X-Git-Url: https://perl5.git.perl.org/perl5.git/blobdiff_plain/cefafd73018b048fa66d2b22250431112141955a..4637d003d7b10148df1e72a5677e466c6b11f731:/regcomp.h diff --git a/regcomp.h b/regcomp.h index 7b73b02..1fb6059 100644 --- a/regcomp.h +++ b/regcomp.h @@ -9,34 +9,12 @@ */ #include "regcharclass.h" -typedef OP OP_4tree; /* Will be redefined later. */ - - /* Convert branch sequences to more efficient trie ops? */ #define PERL_ENABLE_TRIE_OPTIMISATION 1 -/* Be really agressive about optimising patterns with trie sequences? */ +/* Be really aggressive about optimising patterns with trie sequences? */ #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1 -/* Use old style unicode mappings for perl and posix character classes - * - * NOTE: Enabling this essentially breaks character class matching against unicode - * strings, so that POSIX char classes match when they shouldn't, and \d matches - * way more than 10 characters, and sometimes a charclass and its complement either - * both match or neither match. - * NOTE: Disabling this will cause various backwards compatibility issues to rear - * their head, and tests to fail. However it will make the charclass behaviour - * consistant regardless of internal string type, and make character class inversions - * consistant. The tests that fail in the regex engine are basically broken tests. - * - * Personally I think 5.12 should disable this for sure. Its a bit more debatable for - * 5.10, so for now im leaving it enabled. - * XXX: It is now enabled for 5.11/5.12 - * - * -demerphq - */ -#define PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS 1 - /* Should the optimiser take positive assertions into account? */ #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0 @@ -140,6 +118,8 @@ typedef OP OP_4tree; /* Will be redefined later. */ Used to make it easier to clone and free arbitrary data that the regops need. Often the ARG field of a regop is an index into this structure */ + struct reg_code_block *code_blocks;/* positions of literal (?{}) */ + int num_code_blocks; /* size of code_blocks[] */ regnode program[1]; /* Unwarranted chumminess with compiler. */ } regexp_internal; @@ -157,6 +137,7 @@ typedef OP OP_4tree; /* Will be redefined later. */ #define PREGf_NAUGHTY 0x00000004 /* how exponential is this pattern? */ #define PREGf_VERBARG_SEEN 0x00000008 #define PREGf_CUTGROUP_SEEN 0x00000010 +#define PREGf_USE_RE_EVAL 0x00000020 /* compiled with "use re 'eval'" */ /* this is where the old regcomp.h started */ @@ -272,8 +253,9 @@ struct regnode_charclass_class { #undef STRING #define OP(p) ((p)->type) -#define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \ - regnode types */ +#define FLAGS(p) ((p)->flags) /* Caution: Doesn't apply to all \ + regnode types. For some, it's the \ + character set of the regnode */ #define OPERAND(p) (((struct regnode_string *)p)->string) #define MASK(p) ((char*)OPERAND(p)) #define STR_LEN(p) (((struct regnode_string *)p)->str_len) @@ -309,31 +291,101 @@ struct regnode_charclass_class { #define SIZE_ONLY (RExC_emit == &PL_regdummy) -/* Flags for node->flags of several of the node types */ -#define USE_UNI 0x01 - -/* Flags for node->flags of ANYOF */ +/* If the bitmap doesn't fully represent what this ANYOF node can match, the + * ARG is set to this special value (since 0, 1, ... are legal, but will never + * reach this