): Made `pend' and `dend' not register variables. Only set string2 to string1 if string1 isn't null. Send address of p, d, regstart, regend, and reg_info to pop_failure_point. Put in more debug statements. * regex.c [debug]: Added global variable. (DEBUG_*PRINT*): Only print if `debug' is true. (DEBUG_DOUBLE_STRING_PRINTER): Changed DEBUG_STRING_PRINTER's name to this. Changed some comments. (PUSH_FAILURE_POINT): Moved and added some debugging statements. Was saving regstart on the stack twice instead of saving both regstart and regend; remedied this. [NUM_REGS_ITEMS]: Changed from 3 to 4, as now save lowest and highest active registers instead of highest used one. [NUM_NON_REG_ITEMS]: Changed name of NUM_OTHER_ITEMS to this. (NUM_FAILURE_ITEMS): Use active registers instead of number 0 through highest used one. (re_match_2): Have pop_failure_point put things in the variables. (pop_failure_point): Have it do what the fail case in re_match_2 did with the failure stack, instead of throwing away the stuff popped off. re_match_2 can ignore results when it doesn't need them. Thu Sep 5 13:23:28 1991 Kathy Hargreaves (kathy at fosse) * regex.c (banner): Changed copyright years to be separate. * regex.c [CHAR_UNSIGNED]: Put __ at both ends of this name. [DEBUG, debug_count, *debug_p, DEBUG_PRINT_1, DEBUG_PRINT_2, DEBUG_COMPILED_PATTERN_PRINTER ,DEBUG_STRING_PRINTER]: defined these for debugging. (extract_number): Added this (debuggable) routine version of the macro EXTRACT_NUMBER. Ditto for EXTRACT_NUMBER_AND_INCR. (re_compile_pattern): Set return_default_num_regs if the syntax bit RE_ALLOCATE_REGISTERS is set. [REGEX_MALLOC]: Renamed USE_ALLOCA to this. (BUF_POP): Got rid of this, as don't ever use it. (regex_compile): Made the type of `pattern' not be register. If DEBUG, print the pattern to compile. (re_match_2): If had a `$' in the pattern before a `^' then don't record the `^' as an anchor. Put (enum regexpcode) before references to b, as suggested [RE_NO_BK_BRACES]: Changed RE_NO_BK_CURLY_BRACES to this. (remove_pattern_offset): Removed this unused routine. (PUSH_FAILURE_POINT): Changed to only save active registers. Put in debugging statements. (re_compile_fastmap): Made `pattern' not a register variable. Use routine for extracting numbers instead of macro. (re_match_2): Made `p', `mcnt' and `mcnt2' not register variables. Added `num_regs_pushed' for debugging. Only malloc registers if the syntax bit RE_ALLOCATE_REGISTERS is set. Put in debug statements. Put the macro NOTE_INNER_GROUP's code inline, as it was the only called in one place. For debugging, extract numbers using routines instead of macros. In case fail: only restore pushed active registers, and added debugging statements. (pop_failure_point): Test for underfull stack. (group_can_match_nothing, common_op_can_match_nothing): For debugging, extract numbers using routines instead of macros. (regexec): Changed formal parameters to not be prototypes. Don't initialize `regs' or `private_preg' in their declarations. Tue Jul 23 18:38:36 1991 Kathy Hargreaves (kathy at hayley) * regex.h [RE_CONTEX_INDEP_OPS]: Moved the anchor stuff out of this bit. [RE_UNMATCHED_RIGHT_PAREN_ORD]: Defined this bit. [RE_CONTEXT_INVALID_ANCHORS]: Defined this bit. [RE_CONTEXT_INDEP_ANCHORS]: Defined this bit. Added RE_CONTEXT_INDEP_ANCHORS to all syntaxes which had RE_CONTEXT_INDEP_OPS. Took RE_ANCHORS_ONLY_AT_ENDS out of the POSIX basic syntax. Added RE_UNMATCHED_RIGHT_PAREN_ORD to the POSIX extended syntax. Took RE_REPEATED_ANCHORS_AWAY out of the POSIX extended syntax. Defined REG_NOERROR (which will probably have to go away again). Changed the type `off_t' to `regoff_t'. * regex.c: Changed some commments. (regex_compile): Added variable `had_an_endline' to keep track of if hit a `$' since the beginning of the pattern or the last alternative (if any). Changed RE_CONTEXT_INVALID_OPS and RE_CONTEXT_INDEP_OPS to RE_CONTEXT_INVALID_ANCHORS and RE_CONTEXT_INDEP_ANCHORS where appropriate. Put a `no_op' in the pattern if a repeat is only zero or one times; in this case and if it is many times (whereupon a jump backwards is pushed instead), keep track of the operator for verify_and_adjust_endlines. If RE_UNMATCHED_RIGHT_PAREN is set, make an unmatched close-group operator match `)'. Changed all error exits to exit (1). (remove_pattern_offset): Added this routine, but don't use it. (verify_and_adjust_endlines): At top of routine, if initialize routines run out of memory, return true after setting enough_memory false. At end of endline, et al. case, don't set *p to no_op. Repetition operators also set the level and active groups' match statuses, unless RE_REPEATED_ANCHORS_AWAY is set. (get_group_match_status): Put a return in front of call to get_bit. (re_compile_fastmap): Changed is_a_succeed_n to a boolean. If at end of pattern, then if the failure stack isn't empty, go back to the failure point. In *jump* case, only pop the stack if what's on top of it is where we've just jumped to. (re_search_2): Return -2 instead of val if val is -2. (group_can_match_nothing, alternative_can_match_nothing, common_op_can-match_nothing): Now pass in reg_info for the `duplicate' case. (re_match_2): Don't skip over the next alternative also if empty alternatives aren't allowed. In fail case, if failed to a backwards jump that's part of a repetition loop, pop the current failure point and use the next one. (pop_failure_point): Check that there's as many register items on the failure stack as the stack says there are. (common_op_can_match_nothing): Added variables `ret' and `reg_no' so can set reg_info for the group encountered. Also break without doing anything if hit a no_op or the other kinds of `endline's. If not done already, set reg_info in start_memory case. Put in no_pop_jump for an optimized succeed_n of zero repetitions. In succeed_n case, if the number isn't zero, then return false. Added `duplicate' case. Sat Jul 13 11:27:38 1991 Kathy Hargreaves (kathy at hayley) * regex.h (REG_NOERROR): Added this error code definition. * regex.c: Took some redundant parens out of macros. (enum regexpcode): Added jump_past_next_alt. Wrapped some macros in `do..while (0)'. Changed some comments. (regex_compile): Use `fixup_alt_jump' instead of `fixup_jump'. Use `maybe_pop_jump' instead of `maybe_pop_failure_jump'. Use `jump_past_next_alt' instead of `no_pop_jump' when at the end of an alternative. (re_match_2): Used REGEX_ALLOCATE for the registers stuff. In stop_memory case: Add more boolean tests to see if the group is in a loop. Added jump_past_next_alt case, which doesn't jump over the next alternative if the last one didn't match anything. Unfortunately, to make this work with, e.g., `(a+?*|b)*' against `bb', I also had to pop the alternative's failure point, which in turn broke backtracking! In fail case: Detect a dummy failure point by looking at failure_stack.avail - 2, not stack[-2]. (pop_failure_point): Only pop if the stack isn't empty; don't give an error if it is. (Not sure yet this is correct.) (group_can_match_nothing): Make it return a boolean instead of int. Make it take an argument indicating the end of where it should look. If find a group that can match nothing, set the pointer argument to past the group in the pattern. Took out cases which can share with alternative_can_match_nothing and call common_op_can_match_nothing. Took ++ out of switch, so could call common_op_can_match_nothing. Wrote lots more for on_failure_jump case to handle alternatives. Main loop now doesn't look for matching stop_memory, but rather the argument END; return true if hit the matching stop_memory; this way can call itself for inner groups. (alternative_can_match_nothing): Added for alternatives. (common_op_can_match_nothing): Added for previous two routines' common operators. (regerror): Returns a message saying there's no error if gets sent REG_NOERROR. Wed Jul 3 10:43:15 1991 Kathy Hargreaves (kathy at hayley) * regex.c: Removed unnecessary enclosing parens from several macros. Put `do..while (0)' around a few. Corrected some comments. (INIT_FAILURE_STACK_SIZE): Deleted in favor of using INIT_FAILURE_ALLOC. (INIT_FAILURE_STACK, DOUBLE_FAILURE_STACK, PUSH_PATTERN_OP, PUSH_FAILURE_POINT): Made routines of the same name (but with all lowercase letters) into these macros, so could use `alloca' when USE_ALLOCA is defined. The reason is stated below for bits lists. Deleted analogous routines. (re_compile_fastmap): Added variable void *destination for PUSH_PATTERN_OP. (re_match_2): Added variable void *destination for REGEX_REALLOCATE. Used the failure stack macros in place of the routines. Detected a dummy failure point by inspecting the failure stack's (avail - 2)th element, not failure_stack.stack[-2]. This bug arose when used the failure stack macros instead of the routines. * regex.c [USE_ALLOCA]: Put this conditional around previous alloca stuff and defined these to work differently depending on whether or not USE_ALLOCA is defined: (REGEX_ALLOCATE): Uses either `alloca' or `malloc'. (REGEX_REALLOCATE): Uses either `alloca' or `realloc'. (INIT_BITS_LIST, EXTEND_BITS_LIST, SET_BIT_TO_VALUE): Defined macro versions of routines with the same name (only with all lowercase letters) so could use `alloc' in re_match_2. This is to prevent core leaks when C-g is used in Emacs and to make things faster and avoid storage fragmentation. These things have to be macros because the results of `alloca' go away with the routine by which it's called. (BITS_BLOCK_SIZE, BITS_BLOCK, BITS_MASK): Moved to above the above-mentioned macros instead of before the routines defined below regex_compile. (set_bit_to_value): Compacted some code. (reg_info_type): Changed inner_groups field to be bits_list_type so could be arbitrarily long and thus handle arbitrary nesting. (NOTE_INNER_GROUP): Put `do...while (0)' around it so could use as a statement. Changed code to use bits lists. Added variable void *destination for REGEX_REALLOCATE (whose call is several levels in). Changed variable name of `this_bit' to `this_reg'. (FREE_VARIABLES): Only define and use if USE_ALLOCA is defined. (re_match_2): Use REGEX_ALLOCATE instead of malloc. Instead of setting INNER_GROUPS of reg_info to zero, have to use INIT_BITS_LIST and return -2 (and free variables if USE_ALLOCA isn't defined) if it fails. Fri Jun 28 13:45:07 1991 Karl Berry (karl at hayley) * regex.c (re_match_2): set value of `dend' when we restore `d'. * regex.c: remove declaration of alloca. * regex.c (MISSING_ISGRAPH): rename to `ISGRAPH_MISSING'. * regex.h [_POSIX_SOURCE]: remove these conditionals; always define POSIX stuff. * regex.c (_POSIX_SOURCE): change conditionals to use `POSIX' instead. Sat Jun 1 16:56:50 1991 Kathy Hargreaves (kathy at hayley) * regex.*: Changed RE_CONTEXTUAL_* to RE_CONTEXT_*, RE_TIGHT_VBAR to RE_TIGHT_ALT, RE_NEWLINE_OR to RE_NEWLINE_ALT, and RE_DOT_MATCHES_NEWLINE to RE_DOT_NEWLINE. Wed May 29 09:24:11 1991 Karl Berry (karl at hayley) * regex.texinfo (POSIX Pattern Buffers): cross-reference the correct node name (Match-beginning-of-line, not ..._line). (Syntax Bits): put @code around all syntax bits. Sat May 18 16:29:58 1991 Karl Berry (karl at hayley) * regex.c (global): add casts to keep broken compilers from complaining about malloc and realloc calls. * regex.c (isgraph) [MISSING_ISGRAPH]: change test to this, instead of `#ifndef isgraph', since broken compilers can't have both a macro and a symbol by the same name. * regex.c (re_comp, re_exec) [_POSIX_SOURCE]: do not define. (regcomp, regfree, regexec, regerror) [_POSIX_SOURCE && !emacs]: only define in this case. Mon May 6 17:37:04 1991 Kathy Hargreaves (kathy at hayley) * regex.h (re_search, re_search_2): Changed BUFFER to not be const. * regex.c (re_compile_pattern): `^' is in a leading position if it precedes a newline. (various routines): Added or changed header comments. (double_pattern_offsets_list): Changed name from `extend_pattern_offsets_list'. (adjust_pattern_offsets_list): Changed return value from unsigned to void. (verify_and_adjust_endlines): Now returns `true' and `false' instead of 1 and 0. `$' is in a leading position if it follows a newline. (set_bit_to_value, get_bit_value): Exit with error if POSITION < 0 so now calling routines don't have to. (init_failure_stack, inspect_failure_stack_top, pop_failure_stack_top, push_pattern_op, double_failure_stack): Now return value unsigned instead of boolean. (re_search, re_search_2): Changed BUFP to not be const. (re_search_2): Added variable const `private_bufp' to send to re_match_2. (push_failure_point): Made return value unsigned instead of boolean. Sat May 4 15:32:22 1991 Kathy Hargreaves (kathy at hayley) * regex.h (re_compile_fastmap): Added extern for this. Changed some comments. * regex.c (re_compile_pattern): In case handle_bar: put invalid pattern test before levels matching stuff. Changed some commments. Added optimizing test for detecting an empty alternative that ends with a trailing '$' at the end of the pattern. (re_compile_fastmap): Moved failure_stack stuff to before this so could use it. Made its stack dynamic. Made it return an int so that it could return -2 if its stack couldn't be allocated. Added to header comment (about the return values). (init_failure_stack): Wrote so both re_match_2 and re_compile_fastmap could use it similar stacks. (double_failure_stack): Added for above reasons. (push_pattern_op): Wrote for re_compile_fastmap. (re_search_2): Now return -2 if re_compile_fastmap does. (re_match_2): Made regstart and regend type failure_stack_element*. (push_failure_point): Made pattern_place and string_place type failure_stack_element*. Call double_failure_stack now. Return true instead of 1. Wed May 1 12:57:21 1991 Kathy Hargreaves (kathy at hayley) * regex.c (remove_intervening_anchors): Avoid erroneously making ops into no_op's by making them no_op only when they're beglines. (verify_and_adjust_endlines): Don't make '$' a normal character if it's before a newline. Look for the endline op in *p, not p[1]. (failure_stack_element): Added this declaration. (failure_stack_type): Added this declaration. (INIT_FAILURE_STACK_SIZE, FAILURE_STACK_EMPTY, FAILURE_STACK_PTR_EMPTY, REMAINING_AVAIL_SLOTS): Added for failure stack. (FAILURE_ITEM_SIZE, PUSH_FAILURE_POINT): Deleted. (FREE_VARIABLES): Now free failure_stack.stack instead of stackb. (re_match_2): deleted variables `initial_stack', `stackb', `stackp', and `stacke' and added `failure_stack' to replace them. Replaced calls to PUSH_FAILURE_POINT with those to push_failure_point. (push_failure_point): Added for re_match_2. (pop_failure_point): Rewrote to use a failure_stack_type of stack. (can_match_nothing): Moved definition to below re_match_2. (bcmp_translate): Moved definition to below re_match_2. Mon Apr 29 14:20:54 1991 Kathy Hargreaves (kathy at hayley) * regex.c (enum regexpcode): Added codes endline_before_newline and repeated_endline_before_newline so could detect these types of endlines in the intermediate stages of a compiled pattern. (INIT_FAILURE_ALLOC): Renamed NFAILURES to this and set it to 5. (BUF_PUSH): Put `do {...} while 0' around this. (BUF_PUSH_2): Defined this to cut down on expansion of EXTEND_BUFFER. (regex_compile): Changed some comments. Now push endline_before_newline if find a `$' before a newline in the pattern. If a `$' might turn into an ordinary character, set laststart to point to it. In '^' case, if syntax bit RE_TIGHT_VBAR is set, then for `^' to be in a leading position, it must be first in the pattern. Don't have to check in one of the else clauses that it's not set. If RE_CONTEXTUAL_INDEP_OPS isn't set but RE_ANCHORS_ONLY_AT_ENDS is, make '^' a normal character if it isn't first in the pattern. Can only detect at the end if a '$' after an alternation op is a trailing one, so can't immediately detect empty alternatives if a '$' follows a vbar. Added a picture of the ``success jumps'' in alternatives. Have to set bufp->used before calling verify_and_adjust_endlines. Also do it before returning all error strings. (remove_intervening_anchors): Now replaces the anchor with repeated_endline_before_newline if it's an endline_before_newline. (verify_and_adjust_endlines): Deleted SYNTAX parameter (could use bufp's) and added GROUP_FORWARD_MATCH_STATUS so could detect back references referring to empty groups. Added variable `bend' to point past the end of the pattern buffer. Added variable `previous_p' so wouldn't have to reinspect the pattern buffer to see what op we just looked at. Added endline_before_newline and repeated_endline_before_newline cases. When checking if in a trailing position, added case where '$' has to be at the pattern's end if either of the syntax bits RE_ANCHORS_ONLY_AT_ENDS or RE_TIGHT_VBAR are set. Since `endline' can have the intermediate form `endline_in_repeat', have to change it to `endline' if RE_REPEATED_ANCHORS_AWAY isn't set. Now disallow empty alternatives with trailing endlines in them if RE_NO_EMPTY_ALTS is set. Now don't make '$' an ordinary character if it precedes a newline. Don't make it an ordinary character if it's before a newline. Back references now affect the level matching something only if they refer to nonempty groups. (can_match_nothing): Now increment p1 in the switch, which changes many of the cases, but makes the code more like what it was derived from. Adjust the return statement to reflect above. (struct register_info): Made `can_match_nothing' field an int instead of a bit so could have -1 in it if never set. (MAX_FAILURE_ITEMS): Changed name from MAX_NUM_FAILURE_ITEMS. (FAILURE_ITEM_SIZE): Defined how much space a failure items uses. (PUSH_FAILURE_POINT): Changed variable `last_used_reg's name to `highest_used_reg'. Added variable `num_stack_items' and changed `len's name to `stack_length'. Test failure stack limit in terms of number of items in it, not in terms of its length. rms' fix tested length against number of items, which was a misunderstanding. Use `realloc' instead of `alloca' to extend the failure stack. Use shifts instead of multiplying by 2. (FREE_VARIABLES): Free `stackb' instead of `initial_stack', as might may have been reallocated. (re_match_2): When mallocing `initial_stack', now multiply the number of items wanted (what was there before) by FAILURE_ITEM_SIZE. (pop_failure_point): Need this procedure form of the macro of the same name for debugging, so left it in and deleted the macro. (recomp): Don't free the pattern buffer's translate field. Mon Apr 15 09:47:47 1991 Kathy Hargreaves (kathy at hayley) * regex.h (RE_DUP_MAX): Moved to outside of #ifdef _POSIX_SOURCE. * regex.c (#include ): Removed #ifdef _POSIX_SOURCE condition. (malloc, realloc): Made return type void* #ifdef __STDC__. (enum regexpcode): Added endline_in_repeat for the compiler's use; this never ends up on the final compiled pattern. (INIT_PATTERN_OFFSETS_LIST_SIZE): Initial size for pattern_offsets_list_type. (pattern_offset_type): Type for pattern offsets. (pattern_offsets_list_type): Type for keeping a list of pattern offsets. (anchor_list_type): Changed to above type. (PATTERN_OFFSETS_LIST_PTR_FULL): Tests if a pattern offsets list is full. (ANCHOR_LIST_PTR_FULL): Changed to above. (BIT_BLOCK_SIZE): Changed to BITS_BLOCK_SIZE and moved to above bits list routines below regex_compile. (op_list_type): Defined to be pattern_offsets_list_type. (compile_stack_type): Changed offsets to be pattern_offset_type instead of unsigned. (pointer): Changed the name of all structure fields from this to `avail'. (COMPILE_STACK_FULL): Changed so the stack is full if `avail' is equal to `size' instead of `size' - 1. (GET_BUFFER_SPACE): Changed `>=' to `>' in the while statement. (regex_compile): Added variable `enough_memory' so could check that routine that verifies '$' positions could return an allocation error. (group_count): Deleted this variable, as `regnum' already does this work. (op_list): Added this variable to keep track of operations needed for verifying '$' positions. (anchor_list): Now initialize using routine `init_pattern_offsets_list'. Consolidated the three bits_list initializations. In case '$': Instead of trying to go past constructs which can follow '$', merely detect the special case where it has to be at the pattern's end, fix up any fixup jumps if necessary, record the anchor if necessary and add an `endline' (and possibly two `no-op's) to the pattern; will call a routine at the end to verify if it's in a valid position or not. (init_pattern_offsets_list): Added to initialize pattern offsets lists. (extend_anchor_list): Renamed this extend_pattern_offsets_list and renamed parameters and internal variables appropriately. (add_pattern_offset): Added this routine which both record_anchor_position and add_op call. (adjust_pattern_offsets_list): Add this routine to adjust by some increment all the pattern offsets a list of such after a given position. (record_anchor_position): Now send in offset instead of calculating it and just call add_pattern_offset. (adjust_anchor_list): Replaced by above routine. (remove_intervening_anchors): If the anchor is an `endline' then replace it with `endline_in_repeat' instead of `no_op'. (add_op): Added this routine to call in regex_compile wherever push something relevant to verifying '$' positions. (verify_and_adjust_endlines): Added routine to (1) verify that '$'s in a pattern buffer (represented by `endline') were in valid positions and (2) whether or not they were anchors. (BITS_BLOCK_SIZE): Renamed BIT_BLOCK_SIZE and moved to right above bits list routines. (BITS_BLOCK): Defines which array element of a bits list the bit corresponding to a given position is in. (BITS_MASK): Has a 1 where the bit (in a bit list array element) for a given position is. Mon Apr 1 12:09:06 1991 Kathy Hargreaves (kathy at hayley) * regex.c (BIT_BLOCK_SIZE): Defined this for using with bits_list_type, abstracted from level_list_type so could use for more things than just the level match status. (regex_compile): Renamed `level_list' variable to `level_match_status'. Added variable `group_match_status' of type bits_list_type. Kept track of whether or not for all groups any of them matched other than the empty string, so detect if a back reference in front of a '^' made it nonleading or not. Do this by setting a match status bit for all active groups whenever leave a group that matches other than the empty string. Could detect which groups are active by going through the stack each time, but or-ing a bits list of active groups with a bits list of group match status is faster, so make a bits list of active groups instead. Have to check that '^' isn't in a leading position before going to normal_char. Whenever set level match status of the current level, also set the match status of all active groups. Increase the group count and make that group active whenever open a group. When close a group, only set the next level down if the current level matches other than the empty string, and make the current group inactive. At a back reference, only set a level's match status if the group to which the back reference refers matches other than the empty string. (init_bits_list): Added to initialize a bits list. (get_level_value): Deleted this. (Made into get_level_match_status.) (extend_bits_list): Added to extend a bits list. (Made this from deleted routine `extend_level_list'.) (get_bit): Added to get a bit value from a bits list. (Made this from deleted routine `get_level_value'.) (set_bit_to_value): Added to set a bit in a bits list. (Made this from deleted routine `set_level_value'.) (get_level_match_status): Added this to get the match status of a given level. (Made from get_level_value.) (set_this_level, set_next_lower_level): Made all routines which set bits extend the bits list if necessary, thus they now return an unsigned value to indicate whether or not the reallocation failed. (increase_level): No longer extends the level list. (make_group_active): Added to mark as active a given group in an active groups list. (make_group_inactive): Added to mark as inactive a given group in an active groups list. (set_match_status_of_active_groups): Added to set the match status of all currently active groups. (get_group_match_status): Added to get a given group's match status. (no_levels_match_anything): Removed the paramenter LEVEL. (PUSH_FAILURE_POINT): Added rms' bug fix and changed RE_NREGS to num_internal_regs. Sun Mar 31 09:04:30 1991 Kathy Hargreaves (kathy at hayley) * regex.h (RE_ANCHORS_ONLY_AT_ENDS): Added syntax so could constrain '^' and '$' to only be anchors if at the beginning and end of the pattern. (RE_SYNTAX_POSIX_BASIC): Added the above bit. * regex.c (enum regexcode): Changed `unused' to `no_op'. (this_and_lower_levels_match_nothing): Deleted forward reference. (regex_compile): case '^': if the syntax bit RE_ANCHORS_ONLY_AT_ENDS is set, then '^' is only an anchor if at the beginning of the pattern; only record anchor position if the syntax bit RE_REPEATED_ANCHORS_AWAY is set; the '^' is a normal char if the syntax bit RE_ANCHORS_ONLY_AT_END is set and we're not at the beginning of the pattern (and neither RE_CONTEXTUAL_INDEP_OPS nor RE_CONTEXTUAL_INDEP_OPS syntax bits are set). Only adjust the anchor list if the syntax bit RE_REPEATED_ANCHORS_AWAY is set. * regex.c (level_list_type): Use to detect when '^' is in a leading position. (regex_compile): Added level_list_type level_list variable in which we keep track of whether or not a grouping level (in its current or most recent incarnation) matches anything besides the empty string. Set the bit for the i-th level when detect it should match something other than the empty string and the bit for the (i-1)-th level when leave the i-th group. Clear all bits for the i-th and higher levels if none of 0--(i - 1)-th's bits are set when encounter an alternation operator on that level. If no levels are set when hit a '^', then it is in a leading position. We keep track of which level we're at by increasing a variable current_level whenever we encounter an open-group operator and decreasing it whenever we encounter a close-group operator. Have to adjust the anchor list contents whenever insert something ahead of them (such as on_failure_jump's) in the pattern. (adjust_anchor_list): Adjusts the offsets in an anchor list by a given increment starting at a given start position. (get_level_value): Returns the bit setting of a given level. (set_level_value): Sets the bit of a given level to a given value. (set_this_level): Sets (to 1) the bit of a given level. (set_next_lower_level): Sets (to 1) the bit of (LEVEL - 1) for a given LEVEL. (clear_this_and_higher_levels): Clears the bits for a given level and any higher levels. (extend_level_list): Adds sizeof(unsigned) more bits to a level list. (increase_level): Increases by 1 the value of a given level variable. (decrease_level): Decreases by 1 the value of a given level variable. (lower_levels_match_nothing): Checks