From 098b07d5cb1d6aa13b81a0f43ea5e151829ad26c Mon Sep 17 00:00:00 2001 From: Karl Williamson Date: Tue, 20 Aug 2013 21:51:23 -0600 Subject: [PATCH] Allow trie use for /iaa matching This adds code so that tries can be formed under /iaa, which formerly weren't handled. A problem occurs when the string contains the LATIN SMALL LETTER SHARP S when the regex pattern is not UTF-8 encoded. I tried several ways to get this to work easily, but ended up deciding it was too hard, to in this one situation, a new regnode is created to prevent the trie code from even trying to turn it into a trie. --- pod/perldebguts.pod | 3 + pod/perldelta.pod | 8 +- regcomp.c | 39 ++++++--- regcomp.sym | 1 + regexec.c | 49 +++++++++--- regnodes.h | 227 +++++++++++++++++++++++++++------------------------- 6 files changed, 190 insertions(+), 137 deletions(-) diff --git a/pod/perldebguts.pod b/pod/perldebguts.pod index 49238b2..ff9f97b 100644 --- a/pod/perldebguts.pod +++ b/pod/perldebguts.pod @@ -636,6 +636,9 @@ will be lost. EXACTFU_SS str Match this string (folded iff in UTF-8, length in folding may change even if not in UTF-8) using /iu rules (w/len). + EXACTFA_NO_TRIE str Match this string (which is not trie-able; + not guaranteed to be folded) using /iaa + rules (w/len). # Do nothing types diff --git a/pod/perldelta.pod b/pod/perldelta.pod index 698c2ac..722abf4 100644 --- a/pod/perldelta.pod +++ b/pod/perldelta.pod @@ -78,16 +78,12 @@ as an updated module in the L section. =head1 Performance Enhancements -XXX Changes which enhance performance without changing behaviour go here. -There may well be none in a stable release. - -[ List each enhancement as a =item entry ] - =over 4 =item * -XXX +The trie performance enhancement for regular expressions has now been +extended to those compiled under C. =back diff --git a/regcomp.c b/regcomp.c index 51141c0..e22dcd8 100644 --- a/regcomp.c +++ b/regcomp.c @@ -2730,7 +2730,7 @@ S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode * this file makes sure that in EXACTFU nodes, the sharp s gets folded to * 'ss', even if the pattern isn't UTF-8. This avoids the issues * described in the next item. - * 4) A problem remains for the sharp s in EXACTF and EXACTFA nodes when the + * 3) A problem remains for the sharp s in EXACTF and EXACTFA nodes when the * pattern isn't in UTF-8. (BTW, there cannot be an EXACTF node with a * UTF-8 pattern.) An assumption that the optimizer part of regexec.c * (probably unwittingly, in Perl_regexec_flags()) makes is that a @@ -2761,7 +2761,14 @@ S_make_trie_failtable(pTHX_ RExC_state_t *pRExC_state, regnode *source, regnode * but in a non-UTF8 pattern, folding it to that above-Latin1 string would * require the pattern to be forced into UTF-8, the overhead of which we * want to avoid.) - */ + * + * Similarly, the code that generates tries doesn't currently handle + * not-already-folded multi-char folds, and it looks like a pain to change + * that. Therefore, trie generation of EXACTFA nodes with the sharp s + * doesn't work. Instead, such an EXACTFA is turned into a new regnode, + * EXACTFA_NO_TRIE, which the trie code knows