high). */ +#define ANYOF_NONBITMAP_EMPTY ((U32) -1) + +/* The information used to be stored as as combination of the ANYOF_UTF8 and + * ANYOF_NONBITMAP_NON_UTF8 bits in the flags field, but was moved out of there + * to free up a bit for other uses. This tries to hide the change from + * existing code as much as possible. Now, the data structure that goes in ARG + * is not allocated unless it is needed, and that is what is used to determine + * if there is something outside the bitmap. The code now assumes that if + * that structure exists, that any UTF-8 encoded string should be tried against + * it, but a non-UTF8-encoded string will be tried only if the + * ANYOF_NONBITMAP_NON_UTF8 bit is also set. */ +#define ANYOF_NONBITMAP(node) (ARG(node) != ANYOF_NONBITMAP_EMPTY) + +/* Flags for node->flags of ANYOF. These are in short supply, so some games + * are done to share them, as described below. Already, the ANYOF_LOCALE and + * ANYOF_CLASS bits are shared, making a space penalty for all locale nodes. + * An option would be to push them into new nodes. E.g. there could be an + * ANYOF_LOCALE node that would be in place of the flag of the same name. But + * there are better options. The UNICODE_ALL bit could be freed up by + * resorting to creating a swash containing everything above 255. This + * introduces a performance penalty. Better would be to split it off into a + * separate node, which actually would improve performance by allowing adding a + * case statement to regexec.c use the bit map for code points under 256, and + * to match everything above. If flags need to be added that are applicable to + * the synthetic start class only, with some work, they could be put in the + * next-node field, or in an unused bit of the classflags field. This could be + * done with the current EOS flag, and a new node type created that is just for + * the scc, freeing up that bit */ + +#define ANYOF_LOCALE 0x01 /* /l modifier */ + +/* The fold is calculated and stored in the bitmap where possible at compile + * time. However there are two cases where it isn't possible. These share + * this bit: 1) under locale, where the actual folding varies depending on + * what the locale is at the time of execution; and 2) where the folding is + * specified in a swash, not the bitmap, such as characters which aren't + * specified in the bitmap, or properties that aren't looked at at compile time + */ +#define ANYOF_LOC_NONBITMAP_FOLD 0x02 -#define ANYOF_LOCALE 0x01 -#define ANYOF_FOLD 0x02 -#define ANYOF_INVERT 0x04 +#define ANYOF_INVERT 0x04 -/* CLASS is never set unless LOCALE is too: has runtime \d, \w, [:posix:], ... */ -/* For now, set it always when LOCALE is set, to save a bit for other uses. */ +/* Set if this is a struct regnode_charclass_class vs a regnode_charclass. This + * is used for runtime \d, \w, [:posix:], ..., which are used only in locale + * and the optimizer's synthetic start class. Non-locale \d, etc are resolved + * at compile-time. Now shared with ANYOF_LOCALE, forcing all locale nodes to + * be large */ #define ANYOF_CLASS ANYOF_LOCALE #define ANYOF_LARGE ANYOF_CLASS /* Same; name retained for back compat */ -/* EOS used for regstclass only */ -#define ANYOF_EOS 0x10 /* Can match an empty string too */ +/* Should this character class warn if matched against a character above + * Unicode */ +#define