if any levels lower than the given one match anything. (no_levels_match_anything): Checks if any levels match anything. (re_match_2): At case wordbeg: before looking at d-1, check that we're not at the string's beginning. At case wordend: Added some illuminating parentheses. Mon Mar 25 13:58:51 1991 Kathy Hargreaves (kathy at hayley) * regex.h (RE_NO_ANCHOR_AT_NEWLINE): Changed syntax bit name from RE_ANCHOR_NOT_NEWLINE because an anchor never matches the newline itself, just the empty string either before or after it. (RE_REPEATED_ANCHORS_AWAY): Added this syntax bit for ignoring anchors inside groups which are operated on by repetition operators. (RE_DOT_MATCHES_NEWLINE): Added this bit so the match-any-character operator could match a newline when it's set. (RE_SYNTAX_POSIX_BASIC): Set RE_DOT_MATCHES_NEWLINE in this. (RE_SYNTAX_POSIX_EXTENDED): Set RE_DOT_MATCHES_NEWLINE and RE_REPEATED_ANCHORS_AWAY in this. (regerror): Changed prototypes to new POSIX spec. * regex.c (anchor_list_type): Added so could null out anchors inside repeated groups. (ANCHOR_LIST_PTR_FULL): Added for above type. (compile_stack_element): Changed name from stack_element. (compile_stack_type): Changed name from compile_stack. (INIT_COMPILE_STACK_SIZE): Changed name from INIT_STACK_SIZE. (COMPILE_STACK_EMPTY): Changed name from STACK_EMPTY. (COMPILE_STACK_FULL): Changed name from STACK_FULL. (regex_compile): Changed SYNTAX parameter to non-const. Changed variable name `stack' to `compile_stack'. If syntax bit RE_REPEATED_ANCHORS_AWAY is set, then naively put anchors in a list when encounter them and then set them to `unused' when detect they are within a group operated on by a repetition operator. Need something more sophisticated than this, as they should only get set to `unused' if they are in positions where they would be anchors. Also need a better way to detect contextually invalid anchors. Changed some commments. (is_in_compile_stack): Changed name from `is_in_stack'. (extend_anchor_list): Added to do anchor stuff. (record_anchor_position): Added to do anchor stuff. (remove_intervening_anchors): Added to do anchor stuff. (re_match_2): Now match a newline with the match-any-character operator if RE_DOT_MATCHES_NEWLINE is set. Compacted some code. (regcomp): Added new POSIX newline information to the header commment. If REG_NEWLINE cflag is set, then now unset RE_DOT_MATCHES_NEWLINE in syntax. (put_in_buffer): Added to do new POSIX regerror spec. Called by regerror. (regerror): Changed to take a pattern buffer, error buffer and its size, and return type `size_t', the size of the full error message, and the first ERRBUF_SIZE - 1 characters of the full error message in the error buffer. Wed Feb 27 16:38:33 1991 Kathy Hargreaves (kathy at hayley) * regex.h (#include ): Removed this as new POSIX standard has the user include it. (RE_SYNTAX_POSIX_BASIC and RE_SYNTAX_POSIX_EXTENDED): Removed RE_HAT_LISTS_NOT_NEWLINE as new POSIX standard has the cflag REG_NEWLINE now set this. Similarly, added syntax bit RE_ANCHOR_NOT_NEWLINE as this is now unset by REG_NEWLINE. (RE_SYNTAX_POSIX_BASIC): Removed syntax bit RE_NO_CONSECUTIVE_REPEATS as POSIX now allows them. * regex.c (#include ): Added this as new POSIX standard has the user include it instead of us putting it in regex.h. (extern char *re_syntax_table): Made into an extern so the user could allocate it. (DO_RANGE): If don't find a range end, now goto invalid_range_end instead of unmatched_left_bracket. (regex_compile): Made variable SYNTAX non-const.???? Reformatted some code. (re_compile_fastmap): Moved is_a_succeed_n's declaration to inner braces. Compacted some code. (SET_NEWLINE_FLAG): Removed and put inline. (regcomp): Made variable `syntax' non-const so can unset RE_ANCHOR_NOT_NEWLINE syntax bit if cflag RE_NEWLINE is set. If cflag RE_NEWLINE is set, set the RE_HAT_LISTS_NOT_NEWLINE syntax bit and unset RE_ANCHOR_NOT_NEWLINE one of `syntax'. Wed Feb 20 16:33:38 1991 Kathy Hargreaves (kathy at hayley) * regex.h (RE_NO_CONSECUTIVE_REPEATS): Changed name from RE_NO_CONSEC_REPEATS. (REG_ENESTING): Deleted this POSIX return value, as the stack is now unbounded. (struct re_pattern_buffer): Changed some comments. (re_compile_pattern): Changed a comment. Deleted check on stack upper bound and corresponding error. Now when there's no interval contents and it's the end of the pattern, go to unmatched_left_curly_brace instead of end_of_pattern. Removed nesting_too_deep error, as the stack is now unbounded. (regcomp): Removed REG_ENESTING case, as the stack is now unbounded. (regerror): Removed REG_ENESTING case, as the stack is now unbounded. * regex.c (MAX_STACK_SIZE): Deleted because don't need upper bound on array indexed with an unsigned number. Sun Feb 17 15:50:24 1991 Kathy Hargreaves (kathy at hayley) * regex.h: Changed and added some comments. * regex.c (init_syntax_once): Made `_' a word character. (re_compile_pattern): Added a comment. (re_match_2): Redid header comment. (regexec): With header comment about PMATCH, corrected and removed details found regex.h, adding a reference. Fri Feb 15 09:21:31 1991 Kathy Hargreaves (kathy at hayley) * regex.c (DO_RANGE): Removed argument parentheses. Now get untranslated range start and end characters and set list bits for the translated (if at all) versions of them and all characters between them. (re_match_2): Now use regs->num_regs instead of num_regs_wanted wherever possible. (regcomp): Now build case-fold translate table using isupper and tolower facilities so will work on foreign language characters. Sat Feb 9 16:40:03 1991 Kathy Hargreaves (kathy at hayley) * regex.h (RE_HAT_LISTS_NOT_NEWLINE): Changed syntax bit name from RE_LISTS_NOT_NEWLINE as it only affects nonmatching lists. Changed all references to the match-beginning-of-string operator to match-beginning-of-line operator, as this is what it does. (RE_NO_CONSEC_REPEATS): Added this syntax bit. (RE_SYNTAX_POSIX_BASIC): Added above bit to this. (REG_PREMATURE_END): Changed name to REG_EEND. (REG_EXCESS_NESTING): Changed name to REG_ENESTING. (REG_TOO_BIG): Changed name to REG_ESIZE. (REG_INVALID_PREV_RE): Deleted this return POSIX value. Added and changed some comments. * regex.c (re_compile_pattern): Now sets the pattern buffer's `return_default_num_regs' field. (typedef struct stack_element, stack_type, INIT_STACK_SIZE, MAX_STACK_SIZE, STACK_EMPTY, STACK_FULL): Added for regex_compile. (INIT_BUF_SIZE): Changed value from 28 to 32. (BUF_PUSH): Changed name from BUFPUSH. (MAX_BUF_SIZE): Added so could use in many places. (IS_CHAR_CLASS_STRING): Replaced is_char_class with this. (regex_compile): Added a stack which could grow dynamically and which has struct elements. Go back to initializing `zero_times_ok' and `many_time_ok' to 0 and |=ing them inside the loop. Now disallow consecutive repetition operators if the syntax bit RE_NO_CONSEC_REPEATS is set. Now detect trailing backslash when the compiler is expecting a `?' or a `+'. Changed calls to GET_BUFFER_SPACE which asked for 6 to ask for 3, as that's all they needed. Now check for trailing backslash inside lists. Now disallow an empty alternative right before an end-of-line operator. Now get buffer space before leaving space for a fixup jump. Now check if at pattern end when at open-interval operator. Added some comments. Now check if non-interval repetition operators follow an interval one if the syntax bit RE_NO_CONSEC_REPEATS is set. Now only check if what precedes an interval repetition operator isn't a regular expression which matches one character if the syntax bit RE_NO_CONSEC_REPEATS is set. Now return "Unmatched [ or [^" instead of "Unmatched [". (is_in_stack): Added to check if a given register number is in the stack. (re_match_2): If initial variable allocations fail, return -2, instead of -1. Now set reg's `num_regs' field when allocating regs. Now before allocating them, free regs->start and end if they aren't NULL and return -2 if either allocation fails. Now use regs->num_regs instead of num_regs_wanted to control regs loops. Now increment past the newline when matching it with an end-of-line operator. (recomp): Added to the header comment. Now return REG_ESUBREG if regex_compile returns "Unmatched [ or [^" instead of doing so if it returns "Unmatched [". Now return REG_BADRPT if in addition to returning "Missing preceding regular expression", regex_compile returns "Invalid preceding regular expression". Now return new return value names (see regex.h changes). (regexec): Added to header comment. Initialize regs structure. Now match whole string. Now always free regs.start and regs.end instead of just when the string matched. (regerror): Now return "Regex error: Unmatched [ or [^.\n" instead of "Regex error: Unmatched [.\n". Now return "Regex error: Preceding regular expression either missing or not simple.\n" instead of "Regex error: Missing preceding regular expression.