not to handle. Most people + * using /iaa matching will be doing so almost entirely with ASCII + * strings, so this should rarely be encountered in practice */ #define JOIN_EXACT(scan,min_subtract,has_exactf_sharp_s, flags) \ if (PL_regkind[OP(scan)] == EXACT) \ @@ -2882,9 +2889,12 @@ S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, UV *min_subtract, b } /* Nodes with 'ss' require special handling, except for EXACTFL - * and EXACTFA for which there is no multi-char fold to this */ + * and EXACTFA-ish for which there is no multi-char fold to + * this */ if (len == 2 && *s == 's' && *(s+1) == 's' - && OP(scan) != EXACTFL && OP(scan) != EXACTFA) + && OP(scan) != EXACTFL + && OP(scan) != EXACTFA + && OP(scan) != EXACTFA_NO_TRIE) { count = 2; OP(scan) = EXACTFU_SS; @@ -2902,7 +2912,10 @@ S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, UV *min_subtract, b * test for them. The code that generates the * is_MULTI_foo() macros croaks should one actually get put * into Unicode .) */ - if (OP(scan) != EXACTFL && OP(scan) != EXACTFA) { + if (OP(scan) != EXACTFL + && OP(scan) != EXACTFA + && OP(scan) != EXACTFA_NO_TRIE) + { count = utf8_length(s, multi_end); s = multi_end; } @@ -2931,9 +2944,12 @@ S_join_exact(pTHX_ RExC_state_t *pRExC_state, regnode *scan, UV *min_subtract, b /* Non-UTF-8 pattern, EXACTFA node. There can't be a multi-char * fold to the ASCII range (and there are no existing ones in the * upper latin1 range). But, as outlined in the comments preceding - * this function, we need to flag any occurrences of the sharp s */ + * this function, we need to flag any occurrences of the sharp s. + * This character forbids trie formation (because of added + * complexity) */ while (s < s_end) { if (*s == LATIN_SMALL_LETTER_SHARP_S) { + OP(scan) = EXACTFA_NO_TRIE; *has_exactf_sharp_s = TRUE; break; } @@ -3353,13 +3369,14 @@ S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp, EXACT | EXACT EXACTFU | EXACTFU EXACTFU_SS | EXACTFU - EXACTFA | 0 + EXACTFA | EXACTFA */ #define TRIE_TYPE(X) ( ( NOTHING == (X) ) ? NOTHING : \ ( EXACT == (X) ) ? EXACT : \ ( EXACTFU == (X) || EXACTFU_SS == (X) ) ? EXACTFU : \ + ( EXACTFA == (X) ) ? EXACTFA : \ 0 ) /* dont use tail as the end marker for this traverse */ @@ -3739,7 +3756,8 @@ S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp, /* All other (EXACTFL handled above) folds except under * /iaa that include s, S, and sharp_s also may include * the others */ - if (OP(scan) != EXACTFA) { + if (OP(scan) != EXACTFA && OP(scan) != EXACTFA_NO_TRIE) + { if (uc == 's' || uc == 'S') { ANYOF_BITMAP_SET(data->start_class, LATIN_SMALL_LETTER_SHARP_S); @@ -3776,7 +3794,9 @@ S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp, /* All folds except under /iaa that include s, S, * and sharp_s also may include the others */ - if (OP(scan) != EXACTFA) { + if (OP(scan) != EXACTFA + && OP(scan) != EXACTFA_NO_TRIE) + { if (uc == 's' || uc == 'S') { ANYOF_BITMAP_SET(data->start_class, LATIN_SMALL_LETTER_SHARP_S); @@ -14435,6 +14455,7 @@ S_regtail_study(pTHX_ RExC_state_t *pRExC_state, regnode *p, const regnode *val, switch (OP(scan)) { case