ANYOF_WARN_SUPER 0x08 -#define ANYOF_UNICODE 0x20 /* Matches >= one thing past 0xff */ -#define ANYOF_UNICODE_ALL 0x40 /* Matches 0x100 - infinity */ +/* EOS, meaning that it can match an empty string too, is used for the + * synthetic start class only. */ +#define ANYOF_EOS 0x10 + +/* ? Is this node the synthetic start class (ssc). This bit is shared with + * ANYOF_EOS, as the latter is used only for the ssc, and then not used by + * regexec.c. And, the code is structured so that if it is set, the ssc is + * not used, so it is guaranteed to be 0 for the ssc by the time regexec.c + * gets executed, and 0 for a non-ssc ANYOF node, as it only ever gets set for + * a potential ssc candidate. Thus setting it to 1 after it has been + * determined that the ssc will be used is not ambiguous */ +#define ANYOF_IS_SYNTHETIC ANYOF_EOS + +/* Can match something outside the bitmap that isn't in utf8 */ +#define ANYOF_NONBITMAP_NON_UTF8 0x20 + +/* Matches every code point 0x100 and above*/ +#define ANYOF_UNICODE_ALL 0x40 + +/* Match all Latin1 characters that aren't ASCII when the target string is not + * in utf8. */ +#define ANYOF_NON_UTF8_LATIN1_ALL 0x80 #define ANYOF_FLAGS_ALL 0xff +/* These are the flags that ANYOF_INVERT being set or not doesn't affect + * whether they are operative or not. e.g., the node still has LOCALE + * regardless of being inverted; whereas ANYOF_UNICODE_ALL means something + * different if inverted */ +#define INVERSION_UNAFFECTED_FLAGS (ANYOF_LOCALE \ + |ANYOF_LOC_NONBITMAP_FOLD \ + |ANYOF_CLASS \ + |ANYOF_EOS \ + |ANYOF_NONBITMAP_NON_UTF8) + /* Character classes for node->classflags of ANYOF */ /* Should be synchronized with a table in regprop() */ -/* 2n should pair with 2n+1 */ +/* 2n should be the normal one, paired with its complement at 2n+1 */ #define ANYOF_ALNUM 0 /* \w, PL_utf8_alnum, utf8::IsWord, ALNUM */ #define ANYOF_NALNUM 1 @@ -397,13 +449,9 @@ struct regnode_charclass_class { #define ANYOF_CLASS_CLEAR(p, c) (ANYOF_CLASS_BYTE(p, c) &= ~ANYOF_BIT(c)) #define ANYOF_CLASS_TEST(p, c) (ANYOF_CLASS_BYTE(p, c) & ANYOF_BIT(c)) -/* Quicker way to see if there are actually any tests. This is because - * currently the set of tests can be empty even when the class bitmap is - * allocated */ -#define ANYOF_CLASS_TEST_ANY_SET(p) /* assumes sizeof(p) = 4 */ \ - memNE (((struct regnode_charclass_class*)(p))->classflags, "0000", ANYOF_CLASS_SIZE) - #define ANYOF_CLASS_ZERO(ret) Zero(((struct regnode_charclass_class*)(ret))->classflags, ANYOF_CLASSBITMAP_SIZE, char) +#define ANYOF_CLASS_SETALL(ret) \ + memset (((struct regnode_charclass_class*)(ret))->classflags, 255, ANYOF_CLASSBITMAP_SIZE) #define ANYOF_BITMAP_ZERO(ret) Zero(((struct regnode_charclass*)(ret))->bitmap, ANYOF_BITMAP_SIZE, char) #define ANYOF_BITMAP(p) (((struct regnode_charclass*)(p))->bitmap) @@ -422,7 +470,15 @@ struct regnode_charclass_class { #define ANYOF_SKIP ((ANYOF_SIZE - 1)/sizeof(regnode)) #define ANYOF_CLASS_SKIP ((ANYOF_CLASS_SIZE - 1)/sizeof(regnode)) -#define ANYOF_CLASS_ADD_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP) + +#if ANYOF_CLASSBITMAP_SIZE != 4 +# error ANYOF_CLASSBITMAP_SIZE is expected to be 4 +#endif +#define ANYOF_CLASS_TEST_ANY_SET(p) ((ANYOF_FLAGS(p) & ANYOF_CLASS) \ + && memNE (((struct regnode_charclass_class*)(p))->classflags, \ + "\0\0\0\0", ANYOF_CLASSBITMAP_SIZE)) +/*#define