\n". Removed REG_INVALID_PREV_RE case (it got subsumed into the REG_BADRPT case). Thu Jan 17 09:52:35 1991 Kathy Hargreaves (kathy at hayley) * regex.h: Changed a comment. * regex.c: Changed and added large header comments. (re_compile_pattern): Now if detect that `laststart' for an interval points to a byte code for a regular expression which matches more than one character, make it an internal error. (regerror): Return error message, don't print it. Tue Jan 15 15:32:49 1991 Kathy Hargreaves (kathy at hayley) * regex.h (regcomp return codes): Added GNU ones. Updated some comments. * regex.c (DO_RANGE): Changed `obscure_syntax' to `syntax'. (regex_compile): Added `following_left_brace' to keep track of where pseudo interval following a valid interval starts. Changed some instances that returned "Invalid regular expression" to instead return error strings coinciding with POSIX error codes. Changed some comments. Now consider only things between `[:' and `:]' to be possible character class names. Now a character class expression can't end a pattern; at least a `]' must close the list. Now if the syntax bit RE_NO_BK_CURLY_BRACES is set, then a valid interval must be followed by yet another to get an error for preceding an interval (in this case, the second one) with a regular expression that matches more than one character. Now if what follows a valid interval begins with a open interval operator but doesn't begin a valid interval, then set following_left_bracket to it, put it in C and go to normal_char label. Added some comments. Return "Invalid character class name" instead of "Invalid character class". (regerror): Return messages for all POSIX error codes except REG_ECOLLATE and REG_NEWLINE, along with all GNU error codes. Added `break's after all cases. (main): Call re_set_syntax instead of setting `obscure_syntax' directly. Sat Jan 12 13:37:59 1991 Kathy Hargreaves (kathy at hayley) * regex.h (Copyright): Updated date. (#include ): Include unconditionally. (RE_CANNOT_MATCH_NEWLINE): Deleted this syntax bit. (RE_SYNTAX_POSIX_BASIC, RE_SYNTAX_POSIX_EXTENDED): Removed setting the RE_ANCHOR_NOT_NEWLINE syntax bit from these. Changed and added some comments. (struct re_pattern_buffer): Changed some flags from chars to bits. Added field `syntax'; holds which syntax pattern was compiled with. Added bit flag `return_default_num_regs'. (externs for GNU and Berkeley UNIX routines): Added `const's to parameter types to be compatible with POSIX. (#define const): Added to support old C compilers. * regex.c (Copyright): Updated date. (enum regexpcode): Deleted `newline'. (regex_compile): Renamed re_compile_pattern to this, added a syntax parameter so it can set the pattern buffer's `syntax' field. Made `pattern', and `size' `const's so could pass to POSIX interface routines; also made `const' whatever interval variables had to be to make this work. Changed references to `obscure_syntax' to new parameter `syntax'. Deleted putting `newline' in buffer when see `\n'. Consider invalid character classes which have nothing wrong except the character class name; if so, return character-class error. (is_char_class): Added routine for regex_compile. (re_compile_pattern): added a new one which calls regex_compile with `obscure_syntax' as the actual parameter for the formal `syntax'. Gave this the old routine's header comments. Made `pattern', and `size' `const's so could use POSIX interface routine parameters. (re_search, re_search_2, re_match, re_match_2): Changed `pbufp' to `bufp'. (re_search_2, re_match_2): Changed `mstop' to `stop'. (re_search, re_search_2): Made all parameters except `regs' `const's so could use POSIX interface routines parameters. (re_search_2): Added private copies of `const' parameters so could change their values. (re_match_2): Made all parameters except `regs' `const's so could use POSIX interface routines parameters. Changed `size1' and `size2' parameters to `size1_arg' and `size2_arg' and so could change; added local `size1' and `size2' and set to these. Added some comments. Deleted `newline' case. `begline' can also possibly match if `d' contains a newline; if it does, we have to increment d to point past the newline. Replaced references to `obscure_syntax' with `bufp->syntax'. (re_comp, re_exec): Made parameter `s' a `const' so could use POSIX interface routines parameters. Now call regex_compile, passing `obscure_syntax' via the `syntax' parameter. (re_exec): Made local `len' a `const' so could pass to re_search. (regcomp): Added header comment. Added local `syntax' to set and pass to regex_compile rather than setting global `obscure_syntax' and passing it. Call regex_compile with its `syntax' parameter rather than re_compile_pattern. Return REG_ECTYPE if character-class error. (regexec): Don't initialize `regs' to anything. Made `private_preg' a nonpointer so could set to what the constant `preg' points. Initialize `private_preg's `return_default_num_regs' field to zero because want to return `nmatch' registers, not however many there are subexpressions in the pattern. Also test if `nmatch' > 0 to see if should pass re_match `regs'. Tue Jan 8 15:57:17 1991 Kathy Hargreaves (kathy at hayley) * regex.h (struct re_pattern_buffer): Reworded comment. * regex.c (EXTEND_BUFFER): Also reset beg_interval. (re_search_2): Return val if val = -2. (NUM_REG_ITEMS): Listed items in comment. (NUM_OTHER_ITEMS): Defined this for using in > 1 definition. (MAX_NUM_FAILURE_ITEMS): Replaced `+ 2' with NUM_OTHER_ITEMS. (NUM_FAILURE_ITEMS): As with definition above and added to comment. (PUSH_FAILURE_POINT): Replaced `* 2's with `<< 1's. (re_match_2): Test with equality with 1 to see pbufp->bol and pbufp->eol are set. Fri Jan 4 15:07:22 1991 Kathy Hargreaves (kathy at hayley) * regex.h (struct re_pattern_buffer): Reordered some fields. Updated some comments. Added not_bol and not_eol fields. (extern regcomp, regexec, regerror): Added return types. (extern regfree): Added `extern'. * regex.c (min): Deleted unused macro. (re_match_2): Compacted some code. Removed call to macro `min' from `for' loop. Fixed so unused registers get filled with -1's. Fail if the pattern buffer's `not_bol' field is set and encounter a `begline'. Fail if the pattern buffer's `not_eol' field is set and encounter a `endline'. Deleted redundant check for empty stack in fail case. Don't free pattern buffer's components in re_comp. (regexec): Initialize variable regs. Added `private_preg' pattern buffer so could set `not_bol' and `not_eol' fields and hand to re_match. Deleted naive attempt to detect anchors. Set private pattern buffer's `not_bol' and `not_eol' fields according to eflags value. `nmatch' must also be > 0 for us to bother allocating registers to send to re_match and filling pmatch with their results after the call to re_match. Send private pattern buffer instead of argument to re_match. If use the registers, always free them and then set them to NULL. (regerror): Added this Posix routine. (regfree): Added this Posix routine. Tue Jan 1 15:02:45 1991 Kathy Hargreaves (kathy at hayley) * regex.h (RE_NREGS): Deleted this definition, as now the user can choose how many registers to have. (REG_NOTBOL, REG_NOTEOL): Defined these Posix eflag bits. (REG_NOMATCH, REG_BADPAT, REG_ECOLLATE, REG_ECTYPE, REG_EESCAPE, REG_ESUBREG, REG_EBRACK, REG_EPAREN, REG_EBRACE, REG_BADBR, REG_ERANGE, REG_ESPACE, REG_BADRPT, REG_ENEWLINE): Defined these return values for Posix's regcomp and regexec. Updated some comments. (struct re_pattern_buffer): Now typedef this as regex_t instead of the other way around. (struct re_registers): Added num_regs field. Made start and end fields pointers to char instead of fixed size arrays. (regmatch_t): Added this Posix register type. (regcomp, regexec, regerror, regfree): Added externs for these Posix routines. * regex.c (enum boolean): Typedefed this. (re_pattern_buffer): Reformatted some comments. (re_compile_pattern): Updated some comments. Always push start_memory and its attendant number whenever encounter a group, not just when its number is less than the previous maximum number of registers; same for stop_memory. Get 4 bytes of buffer space instead of 2 when pushing a set_number_at. (can_match_nothing): Added this to elaborate on and replace code in re_match_2. (reg_info_type): Made can_match_nothing field a bit instead of int. (MIN): Added for re_match_2. (re_match_2 macros): Changed all `for' loops which used RE_NREGS to now use num_internal_regs as upper bounds. (MAX_NUM_FAILURE_ITEMS): Use num_internal_regs instead of RE_NREGS. (POP_FAILURE_POINT): Added check for empty stack. (FREE_VARIABLES): Added this to free (and set to NULL) variables allocated in re_match_2. (re_match_2): Rearranged parameters to be in order. Added variables num_regs_wanted (how many registers the user wants) and num_internal_regs (how many groups there are). Allocated initial_stack, regstart, regend, old_regstart, old_regend, reginfo, best_regstart, and best_regend---all which used to be fixed size arrays. Free them all and return -1 if any