EXACT: case EXACTF: + case EXACTFA_NO_TRIE: case EXACTFA: case EXACTFU: case EXACTFU_SS: diff --git a/regcomp.sym b/regcomp.sym index 8b19bff..7d3cdc0 100644 --- a/regcomp.sym +++ b/regcomp.sym @@ -100,6 +100,7 @@ EXACTFL EXACT, str ; Match this string (not guaranteed to be fold EXACTFU EXACT, str ; Match this string (folded iff in UTF-8, length in folding doesn't change if not in UTF-8) using /iu rules (w/len). EXACTFA EXACT, str ; Match this string (not guaranteed to be folded) using /iaa rules (w/len). EXACTFU_SS EXACT, str ; Match this string (folded iff in UTF-8, length in folding may change even if not in UTF-8) using /iu rules (w/len). +EXACTFA_NO_TRIE EXACT, str ; Match this string (which is not trie-able; not guaranteed to be folded) using /iaa rules (w/len). #*Do nothing types diff --git a/regexec.c b/regexec.c index 30d7eea..b441682 100644 --- a/regexec.c +++ b/regexec.c @@ -207,13 +207,13 @@ static const char* const non_utf8_target_but_utf8_required /* Currently these are only used when PL_regkind[OP(rn)] == EXACT so we don't need this definition. */ #define IS_TEXT(rn) ( OP(rn)==EXACT || OP(rn)==REF || OP(rn)==NREF ) -#define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFA || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF ) +#define IS_TEXTF(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn)==EXACTFA || OP(rn)==EXACTFA_NO_TRIE || OP(rn)==EXACTF || OP(rn)==REFF || OP(rn)==NREFF ) #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL || OP(rn)==REFFL || OP(rn)==NREFFL ) #else /* ... so we use this as its faster. */ #define IS_TEXT(rn) ( OP(rn)==EXACT ) -#define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn) == EXACTFA) +#define IS_TEXTFU(rn) ( OP(rn)==EXACTFU || OP(rn)==EXACTFU_SS || OP(rn) == EXACTFA || OP(rn) == EXACTFA_NO_TRIE) #define IS_TEXTF(rn) ( OP(rn)==EXACTF ) #define IS_TEXTFL(rn) ( OP(rn)==EXACTFL ) @@ -1227,15 +1227,22 @@ Perl_re_intuit_start(pTHX_ } #define DECL_TRIE_TYPE(scan) \ - const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold } \ + const enum { trie_plain, trie_utf8, trie_utf8_fold, trie_latin_utf8_fold, \ + trie_utf8_exactfa_fold, trie_latin_utf8_exactfa_fold } \ trie_type = ((scan->flags == EXACT) \ ? (utf8_target ? trie_utf8 : trie_plain) \ - : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold)) + : (scan->flags == EXACTFA) \ + ? (utf8_target ? trie_utf8_exactfa_fold : trie_latin_utf8_exactfa_fold) \ + : (utf8_target ? trie_utf8_fold : trie_latin_utf8_fold)) #define REXEC_TRIE_READ_CHAR(trie_type, trie, widecharmap, uc, uscan, len, uvc, charid, foldlen, foldbuf, uniflags) \ STMT_START { \ STRLEN skiplen; \ + U8 flags = FOLD_FLAGS_FULL; \ switch (trie_type) { \ + case trie_utf8_exactfa_fold: \ + flags |= FOLD_FLAGS_NOMIX_ASCII; \ + /* FALL THROUGH */ \ case trie_utf8_fold: \ if ( foldlen>0 ) { \ uvc = utf8n_to_uvchr( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ @@ -1243,13 +1250,16 @@ STMT_START { \ uscan += len; \ len=0; \ } else { \ - uvc = to_utf8_fold( (const U8*) uc, foldbuf, &foldlen ); \ + uvc = _to_utf8_fold_flags( (const U8*) uc, foldbuf, &foldlen, flags, NULL); \ len = UTF8SKIP(uc); \ skiplen = UNISKIP( uvc ); \ foldlen -= skiplen; \ uscan = foldbuf + skiplen; \ } \ break; \ + case trie_latin_utf8_exactfa_fold: \ + flags |= FOLD_FLAGS_NOMIX_ASCII; \ + /* FALL THROUGH */ \ case trie_latin_utf8_fold: \ if ( foldlen>0 ) { \ uvc = utf8n_to_uvchr( (const U8*) uscan, UTF8_MAXLEN, &len, uniflags ); \ @@ -1258,7 +1268,7 @@ STMT_START { \ len=0; \ } else { \ len = 1; \ - uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, FOLD_FLAGS_FULL); \ + uvc = _to_fold_latin1( (U8) *uc, foldbuf, &foldlen, flags); \ skiplen = UNISKIP( uvc ); \ foldlen -= skiplen; \ uscan = foldbuf + skiplen; \ @@ -1487,6 +1497,9 @@ S_find_byclass(pTHX_ regexp * prog, const regnode *c, char *s, ); break; + case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 patterns */ + assert(! is_utf8_pat); + /* FALL THROUGH */ case EXACTFA: if (is_utf8_pat || utf8_target) { utf8_fold_flags = FOLDEQ_UTF8_NOMIX_ASCII; @@ -3485,7 +3498,8 @@ S_setup_EXACTISH_ST_c1_c2(pTHX_ const regnode * const text_node, int *c1p, * which is the one above 255 */ if ((c1 < 256) != (c2 < 256)) { if (OP(text_node) == EXACTFL - || (OP(text_node) == EXACTFA + || ((OP(text_node) == EXACTFA + || OP(text_node) == EXACTFA_NO_TRIE) && (isASCII(c1) || isASCII(c2)))) { if (c1 < 256) { @@ -3503,7 +3517,9 @@ S_setup_EXACTISH_ST_c1_c2(pTHX_ const regnode * const text_node, int *c1p, if (utf8_target && HAS_NONLATIN1_FOLD_CLOSURE(c1) && OP(text_node) != EXACTFL - && (OP(text_node) != EXACTFA || ! isASCII(c1))) + && ((OP(text_node) != EXACTFA + && OP(text_node) != EXACTFA_NO_TRIE) + || ! isASCII(c1))) { /* Here, there could be something above Latin1 in the target which * folds to this character in the pattern. All such cases except @@ -3534,6 +3550,10 @@ S_setup_EXACTISH_ST_c1_c2(pTHX_ const regnode * const text_node, int *c1p, /* FALLTHROUGH */ /* /u rules for all these. This happens to work for * EXACTFA as nothing in Latin1 folds to ASCII */ + case EXACTFA_NO_TRIE: /* This node only generated for + non-utf8 patterns */ + assert(! is_utf8_pat); + /* FALL THROUGH */ case EXACTFA: case EXACTFU_SS: case EXACTFU: @@ -4221,6 +4241,10 @@ S_regmatch(pTHX_ regmatch_info *reginfo, char *startpos, regnode *prog) fold_utf8_flags = is_utf8_pat ? FOLDEQ_S1_ALREADY_FOLDED : 0; goto do_exactf; + case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 + patterns */ + assert(! is_utf8_pat); + /* FALL THROUGH */ case EXACTFA: /* /abc/iaa */ folder = foldEQ_latin1; fold_array = PL_fold_latin1; @@ -6890,8 +6914,11 @@ S_regrepeat(pTHX_ regexp *prog, char **startposp, const regnode *p, } break; + case EXACTFA_NO_TRIE: /* This node only generated for non-utf8 patterns */ + assert(! reginfo->is_utf8_pat); + /* FALL THROUGH */ case EXACTFA: - utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; + utf8_flags = FOLDEQ_UTF8_NOMIX_ASCII; goto do_exactf; case EXACTFL: @@ -6901,8 +6928,8 @@ S_regrepeat(pTHX_ regexp *prog, char **startposp, const regnode *p, case EXACTF: /* This node only generated for non-utf8 patterns */ assert(! reginfo->is_utf8_pat); - utf8_flags = 0; - goto do_exactf; + utf8_flags = 0; + goto do_exactf; case