ANYOF_CLASS_ADD_SKIP (ANYOF_CLASS_SKIP - ANYOF_SKIP) + * */ /* @@ -439,7 +495,7 @@ struct regnode_charclass_class { #define REG_SEEN_ZERO_LEN 0x00000001 #define REG_SEEN_LOOKBEHIND 0x00000002 #define REG_SEEN_GPOS 0x00000004 -#define REG_SEEN_EVAL 0x00000008 +/* spare */ #define REG_SEEN_CANY 0x00000010 #define REG_SEEN_SANY REG_SEEN_CANY /* src bckwrd cmpt */ #define REG_SEEN_RECURSE 0x00000020 @@ -447,6 +503,7 @@ struct regnode_charclass_class { #define REG_SEEN_VERBARG 0x00000080 #define REG_SEEN_CUTGROUP 0x00000100 #define REG_SEEN_RUN_ON_COMMENT 0x00000200 +#define REG_SEEN_EXACTF_SHARP_S 0x00000400 START_EXTERN_C @@ -473,8 +530,9 @@ EXTCONST regexp_engine PL_core_reg_engine = { Perl_reg_named_buff_iter, Perl_reg_qr_package, #if defined(USE_ITHREADS) - Perl_regdupe_internal + Perl_regdupe_internal, #endif + Perl_re_op_compile }; #endif /* DOINIT */ #endif /* PLUGGABLE_RE_EXTENSION */ @@ -487,9 +545,9 @@ END_EXTERN_C * The character describes the function of the corresponding .data item: * a - AV for paren_name_list under DEBUGGING * f - start-class data for regstclass optimization - * n - Root of op tree for (?{EVAL}) item - * o - Start op for (?{EVAL}) item - * p - Pad for (?{EVAL}) item + * l - start op for literal (?{EVAL}) item + * L - start op for literal (?{EVAL}) item, with separate CV (qr//) + * r - pointer to an embedded code-containing qr, e.g. /ab$qr/ * s - swash for Unicode-style character class, and the multicharacter * strings resulting from casefolding the single-character entries * in the character class @@ -630,7 +688,7 @@ struct _reg_ac_data { }; typedef struct _reg_ac_data reg_ac_data; -/* ANY_BIT doesnt use the structure, so we can borrow it here. +/* ANY_BIT doesn't use the structure, so we can borrow it here. This is simpler than refactoring all of it as wed end up with three different sets... */ @@ -791,26 +849,27 @@ re.pm, especially to the documentation. #ifdef DEBUGGING -#define GET_RE_DEBUG_FLAGS_DECL VOL IV re_debug_flags = 0; GET_RE_DEBUG_FLAGS; +#define GET_RE_DEBUG_FLAGS_DECL VOL IV re_debug_flags \ + PERL_UNUSED_DECL = 0; GET_RE_DEBUG_FLAGS; #define RE_PV_COLOR_DECL(rpv,rlen,isuni,dsv,pv,l,m,c1,c2) \ const char * const rpv = \ pv_pretty((dsv), (pv), (l), (m), \ PL_colors[(c1)],PL_colors[(c2)], \ - PERL_PV_ESCAPE_RE |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \ + PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ); \ const int rlen = SvCUR(dsv) #define RE_SV_ESCAPE(rpv,isuni,dsv,sv,m) \ const char * const rpv = \ pv_pretty((dsv), (SvPV_nolen_const(sv)), (SvCUR(sv)), (m), \ PL_colors[(c1)],PL_colors[(c2)], \ - PERL_PV_ESCAPE_RE |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ) + PERL_PV_ESCAPE_RE|PERL_PV_ESCAPE_NONASCII |((isuni) ? PERL_PV_ESCAPE_UNI : 0) ) #define RE_PV_QUOTED_DECL(rpv,isuni,dsv,pv,l,m) \ const char * const rpv = \ pv_pretty((dsv), (pv), (l), (m), \ PL_colors[0], PL_colors[1], \ - ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_PRETTY_ELLIPSES | \ + ( PERL_PV_PRETTY_QUOTE | PERL_PV_ESCAPE_RE | PERL_PV_ESCAPE_NONASCII | PERL_PV_PRETTY_ELLIPSES | \ ((isuni) ? PERL_PV_ESCAPE_UNI : 0)) \ ) @@ -832,8 +891,8 @@ re.pm, especially to the documentation. * Local variables: * c-indentation-style: bsd * c-basic-offset: 4 - * indent-tabs-mode: t + * indent-tabs-mode: nil * End: * - * ex: set ts=8 sts=4 sw=4 noet: + * ex: set ts=8 sts=4 sw=4 et: */