fail. Free above variables if starting position pos isn't valid. Changed all `for' loops which used RE_NREGS to now use num_internal_regs as upper bounds---except for the loops which fill regs; then use num_regs_wanted. Allocate regs if the user has passed it and wants more than 0 registers filled. Set regs->start[i] and regs->end[i] to -1 if either regstart[i] or regend[i] equals -1, not just the first. Free allocated variables before returning. Updated some comments. (regcomp): Return REG_ESPACE, REG_BADPAT, REG_EPAREN when appropriate. Free translate array. (regexec): Added this Posix interface routine. Mon Dec 24 14:21:13 1990 Kathy Hargreaves (kathy at hayley) * regex.h: If _POSIX_SOURCE is defined then #include . Added syntax bit RE_CANNOT_MATCH_NEWLINE. Defined Posix cflags: REG_EXTENDED, REG_NEWLINE, REG_ICASE, and REG_NOSUB. Added fields re_nsub and no_sub to struct re_pattern_buffer. Typedefed regex_t to be `struct re_pattern_buffer'. * regex.c (CHAR_SET_SIZE): Defined this to be 256 and replaced incidences of this value with this constant. (re_compile_pattern): Added switch case for `\n' and put `newline' into the pattern buffer when encounter this. Increment the pattern_buffer's `re_nsub' field whenever open a group. (re_match_2): Match a newline with `newline'---provided the syntax bit RE_CANNOT_MATCH_NEWLINE isn't set. (regcomp): Added this Posix interface routine. (enum test_type): Added interface_test tag. (main): Added Posix interface test. Tue Dec 18 12:58:12 1990 Kathy Hargreaves (kathy at hayley) * regex.h (struct re_pattern_buffer): reformatted so would fit in texinfo documentation. Thu Nov 29 15:49:16 1990 Kathy Hargreaves (kathy at hayley) * regex.h (RE_NO_EMPTY_ALTS): Added this bit. (RE_SYNTAX_POSIX_EXTENDED): Added above bit. * regex.c (re_compile_pattern): Disallow empty alternatives only when RE_NO_EMPTY_ALTS is set, not when RE_CONTEXTUAL_INVALID_OPS is. Changed RE_NO_BK_CURLY_BRACES to RE_NO_BK_PARENS when testing for empty groups at label handle_open. At label handle_bar: disallow empty alternatives if RE_NO_EMPTY_ALTS is set. Rewrote some comments. (re_compile_fastmap): cleaned up code. (re_search_2): Rewrote comment. (struct register_info): Added field `inner_groups'; it records which groups are inside of the current one. Added field can_match_nothing; it's set if the current group can match nothing. Added field ever_match_something; it's set if current group ever matched something. (INNER_GROUPS): Added macro to access inner_groups field of struct register_info. (CAN_MATCH_NOTHING): Added macro to access can_match_nothing field of struct register_info. (EVER_MATCHED_SOMETHING): Added macro to access ever_matched_something field of struct register_info. (NOTE_INNER_GROUP): Defined macro to record that a given group is inside of all currently active groups. (re_match_2): Added variables *p1 and mcnt2 (multipurpose). Added old_regstart and old_regend arrays to hold previous register values if they need be restored. Initialize added fields and variables. case start_memory: Find out if the group can match nothing. Save previous register values in old_restart and old_regend. Record that current group is inside of all currently active groups. If the group is inside a loop and it ever matched anything, restore its registers to values before the last failed match. Restore the registers for the inner groups, too. case duplicate: Can back reference to a group that never matched if it can match nothing. Thu Nov 29 11:12:54 1990 Karl Berry (karl at hayley) * regex.c (bcopy, ...): define these if either _POSIX_SOURCE or STDC_HEADERS is defined; same for including . Sat Oct 6 16:04:55 1990 Kathy Hargreaves (kathy at hayley) * regex.h (struct re_pattern_buffer): Changed field comments. * regex.c (re_compile_pattern): Allow a `$' to precede an alternation operator (`|' or `\|'). Disallow `^' and/or `$' in empty groups if the syntax bit RE_NO_EMPTY_GROUPS is set. Wait until have parsed a valid `\{...\}' interval expression before testing RE_CONTEXTUAL_INVALID_OPS to see if it's invalidated by that. Don't use RE_NO_BK_CURLY_BRACES to test whether or not a validly parsed interval expression is invalid if it has no preceding re; rather, use RE_CONTEXTUAL_INVALID_OPS. If an interval parses, but there is no preceding regular expression, yet the syntax bit RE_CONTEXTUAL_INDEP_OPS is set, then that interval can match the empty regular expression; if the bit isn't set, then the characters in the interval expression are parsed as themselves (sans the backslashes). In unfetch_interval case: Moved PATFETCH to above the test for RE_NO_BK_CURLY_BRACES being set, which would force a goto normal_backslash; the code at both normal_backsl and normal_char expect a character in `c.' Sun Sep 30 11:13:48 1990 Kathy Hargreaves (kathy at hayley) * regex.h: Changed some comments to use the terms used in the documentation. (RE_CONTEXTUAL_INDEP_OPS): Changed name from `RE_CONTEXT_INDEP_OPS'. (RE_LISTS_NOT_NEWLINE): Changed name from `RE_HAT_NOT_NEWLINE.' (RE_ANCHOR_NOT_NEWLINE): Added this syntax bit. (RE_NO_EMPTY_GROUPS): Added this syntax bit. (RE_NO_HYPHEN_RANGE_END): Deleted this syntax bit. (RE_SYNTAX_...): Reformatted. (RE_SYNTAX_POSIX_BASIC, RE_SYNTAX_EXTENDED): Added syntax bits RE_ANCHOR_NOT_NEWLINE and RE_NO_EMPTY_GROUPS, and deleted RE_NO_HYPHEN_RANGE_END. (RE_SYNTAX_POSIX_EXTENDED): Added syntax bit RE_DOT_NOT_NULL. * regex.c (bcopy, bcmp, bzero): Define if _POSIX_SOURCE is defined. (_POSIX_SOURCE): ifdef this, #include (#ifdef emacs): Changed comment of the #endif for the its #else clause to be `not emacs', not `emacs.' (no_pop_jump): Changed name from `jump'. (pop_failure_jump): Changed name from `finalize_jump.' (maybe_pop_failure_jump): Changed name from `maybe_finalize_jump'. (no_pop_jump_n): Changed name from `jump_n.' (EXTEND_BUFFER): Use shift instead of multiplication to double buf->allocated. (DO_RANGE, recompile_pattern): Added macro to set the list bits for a range. (re_compile_pattern): Fixed grammar problems in some comments. Checked that RE_NO_BK_VBAR is set to make `$' valid before a `|' and not set to make it valid before a `\|'. Checked that RE_NO_BK_PARENS is set to make `$' valid before a ')' and not set to make it valid before a `\)'. Disallow ranges starting with `-', unless the range is the first item in a list, rather than disallowing ranges which end with `-'. Disallow empty groups if the syntax bit RE_NO_EMPTY_GROUPS is set. Disallow nothing preceding `{' and `\{' if they represent the open-interval operator and RE_CONTEXTUAL_INVALID_OPS is set. (register_info_type): typedef-ed this using `struct register_info.' (SET_REGS_MATCHED): Compacted the code. (re_match_2): Made it fail if back reference a group which we've never matched. Made `^' not match a newline if the syntax bit RE_ANCHOR_NOT_NEWLINE is set. (really_fail): Added this label so could force a final fail that would not try to use the failure stack to recover. Sat Aug 25 14:23:01 1990 Kathy Hargreaves (kathy at hayley) * regex.h (RE_CONTEXTUAL_OPS): Changed name from RE_CONTEXT_OPS. (global): Rewrote comments and rebroke some syntax #define lines. * regex.c (isgraph): Added definition for sequents. (global): Now refer to character set lists as ``lists.'' Rewrote comments containing ``\('' or ``\)'' to now refer to ``groups.'' (RE_CONTEXTUAL_OPS): Changed name from RE_CONTEXT_OPS. (re_compile_pattern): Expanded header comment. Sun Jul 15 14:50:25 1990 Kathy Hargreaves (kathy at hayley) * regex.h (RE_CONTEX_INDEP_OPS): the comment's sense got turned around when we changed how it read; changed it to be correct. Sat Jul 14 16:38:06 1990 Kathy Hargreaves (kathy at hayley) * regex.h (RE_NO_EMPTY_BK_REF): changed name to RE_NO_MISSING_BK_REF, as this describes it better. * regex.c (re_compile_pattern): changed RE_NO_EMPTY_BK_REF to RE_NO_MISSING_BK_REF, as above. Thu Jul 12 11:45:05 1990 Kathy Hargreaves (kathy at hayley) * regex.h (RE_NO_EMPTY_BRACKETS): removed this syntax bit, as bracket expressions should *never* be empty regardless of the syntax. Removes this bit from RE_SYNTAX_POSIX_BASIC and RE_SYNTAX_POSIX_EXTENDED. * regex.c (SET_LIST_BIT): in the comment, now refer to character sets as (non)matching sets, as bracket expressions can now match other things in addition to characters. (re_compile_pattern): refer to groups as such instead of `\(...