EXACTFU_SS: case EXACTFU: diff --git a/regnodes.h b/regnodes.h index 0da6757..2d55c8b 100644 --- a/regnodes.h +++ b/regnodes.h @@ -6,8 +6,8 @@ /* Regops and State definitions */ -#define REGNODE_MAX 94 -#define REGMATCH_STATE_MAX 134 +#define REGNODE_MAX 95 +#define REGMATCH_STATE_MAX 135 #define END 0 /* 0000 End of program. */ #define SUCCEED 1 /* 0x01 Return from a subroutine, basically. */ @@ -50,60 +50,61 @@ #define EXACTFU 38 /* 0x26 Match this string (folded iff in UTF-8, length in folding doesn't change if not in UTF-8) using /iu rules (w/len). */ #define EXACTFA 39 /* 0x27 Match this string (not guaranteed to be folded) using /iaa rules (w/len). */ #define EXACTFU_SS 40 /* 0x28 Match this string (folded iff in UTF-8, length in folding may change even if not in UTF-8) using /iu rules (w/len). */ -#define NOTHING 41 /* 0x29 Match empty string. */ -#define TAIL 42 /* 0x2a Match empty string. Can jump here from outside. */ -#define STAR 43 /* 0x2b Match this (simple) thing 0 or more times. */ -#define PLUS 44 /* 0x2c Match this (simple) thing 1 or more times. */ -#define CURLY 45 /* 0x2d Match this simple thing {n,m} times. */ -#define CURLYN 46 /* 0x2e Capture next-after-this simple thing */ -#define CURLYM 47 /* 0x2f Capture this medium-complex thing {n,m} times. */ -#define CURLYX 48 /* 0x30 Match this complex thing {n,m} times. */ -#define WHILEM 49 /* 0x31 Do curly processing and see if rest matches. */ -#define OPEN 50 /* 0x32 Mark this point in input as start of #n. */ -#define CLOSE 51 /* 0x33 Analogous to OPEN. */ -#define REF 52 /* 0x34 Match some already matched string */ -#define REFF 53 /* 0x35 Match already matched string, folded using native charset semantics for non-utf8 */ -#define REFFL 54 /* 0x36 Match already matched string, folded in loc. */ -#define REFFU 55 /* 0x37 Match already matched string, folded using unicode semantics for non-utf8 */ -#define REFFA 56 /* 0x38 Match already matched string, folded using unicode semantics for non-utf8, no mixing ASCII, non-ASCII */ -#define NREF 57 /* 0x39 Match some already matched string */ -#define NREFF 58 /* 0x3a Match already matched string, folded using native charset semantics for non-utf8 */ -#define NREFFL 59 /* 0x3b Match already matched string, folded in loc. */ -#define NREFFU 60 /* 0x3c Match already matched string, folded using unicode semantics for non-utf8 */ -#define NREFFA 61 /* 0x3d Match already matched string, folded using unicode semantics for non-utf8, no mixing ASCII, non-ASCII */ -#define IFMATCH 62 /* 0x3e Succeeds if the following matches. */ -#define UNLESSM 63 /* 0x3f Fails if the following matches. */ -#define SUSPEND 64 /* 0x40 "Independent" sub-RE. */ -#define IFTHEN 65 /* 0x41 Switch, should be preceded by switcher. */ -#define GROUPP 66 /* 0x42 Whether the group matched. */ -#define LONGJMP 67 /* 0x43 Jump far away. */ -#define BRANCHJ 68 /* 0x44 BRANCH with long offset. */ -#define EVAL 69 /* 0x45 Execute some Perl code. */ -#define MINMOD 70 /* 0x46 Next operator is not greedy. */ -#define LOGICAL 71 /* 0x47 Next opcode should set the flag only. */ -#define RENUM 72 /* 0x48 Group with independently numbered parens. */ -#define TRIE 73 /* 0x49 Match many EXACT(F[ALU]?)? at once. flags==type */ -#define TRIEC 74 /* 0x4a Same as TRIE, but with embedded charclass data */ -#define AHOCORASICK 75 /* 0x4b Aho Corasick stclass. flags==type */ -#define AHOCORASICKC 76 /* 0x4c Same as AHOCORASICK, but with embedded charclass data */ -#define GOSUB 77 /* 0x4d recurse to paren arg1 at (signed) ofs arg2 */ -#define GOSTART 78 /* 0x4e recurse to start of pattern */ -#define NGROUPP 79 /* 0x4f Whether the group matched. */ -#define INSUBP 80 /* 0x50 Whether we are in a specific recurse. */ -#define DEFINEP 81 /* 0x51 Never execute directly. */ -#define ENDLIKE 82 /* 0x52 Used only for the type field of verbs */ -#define OPFAIL 83 /* 0x53 Same as (?!) */ -#define ACCEPT 84 /* 0x54 Accepts the current matched string. */ -#define VERB 85 /* 0x55 Used only for the type field of verbs */ -#define PRUNE 86 /* 0x56 Pattern fails at this startpoint if no-backtracking through this */ -#define MARKPOINT 87 /* 0x57 Push the current location for rollback by cut. */ -#define SKIP 88 /* 0x58 On failure skip forward (to the mark) before retrying */ -#define COMMIT 89 /* 0x59 Pattern fails outright if backtracking through this */ -#define CUTGROUP 90 /* 0x5a On failure go to the next alternation in the group */ -#define KEEPS 91 /* 0x5b $& begins here. */ -#define LNBREAK 92 /* 0x5c generic newline pattern */ -#define OPTIMIZED 93 /* 0x5d Placeholder for dump. */ -#define PSEUDO 94 /* 0x5e Pseudo opcode for internal use. */ +#define EXACTFA_NO_TRIE 41 /* 0x29 Match this string (which is not trie-able; not guaranteed to be folded) using /iaa rules (w/len). */ +#define NOTHING 42 /* 0x2a Match empty string. */ +#define TAIL 43 /* 0x2b Match empty string. Can jump here from outside. */ +#define STAR 44 /* 0x2c Match this (simple) thing 0 or more times. */ +#define PLUS 45 /* 0x2d Match this (simple) thing 1 or more times. */ +#define CURLY 46 /* 0x2e Match this simple thing {n,m} times. */ +#define CURLYN 47 /* 0x2f Capture next-after-this simple thing */ +#define CURLYM 48 /* 0x30 Capture this medium-complex thing {n,m} times. */ +#define CURLYX 49 /* 0x31 Match this complex thing {n,m} times. */ +#define WHILEM 50 /* 0x32 Do curly processing and see if rest matches. */ +#define OPEN 51 /* 0x33 Mark this point in input as start of #n. */ +#define CLOSE 52 /* 0x34 Analogous to OPEN. */ +#define REF 53 /* 0x35 Match some already matched string */ +#define REFF 54 /* 0x36 Match already matched string, folded using native charset semantics for non-utf8 */ +#define REFFL 55 /* 0x37 Match already matched string, folded in loc. */ +#define REFFU 56 /* 0x38 Match already matched string, folded using unicode semantics for non-utf8 */ +#define REFFA 57 /* 0x39 Match already matched string, folded using unicode semantics for non-utf8, no mixing ASCII, non-ASCII */ +#define NREF 58 /* 0x3a Match some already matched string */ +#define NREFF 59 /* 0x3b Match already matched string, folded using native charset semantics for non-utf8 */ +#define NREFFL 60 /* 0x3c Match already matched string, folded in loc. */ +#define NREFFU 61 /* 0x3d Match already matched string, folded using unicode semantics for non-utf8 */ +#define NREFFA 62 /* 0x3e Match already matched string, folded using unicode semantics for non-utf8, no mixing ASCII, non-ASCII */ +#define IFMATCH 63 /* 0x3f Succeeds if the following matches. */ +#define UNLESSM 64 /* 0x40 Fails if the following matches. */ +#define SUSPEND 65 /* 0x41 "Independent" sub-RE. */ +#define IFTHEN 66 /* 0x42 Switch, should be preceded by switcher. */ +#define GROUPP 67 /* 0x43 Whether the group matched. */ +#define LONGJMP 68 /* 0x44 Jump far away. */ +#define BRANCHJ 69 /* 0x45 BRANCH with long offset. */ +#define EVAL 70 /* 0x46 Execute some Perl code. */ +#define MINMOD 71 /* 0x47 Next operator is not greedy. */ +#define LOGICAL 72 /* 0x48 Next opcode should set the flag only. */ +#define RENUM 73 /* 0x49 Group with independently numbered parens. */ +#define TRIE 74 /* 0x4a Match many EXACT(F[ALU]?)? at once. flags==type */ +#define TRIEC 75 /* 0x4b Same as TRIE, but with embedded charclass data */ +#define AHOCORASICK 76 /* 0x4c Aho Corasick stclass. flags==type */ +#define AHOCORASICKC 77 /* 0x4d Same as AHOCORASICK, but with embedded charclass data */ +#define GOSUB 78 /* 0x4e recurse to paren arg1 at (signed) ofs arg2 */ +#define GOSTART 79 /* 0x4f recurse to start of pattern */ +#define NGROUPP 80 /* 0x50 Whether the group matched. */ +#define INSUBP 81 /* 0x51 Whether we are in a specific recurse. */ +#define DEFINEP 82 /* 0x52 Never execute directly. */ +#define ENDLIKE 83 /* 0x53 Used only for the type field of verbs */ +#define OPFAIL 84 /* 0x54 Same as (?!) */ +#define ACCEPT 85 /* 0x55 Accepts the current matched string. */ +#define VERB 86 /* 0x56 Used only for the type field of verbs */ +#define PRUNE 87 /* 0x57 Pattern fails at this startpoint if no-backtracking through this */ +#define MARKPOINT 88 /* 0x58 Push the current location for rollback by cut. */ +#define SKIP 89 /* 0x59 On failure skip forward (to the mark) before retrying */ +#define COMMIT 90 /* 0x5a Pattern fails outright if backtracking through this */ +#define CUTGROUP 91 /* 0x5b On failure go to the next alternation in the group */ +#define KEEPS 92 /* 0x5c $& begins here. */ +#define LNBREAK 93 /* 0x5d generic newline pattern */ +#define OPTIMIZED 94 /* 0x5e Placeholder for dump. */ +#define PSEUDO 95 /* 0x5f Pseudo opcode for internal use. */ /* ------------ States ------------- */ #define TRIE_next (REGNODE_MAX + 1) /* state for TRIE */ #define TRIE_next_fail (REGNODE_MAX + 2) /* state for TRIE */ @@ -193,6 +194,7 @@ EXTCONST U8 PL_regkind[] = { EXACT, /* EXACTFU */ EXACT, /* EXACTFA */ EXACT, /* EXACTFU_SS */ + EXACT, /* EXACTFA_NO_TRIE */ NOTHING, /* NOTHING */ NOTHING, /* TAIL */ STAR, /* STAR */ @@ -336,6 +338,7 @@ static const U8 regarglen[] = { 0, /* EXACTFU */ 0, /* EXACTFA */ 0, /* EXACTFU_SS */ + 0, /* EXACTFA_NO_TRIE */ 0, /* NOTHING */ 0, /* TAIL */ 0, /* STAR */ @@ -436,6 +439,7 @@ static const char reg_off_by_arg[] = { 0, /* EXACTFU */ 0, /* EXACTFA */ 0, /* EXACTFU_SS */ + 0, /* EXACTFA_NO_TRIE */ 0, /* NOTHING */ 0, /* TAIL */ 0, /* STAR */ @@ -541,60 +545,61 @@ EXTCONST char * const PL_reg_name[] = { "EXACTFU", /* 0x26 */ "EXACTFA", /* 0x27 */ "EXACTFU_SS", /* 0x28 */ - "NOTHING", /* 0x29 */ - "TAIL", /* 0x2a */ - "STAR", /* 0x2b */ - "PLUS", /* 0x2c */ - "CURLY", /* 0x2d */ - "CURLYN", /* 0x2e */ - "CURLYM", /* 0x2f */ - "CURLYX", /* 0x30 */ - "WHILEM", /* 0x31 */ - "OPEN", /* 0x32 */ - "CLOSE", /* 0x33 */ - "REF", /* 0x34 */ - "REFF", /* 0x35 */ - "REFFL", /* 0x36 */ - "REFFU", /* 0x37 */ - "REFFA", /* 0x38 */ - "NREF", /* 0x39 */ - "NREFF", /* 0x3a */ - "NREFFL", /* 0x3b */ - "NREFFU", /* 0x3c */ - "NREFFA", /* 0x3d */ - "IFMATCH", /* 0x3e */ - "UNLESSM", /* 0x3f */ - "SUSPEND", /* 0x40 */ - "IFTHEN", /* 0x41 */ - "GROUPP", /* 0x42 */ - "LONGJMP", /* 0x43 */ - "BRANCHJ", /* 0x44 */ - "EVAL", /* 0x45 */ - "MINMOD", /* 0x46 */ - "LOGICAL", /* 0x47 */ - "RENUM", /* 0x48 */ - "TRIE", /* 0x49 */ - "TRIEC", /* 0x4a */ - "AHOCORASICK", /* 0x4b */ - "AHOCORASICKC", /* 0x4c */ - "GOSUB", /* 0x4d */ - "GOSTART", /* 0x4e */ - "NGROUPP", /* 0x4f */ - "INSUBP", /* 0x50 */ - "DEFINEP", /* 0x51 */ - "ENDLIKE", /* 0x52 */ - "OPFAIL", /* 0x53 */ - "ACCEPT", /* 0x54 */ - "VERB", /* 0x55 */ - "PRUNE", /* 0x56 */ - "MARKPOINT", /* 0x57 */ - "SKIP", /* 0x58 */ - "COMMIT", /* 0x59 */ - "CUTGROUP", /* 0x5a */ - "KEEPS", /* 0x5b */ - "LNBREAK", /* 0x5c */ - "OPTIMIZED", /* 0x5d */ - "PSEUDO", /* 0x5e */ + "EXACTFA_NO_TRIE", /* 0x29 */ + "NOTHING", /* 0x2a */ + "TAIL", /* 0x2b */ + "STAR", /* 0x2c */ + "PLUS", /* 0x2d */ + "CURLY", /* 0x2e */ + "CURLYN", /* 0x2f */ + "CURLYM", /* 0x30 */ + "CURLYX", /* 0x31 */ + "WHILEM", /* 0x32 */ + "OPEN", /* 0x33 */ + "CLOSE", /* 0x34 */ + "REF", /* 0x35 */ + "REFF", /* 0x36 */ + "REFFL", /* 0x37 */ + "REFFU", /* 0x38 */ + "REFFA", /* 0x39 */ + "NREF", /* 0x3a */ + "NREFF", /* 0x3b */ + "NREFFL", /* 0x3c */ + "NREFFU", /* 0x3d */ + "NREFFA", /* 0x3e */ + "IFMATCH", /* 0x3f */ + "UNLESSM", /* 0x40 */ + "SUSPEND", /* 0x41 */ + "IFTHEN", /* 0x42 */ + "GROUPP", /* 0x43 */ + "LONGJMP", /* 0x44 */ + "BRANCHJ", /* 0x45 */ + "EVAL", /* 0x46 */ + "MINMOD", /* 0x47 */ + "LOGICAL", /* 0x48 */ + "RENUM", /* 0x49 */ + "TRIE", /* 0x4a */ + "TRIEC", /* 0x4b */ + "AHOCORASICK", /* 0x4c */ + "AHOCORASICKC", /* 0x4d */ + "GOSUB", /* 0x4e */ + "GOSTART", /* 0x4f */ + "NGROUPP", /* 0x50 */ + "INSUBP", /* 0x51 */ + "DEFINEP", /* 0x52 */ + "ENDLIKE", /* 0x53 */ + "OPFAIL", /* 0x54 */ + "ACCEPT", /* 0x55 */ + "VERB", /* 0x56 */ + "PRUNE", /* 0x57 */ + "MARKPOINT", /* 0x58 */ + "SKIP", /* 0x59 */ + "COMMIT", /* 0x5a */ + "CUTGROUP", /* 0x5b */ + "KEEPS", /* 0x5c */ + "LNBREAK", /* 0x5d */ + "OPTIMIZED", /* 0x5e */ + "PSEUDO", /* 0x5f */ /* ------------ States ------------- */ "TRIE_next", /* REGNODE_MAX +0x01 */ "TRIE_next_fail", /* REGNODE_MAX +0x02 */ @@ -714,7 +719,7 @@ EXTCONST U8 PL_varies[] __attribute__deprecated__ = { EXTCONST U8 PL_varies_bitmask[]; #else EXTCONST U8 PL_varies_bitmask[] = { - 0x00, 0x00, 0x00, 0x00, 0x07, 0xF8, 0xF3, 0x3F, 0x13, 0x00, 0x00, 0x00 + 0x00, 0x00, 0x00, 0x00, 0x07, 0xF0, 0xE7, 0x7F, 0x26, 0x00, 0x00, 0x00 }; #endif /* DOINIT */ -- 1.8.3.1