\)' or somesuch, because groups can now be enclosed in either plain parens or backslashed ones, depending on the syntax. In the '[' case, added a boolean just_had_a_char_class to detect whether or not a character class begins a range (which is invalid). Restore way of breaking out of a bracket expression to original way. Add way to detect a range if the last thing in a bracket expression was a character class. Took out check for c != ']' at the end of a character class in the else clause, as it had already been checked in the if part that also checked the validity of the string. Set or clear just_had_a_char_class as appropriate. Added some comments. Changed references to character sets to ``(non)matching lists.'' Sun Jul 1 12:11:29 1990 Karl Berry (karl at hayley) * regex.h (BYTEWIDTH): moved back to regex.c. * regex.h (re_compile_fastmap): removed declaration; this shouldn't be advertised. Mon May 28 15:27:53 1990 Kathy Hargreaves (kathy at hayley) * regex.c (ifndef Sword): Made comments more specific. (global): include so can write fatal messages on standard error. Replaced calls to assert with fprintfs to stderr and exit (1)'s. (PREFETCH): Reformatted to make more readable. (AT_STRINGS_BEG): Defined to test if we're at the beginning of the virtual concatenation of string1 and string2. (AT_STRINGS_END): Defined to test if at the end of the virtual concatenation of string1 and string2. (AT_WORD_BOUNDARY): Defined to test if are at a word boundary. (IS_A_LETTER(d)): Defined to test if the contents of the pointer D is a letter. (re_match_2): Rewrote the wordbound, notwordbound, wordbeg, wordend, begbuf, and endbuf cases in terms of the above four new macros. Called SET_REGS_MATCHED in the matchsyntax, matchnotsyntax, wordchar, and notwordchar cases. Mon May 14 14:49:13 1990 Kathy Hargreaves (kathy at hayley) * regex.c (re_search_2): Fixed RANGE to not ever take STARTPOS outside of virtual concatenation of STRING1 and STRING2. Updated header comment as to this. (re_match_2): Clarified comment about MSTOP in header. Sat May 12 15:39:00 1990 Kathy Hargreaves (kathy at hayley) * regex.c (re_search_2): Checked for out-of-range STARTPOS. Added comments. When searching backwards, not only get the character with which to compare to the fastmap from string2 if the starting position >= size1, but also if size1 is zero; this is so won't get a segmentation fault if string1 is null. Reformatted code at label advance. Thu Apr 12 20:26:21 1990 Kathy Hargreaves (kathy at hayley) * regex.h: Added #pragma once and #ifdef...endif __REGEXP_LIBRARY. (RE_EXACTN_VALUE): Added for search.c to use. Reworded some comments. regex.c: Punctuated some comments correctly. (NULL): Removed this. (RE_EXACTN_VALUE): Added for search.c to use. (): Moved this include to top of file. (): Added this include. (struct regexpcode): Assigned 0 to unused and 1 to exactn because of RE_EXACTN_VALUE. Added comment. (various macros): Lined up backslashes near end of line. (insert_jump): Cleaned up the header comment. (re_search): Corrected the header comment. (re_search_2): Cleaned up and completed the header comment. (re_max_failures): Updated comment. (struct register_info): Constructed as bits so as to save space on the stack when pushing register information. (IS_ACTIVE): Macro for struct register_info. (MATCHED_SOMETHING): Macro for struct register_info. (NUM_REG_ITEMS): How many register information items for each register we have to push on the stack at each failure. (MAX_NUM_FAILURE_ITEMS): If push all the registers on failure, this is how many items we push on the stack. (PUSH_FAILURE_POINT): Now pushes whether or not the register is currently active, and whether or not it matched something. Checks that there's enough space allocated to accomodate all the items we currently want to push. (Before, a test for an empty stack sufficed because we always pushed and popped the same number of items). Replaced ``2'' with MAX_NUM_FAILURE_POINTS when ``2'' refers to how many things get pushed on the stack each time. When copy the stack into the newly allocated storage, now only copy the area in use. Clarified comment. (POP_FAILURE_POINT): Defined to use in places where put number of registers on the stack into a variable before using it to decrement the stack, so as to not confuse the compiler. (IS_IN_FIRST_STRING): Defined to check if a pointer points into the first string. (SET_REGS_MATCHED): Changed to use the struct register_info bits; also set the matched-something bit to false if the register isn't currently active. (This is a redundant setting.) (re_match_2): Cleaned up and completed the header comment. Updated the failure stack comment. Replaced the ``2'' with MAX_NUM_FAILURE_ITEMS in the static allocation of initial_stack, because now more than two (now up to MAX_FAILURE_ITEMS) items get pushed on the failure stack each time. Ditto for stackb. Trashed restart_seg1, regend_seg1, best_regstart_seg1, and best_regend_seg1 because they could have erroneous information in them, such as when matching ``a'' (in string1) and ``ab'' (in string2) with ``(a)*ab''; before using IS_IN_FIRST_STRING to see whether or not the register starts or ends in string1, regstart[1] pointed past the end of string1, yet regstart_seg1 was 0! Added variable reg_info of type struct register_info to keep track of currently active registers and whether or not they currently match anything. Commented best_regs_set. Trashed reg_active and reg_matched_something and put the information they held into reg_info; saves space on the stack. Replaced NULL with '\000'. In begline case, compacted the code. Used assert to exit if had an internal error. In begbuf case, because now force the string we're working on into string2 if there aren't two strings, now allow d == string2 if there is no string1 (and the check for that is size1 == 0!); also now succeeds if there aren't any strings at all. (main, ifdef canned): Put test type into a variable so could change it while debugging. Sat Mar 24 12:24:13 1990 Kathy Hargreaves (kathy at hayley) * regex.c (GET_UNSIGNED_NUMBER): Deleted references to num_fetches. (re_compile_pattern): Deleted num_fetches because could keep track of the number of fetches done by saving a pointer into the pattern. Added variable beg_interval to be used as a pointer, as above. Assert that beg_interval points to something when it's used as above. Initialize succeed_n's to lower_bound because re_compile_fastmap needs to know it. (re_compile_fastmap): Deleted unnecessary variable is_a_jump_n. Added comment. (re_match_2): Put number of registers on the stack into a variable before using it to decrement the stack, so as to not confuse the compiler. Updated comments. Used error routine instead of printf and exit. In exactn case, restored longer code from ``original'' regex.c which doesn't test translate inside a loop. * regex.h: Moved #define NULL and the enum regexpcode definition and to regex.c. Changed some comments. regex.c (global): Updated comments about compiling and for the re_compile_pattern jump routines. Added #define NULL and the enum regexpcode definition (from regex.h). (enum regexpcode): Added set_number_at to reset the n's of succeed_n's and jump_n's. (re_set_syntax): Updated its comment. (re_compile_pattern): Moved its heading comment to after its macros. Moved its include statement to the top of the file. Commented or added to comments of its macros. In start_memory case: Push laststart value before adding start_memory and its register number to the buffer, as they might not get added. Added code to put a set_number_at before each succeed_n and one after each jump_n; rewrote code in what seemed a more straightforward manner to put all these things in the pattern so the succeed_n's would correctly jump to the set_number_at's of the matching jump_n's, and so the jump_n's would correctly jump to after the set_number_at's of the matching succeed_n's. Initialize succeed_n n's to -1. (insert_op_2): Added this to insert an operation followed by two integers. (re_compile_fastmap): Added set_number_at case. (re_match_2): Moved heading comment to after macros. Added mention of REGS to heading comment. No longer turn a succeed_n with n = 0 into an on_failure_jump, because n needs to be reset each time through a loop. Check to see if a succeed_n's n is set by its set_number_at. Added set_number_at case. Updated some comments. (main): Added another main to run posix tests, which is compiled ifdef both test and canned. (Old main is still compiled ifdef test only). Tue Mar 19 09:22:55 1990 Kathy Hargreaves (kathy at hayley) * regex.[hc]: Change all instances of the word ``legal'' to ``valid'' and all instances of ``illegal'' to ``invalid.'' Sun Mar 4 12:11:31 1990 Kathy Hargreaves (kathy at hayley) * regex.h: Added syntax bit RE_NO_EMPTY_RANGES which is set if an ending range point has to collate higher or equal to the starting range point. Added syntax bit RE_NO_HYPHEN_RANGE_END which is set if a hyphen can't be an ending range point. Set to two above bits in RE_SYNTAX_POSIX_BASIC and RE_SYNTAX_POSIX_EXTENDED. regex.c: (re_compile_pattern): Don't allow empty ranges if the RE_NO_EMPTY_RANGES syntax bit is set. Don't let a hyphen be a range end if the RE_NO_HYPHEN_RANGE_END syntax bit is set. (ESTACK_PUSH_2): renamed this PUSH_FAILURE_POINT and made it push all the used registers on the stack, as well as the number of the highest numbered register used, and (as before) the two failure points. (re_match_2): Fixed up comments. Added arrays best_regstart[], best_regstart_seg1[], best_regend[], and best_regend_seg1[] to keep track of the best match so far whenever reach the end of the pattern but not the end of the string, and there are still failure points on the stack with which to backtrack; if so, do the saving and force a fail. If reach the end of the pattern but not the end of the string, but there are no more failure points to try, restore the best match so far, set the registers and return. Compacted some code. In stop_memory case, if the subexpression we've just left is in a loop, push onto the stack the loop's on_failure_jump failure point along with the current pointer into the string (d). In finalize_jump case, in addition to popping the failure points, pop the saved registers. In the fail case, restore the registers, as well as the failure points. Sun Feb 18 15:08:10 1990 Kathy Hargreaves (kathy at hayley) * regex.c: (global): Defined a macro GET_BUFFER_SPACE which makes sure you have a specified number of buffer bytes allocated. Redefined the macro BUFPUSH to use this. Added comments. (re_compile_pattern): Call GET_BUFFER_SPACE before storing or inserting any jumps. (re_match_2): Set d to string1 + pos and dend to end_match_1 only if string1 isn't null. Force exit from a loop if it's around empty parentheses. In stop_memory case, if found some jumps, increment p2 before extracting address to which to jump. Also, don't need to know how many more times can jump_n. In begline case, d must equal string1 or string2, in that order, only if they are not null. In maybe_finalize_jump case, skip over start_memorys' and stop_memorys' register numbers, too. Thu Feb 15 15:53:55 1990 Kathy Hargreaves (kathy at hayley) * regex.c (BUFPUSH): off by one goof in deciding whether to EXTEND_BUFFER. Wed Jan 24 17:07:46 1990 Kathy Hargreaves (kathy at hayley) * regex.h: Moved definition of NULL to here. Got rid of ``In other words...'' comment. Added to some comments. regex.c: (re_compile_pattern): Tried to bulletproof some code, i.e., checked if backward references (e.g., p[-1]) were within the range of pattern. (re_compile_fastmap): Fixed a bug in succeed_n part where was getting the amount to jump instead of how many times to jump. (re_search_2): Changed the name of the variable ``total'' to ``total_size.'' Condensed some code. (re_match_2): Moved the comment about duplicate from above the start_memory case to above duplicate case. (global): Rewrote some comments. Added commandline arguments to testing. Wed Jan 17 11:47:27 1990 Kathy Hargreaves (kathy at hayley) * regex.c: (global): Defined a macro STORE_NUMBER which stores a number into two contiguous bytes. Also defined STORE_NUMBER_AND_INCR which does the same thing and then increments the pointer to the storage place to point after the number. Defined a macro EXTRACT_NUMBER which extracts a number from two continguous bytes. Also defined EXTRACT_NUMBER_AND_INCR which does the same thing and then increments the pointer to the source to point to after where the number was. Tue Jan 16 12:09:19 1990 Kathy Hargreaves (kathy at hayley) * regex.h: Incorporated rms' changes. Defined RE_NO_BK_REFS syntax bit which is set when want to interpret back reference patterns as literals. Defined RE_NO_EMPTY_BRACKETS syntax bit which is set when want empty bracket expressions to be illegal. Defined RE_CONTEXTUAL_ILLEGAL_OPS syntax bit which is set when want it to be illegal for *, +, ? and { to be first in an re or come immediately after a | or a (, and for ^ not to appear in a nonleading position and $ in a nontrailing position (outside of bracket expressions, that is). Defined RE_LIMITED_OPS syntax bit which is set when want +, ? and | to always be literals instead of ops. Fixed up the Posix syntax. Changed the syntax bit comments from saying, e.g., ``0 means...'' to ``If this bit is set, it means...''. Changed the syntax bit defines to use shifts instead of integers. * regex.c: (global): Incorporated rms' changes. (re_compile_pattern): Incorporated rms' changes Made it illegal for a $ to appear anywhere but inside a bracket expression or at the end of an re when RE_CONTEXTUAL_ILLEGAL_OPS is set. Made the same hold for $ except it has to be at the beginning of an re instead of the end. Made the re "[]" illegal if RE_NO_EMPTY_BRACKETS is set. Made it illegal for | to be first or last in an re, or immediately follow another | or a (. Added and embellished some comments. Allowed \{ to be interpreted as a literal if RE_NO_BK_CURLY_BRACES is set. Made it illegal for *, +, ?, and { to appear first in an re, or immediately follow a | or a ( when RE_CONTEXTUAL_ILLEGAL_OPS is set. Made back references interpreted as literals if RE_NO_BK_REFS is set. Made recursive intervals either illegal (if RE_NO_BK_CURLY_BRACES isn't set) or interpreted as literals (if is set), if RE_INTERVALS is set. Made it treat +, ? and | as literals if RE_LIMITED_OPS is set. Cleaned up some code. Thu Dec 21 15:31:32 1989 Kathy Hargreaves (kathy at hayley) * regex.c: (global): Moved RE_DUP_MAX to regex.h and made it equal 2^15 - 1 instead of 1000. Defined NULL to be zero. Moved the definition of BYTEWIDTH to regex.h. Made the global variable obscure_syntax nonstatic so the tests in another file could use it. (re_compile_pattern): Defined a maximum length (CHAR_CLASS_MAX_LENGTH) for character class strings (i.e., what's between the [: and the :]'s). Defined a macro SET_LIST_BIT(c) which sets the bit for C in a character set list. Took out comments that EXTEND_BUFFER clobbers C. Made the string "^" match itself, if not RE_CONTEXT_IND_OPS. Added character classes to bracket expressions. Change the laststart pointer saved with the start of each subexpression to point to start_memory instead of after the following register number. This is because the subexpression might be in a loop. Added comments and compacted some code. Made intervals only work if preceded by an re matching a single character or a subexpression. Made back references to nonexistent subexpressions illegal if using POSIX syntax. Made intervals work on the last preceding character of a concatenation of characters, e.g., ab{0,} matches abbb, not abab. Moved macro PREFETCH to outside the routine. (re_compile_fastmap): Added succeed_n to work analogously to on_failure_jump if n is zero and jump_n to work analogously to the other backward jumps. (re_match_2): Defined macro SET_REGS_MATCHED to set which current subexpressions had matches within them. Changed some comments. Added reg_active and reg_matched_something arrays to keep track of in which subexpressions currently have matched something. Defined MATCHING_IN_FIRST_STRING and replaced ``dend == end_match_1'' with it to make code easier to understand. Fixed so can apply * and intervals to arbitrarily nested subexpressions. (Lots of previous bugs here.) Changed so won't match a newline if syntax bit RE_DOT_NOT_NULL is set. Made the upcase array nonstatic so the testing file could use it also. (main.c): Moved the tests out to another file. (tests.c): Moved all the testing stuff here. Sat Nov 18 19:30:30 1989 Kathy Hargreaves (kathy at hayley) * regex.c: (re_compile_pattern): Defined RE_DUP_MAX, the maximum number of times an interval can match a pattern. Added macro GET_UNSIGNED_NUMBER (used to get below): Added variables lower_bound and upper_bound for upper and lower bounds of intervals. Added variable num_fetches so intervals could do backtracking. Added code to handle '{' and "\{" and intervals. Added to comments. (store_jump_n): (Added) Stores a jump with a number following the relative address (for intervals). (insert_jump_n): (Added) Inserts a jump_n. (re_match_2): Defined a macro ESTACK_PUSH_2 for the error stack; it checks for overflow and reallocates if necessary. * regex.h: Added bits (RE_INTERVALS and RE_NO_BK_CURLY_BRACES) to obscure syntax to indicate whether or not a syntax handles intervals and recognizes either \{ and \} or { and } as operators. Also added two syntaxes RE_SYNTAX_POSIX_BASIC and RE_POSIX_EXTENDED and two command codes to the enumeration regexpcode; they are succeed_n and jump_n. Sat Nov 18 19:30:30 1989 Kathy Hargreaves (kathy at hayley) * regex.c: (re_compile_pattern): Defined INIT_BUFF_SIZE to get rid of repeated constants in code. Tested with value 1. Renamed PATPUSH as BUFPUSH, since it pushes things onto the buffer, not the pattern. Also made this macro extend the buffer if it's full (so could do the following): Took out code at top of loop that checks to see if buffer is going to be full after 10 additions (and reallocates if necessary). (insert_jump): Rearranged declaration lines so comments would read better. (re_match_2): Compacted exactn code and added more comments. (main): Defined macros TEST_MATCH and MATCH_SELF to do testing; took out loop so could use these instead. Tue Oct 24 20:57:18 1989 Kathy Hargreaves (kathy at hayley) * regex.c (re_set_syntax): Gave argument `syntax' a type. (store_jump, insert_jump): made them void functions. Local Variables: mode: indented-text left-margin: 8 version-control: never End: