**** Alterations to Henry's code are...
****
**** Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
- **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, by Larry Wall and others
+ **** 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007 by Larry Wall and others
****
**** You may distribute under the terms of either the GNU General Public
**** License or the Artistic License, as specified in the README file.
typedef struct RExC_state_t {
U32 flags; /* are we folding, multilining? */
char *precomp; /* uncompiled string. */
- regexp *rx;
+ REGEXP *rx_sv; /* The SV that is the regexp. */
+ regexp *rx; /* perl core regexp structure */
+ regexp_internal *rxi; /* internal data for regexp object pprivate field */
char *start; /* Start of input for compile */
char *end; /* End of input for compile */
char *parse; /* Input-scan pointer. */
I32 whilem_seen; /* number of WHILEM in this expr */
regnode *emit_start; /* Start of emitted-code area */
+ regnode *emit_bound; /* First regnode outside of the allocated space */
regnode *emit; /* Code-emit pointer; ®dummy = don't = compiling */
I32 naughty; /* How bad is this pattern? */
I32 sawback; /* Did we see \1, ...? */
regnode **open_parens; /* pointers to open parens */
regnode **close_parens; /* pointers to close parens */
regnode *opend; /* END node in program */
- I32 utf8;
+ I32 utf8; /* whether the pattern is utf8 or not */
+ I32 orig_utf8; /* whether the pattern was originally in utf8 */
+ /* XXX use this for future optimisation of case
+ * where pattern must be upgraded to utf8. */
HV *charnames; /* cache of named sequences */
HV *paren_names; /* Paren names */
+
regnode **recurse; /* Recurse regops */
I32 recurse_count; /* Number of recurse regops */
#if ADD_TO_REGEXEC
#ifdef DEBUGGING
const char *lastparse;
I32 lastnum;
+ AV *paren_name_list; /* idx -> name */
#define RExC_lastparse (pRExC_state->lastparse)
#define RExC_lastnum (pRExC_state->lastnum)
+#define RExC_paren_name_list (pRExC_state->paren_name_list)
#endif
} RExC_state_t;
#define RExC_flags (pRExC_state->flags)
#define RExC_precomp (pRExC_state->precomp)
+#define RExC_rx_sv (pRExC_state->rx_sv)
#define RExC_rx (pRExC_state->rx)
+#define RExC_rxi (pRExC_state->rxi)
#define RExC_start (pRExC_state->start)
#define RExC_end (pRExC_state->end)
#define RExC_parse (pRExC_state->parse)
#define RExC_whilem_seen (pRExC_state->whilem_seen)
-#define RExC_offsets (pRExC_state->rx->offsets) /* I am not like the others */
+#ifdef RE_TRACK_PATTERN_OFFSETS
+#define RExC_offsets (pRExC_state->rxi->u.offsets) /* I am not like the others */
+#endif
#define RExC_emit (pRExC_state->emit)
#define RExC_emit_start (pRExC_state->emit_start)
+#define RExC_emit_bound (pRExC_state->emit_bound)
#define RExC_naughty (pRExC_state->naughty)
#define RExC_sawback (pRExC_state->sawback)
#define RExC_seen (pRExC_state->seen)
#define RExC_size (pRExC_state->size)
#define RExC_npar (pRExC_state->npar)
-#define RExC_cpar (pRExC_state->cpar)
#define RExC_nestroot (pRExC_state->nestroot)
#define RExC_extralen (pRExC_state->extralen)
#define RExC_seen_zerolen (pRExC_state->seen_zerolen)
#define RExC_seen_evals (pRExC_state->seen_evals)
#define RExC_utf8 (pRExC_state->utf8)
+#define RExC_orig_utf8 (pRExC_state->orig_utf8)
#define RExC_charnames (pRExC_state->charnames)
#define RExC_open_parens (pRExC_state->open_parens)
#define RExC_close_parens (pRExC_state->close_parens)
#define RExC_recurse (pRExC_state->recurse)
#define RExC_recurse_count (pRExC_state->recurse_count)
+
#define ISMULT1(c) ((c) == '*' || (c) == '+' || (c) == '?')
#define ISMULT2(s) ((*s) == '*' || (*s) == '+' || (*s) == '?' || \
((*s) == '{' && regcurly(s)))
* Flags to be passed up and down.
*/
#define WORST 0 /* Worst case. */
-#define HASWIDTH 0x1 /* Known to match non-null strings. */
-#define SIMPLE 0x2 /* Simple enough to be STAR/PLUS operand. */
-#define SPSTART 0x4 /* Starts with * or +. */
-#define TRYAGAIN 0x8 /* Weeded out a declaration. */
+#define HASWIDTH 0x01 /* Known to match non-null strings. */
+#define SIMPLE 0x02 /* Simple enough to be STAR/PLUS operand. */
+#define SPSTART 0x04 /* Starts with * or +. */
+#define TRYAGAIN 0x08 /* Weeded out a declaration. */
+#define POSTPONED 0x10 /* (?1),(?&name), (??{...}) or similar */
#define REG_NODE_NUM(x) ((x) ? (int)((x)-RExC_emit_start) : -1)
IV len = RExC_end - RExC_precomp; \
\
if (!SIZE_ONLY) \
- SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
+ SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
if (len > RegexLengthToShowInErrorMessages) { \
/* chop 10 shorter than the max, to ensure meaning of "..." */ \
len = RegexLengthToShowInErrorMessages - 10; \
*/
#define vFAIL(m) STMT_START { \
if (!SIZE_ONLY) \
- SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
+ SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
Simple_vFAIL(m); \
} STMT_END
*/
#define vFAIL2(m,a1) STMT_START { \
if (!SIZE_ONLY) \
- SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
+ SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
Simple_vFAIL2(m, a1); \
} STMT_END
*/
#define vFAIL3(m,a1,a2) STMT_START { \
if (!SIZE_ONLY) \
- SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx); \
+ SAVEDESTRUCTOR_X(clear_re,(void*)RExC_rx_sv); \
Simple_vFAIL3(m, a1, a2); \
} STMT_END
* Element 0 holds the number n.
* Position is 1 indexed.
*/
-
+#ifndef RE_TRACK_PATTERN_OFFSETS
+#define Set_Node_Offset_To_R(node,byte)
+#define Set_Node_Offset(node,byte)
+#define Set_Cur_Node_Offset
+#define Set_Node_Length_To_R(node,len)
+#define Set_Node_Length(node,len)
+#define Set_Node_Cur_Length(node)
+#define Node_Offset(n)
+#define Node_Length(n)
+#define Set_Node_Offset_Length(node,offset,len)
+#define ProgLen(ri) ri->u.proglen
+#define SetProgLen(ri,x) ri->u.proglen = x
+#else
+#define ProgLen(ri) ri->u.offsets[0]
+#define SetProgLen(ri,x) ri->u.offsets[0] = x
#define Set_Node_Offset_To_R(node,byte) STMT_START { \
if (! SIZE_ONLY) { \
MJD_OFFSET_DEBUG(("** (%d) offset of node %d is %d.\n", \
Set_Node_Offset_To_R((node)-RExC_emit_start, (offset)); \
Set_Node_Length_To_R((node)-RExC_emit_start, (len)); \
} STMT_END
-
+#endif
#if PERL_ENABLE_EXPERIMENTAL_REGEX_OPTIMISATIONS
#define EXPERIMENTAL_INPLACESCAN
-#endif
+#endif /*RE_TRACK_PATTERN_OFFSETS*/
-#define DEBUG_STUDYDATA(data,depth) \
-DEBUG_OPTIMISE_MORE_r(if(data){ \
+#define DEBUG_STUDYDATA(str,data,depth) \
+DEBUG_OPTIMISE_MORE_r(if(data){ \
PerlIO_printf(Perl_debug_log, \
- "%*s"/* Len:%"IVdf"/%"IVdf" */"Pos:%"IVdf"/%"IVdf \
- " Flags: %"IVdf" Whilem_c: %"IVdf" Lcp: %"IVdf" ", \
+ "%*s" str "Pos:%"IVdf"/%"IVdf \
+ " Flags: 0x%"UVXf" Whilem_c: %"IVdf" Lcp: %"IVdf" %s", \
(int)(depth)*2, "", \
(IV)((data)->pos_min), \
(IV)((data)->pos_delta), \
- (IV)((data)->flags), \
+ (UV)((data)->flags), \
(IV)((data)->whilem_c), \
- (IV)((data)->last_closep ? *((data)->last_closep) : -1) \
+ (IV)((data)->last_closep ? *((data)->last_closep) : -1), \
+ is_inf ? "INF " : "" \
); \
if ((data)->last_found) \
PerlIO_printf(Perl_debug_log, \
floating substrings if needed. */
STATIC void
-S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data, I32 *minlenp)
+S_scan_commit(pTHX_ const RExC_state_t *pRExC_state, scan_data_t *data, I32 *minlenp, int is_inf)
{
const STRLEN l = CHR_SVLEN(data->last_found);
const STRLEN old_l = CHR_SVLEN(*data->longest);
data->minlen_fixed=minlenp;
data->lookbehind_fixed=0;
}
- else {
+ else { /* *data->longest == data->longest_float */
data->offset_float_min = l ? data->last_start_min : data->pos_min;
data->offset_float_max = (l
? data->last_start_max
: data->pos_min + data->pos_delta);
- if ((U32)data->offset_float_max > (U32)I32_MAX)
+ if (is_inf || (U32)data->offset_float_max > (U32)I32_MAX)
data->offset_float_max = I32_MAX;
if (data->flags & SF_BEFORE_EOL)
data->flags
}
data->last_end = -1;
data->flags &= ~SF_BEFORE_EOL;
- DEBUG_STUDYDATA(data,0);
+ DEBUG_STUDYDATA("commit: ",data,0);
}
/* Can match anything (initialization) */
#ifdef DEBUGGING
/*
- dump_trie(trie)
- dump_trie_interim_list(trie,next_alloc)
- dump_trie_interim_table(trie,next_alloc)
+ dump_trie(trie,widecharmap,revcharmap)
+ dump_trie_interim_list(trie,widecharmap,revcharmap,next_alloc)
+ dump_trie_interim_table(trie,widecharmap,revcharmap,next_alloc)
These routines dump out a trie in a somewhat readable format.
The _interim_ variants are used for debugging the interim
*/
/*
- dump_trie(trie)
Dumps the final compressed table form of the trie to Perl_debug_log.
Used for debugging make_trie().
*/
STATIC void
-S_dump_trie(pTHX_ const struct _reg_trie_data *trie,U32 depth)
+S_dump_trie(pTHX_ const struct _reg_trie_data *trie, HV *widecharmap,
+ AV *revcharmap, U32 depth)
{
U32 state;
SV *sv=sv_newmortal();
- int colwidth= trie->widecharmap ? 6 : 4;
+ int colwidth= widecharmap ? 6 : 4;
GET_RE_DEBUG_FLAGS_DECL;
"Match","Base","Ofs" );
for( state = 0 ; state < trie->uniquecharcount ; state++ ) {
- SV ** const tmp = av_fetch( trie->revcharmap, state, 0);
+ SV ** const tmp = av_fetch( revcharmap, state, 0);
if ( tmp ) {
PerlIO_printf( Perl_debug_log, "%*s",
colwidth,
}
}
/*
- dump_trie_interim_list(trie,next_alloc)
Dumps a fully constructed but uncompressed trie in list form.
List tries normally only are used for construction when the number of
possible chars (trie->uniquecharcount) is very high.
Used for debugging make_trie().
*/
STATIC void
-S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie, U32 next_alloc,U32 depth)
+S_dump_trie_interim_list(pTHX_ const struct _reg_trie_data *trie,
+ HV *widecharmap, AV *revcharmap, U32 next_alloc,
+ U32 depth)
{
U32 state;
SV *sv=sv_newmortal();
- int colwidth= trie->widecharmap ? 6 : 4;
+ int colwidth= widecharmap ? 6 : 4;
GET_RE_DEBUG_FLAGS_DECL;
/* print out the table precompression. */
PerlIO_printf( Perl_debug_log, "%*sState :Word | Transition Data\n%*s%s",
);
}
for( charid = 1 ; charid <= TRIE_LIST_USED( state ) ; charid++ ) {
- SV ** const tmp = av_fetch( trie->revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0);
+ SV ** const tmp = av_fetch( revcharmap, TRIE_LIST_ITEM(state,charid).forid, 0);
if ( tmp ) {
PerlIO_printf( Perl_debug_log, "%*s:%3X=%4"UVXf" | ",
colwidth,
}
/*
- dump_trie_interim_table(trie,next_alloc)
Dumps a fully constructed but uncompressed trie in table form.
This is the normal DFA style state transition table, with a few
twists to facilitate compression later.
Used for debugging make_trie().
*/
STATIC void
-S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie, U32 next_alloc, U32 depth)
+S_dump_trie_interim_table(pTHX_ const struct _reg_trie_data *trie,
+ HV *widecharmap, AV *revcharmap, U32 next_alloc,
+ U32 depth)
{
U32 state;
U16 charid;
SV *sv=sv_newmortal();
- int colwidth= trie->widecharmap ? 6 : 4;
+ int colwidth= widecharmap ? 6 : 4;
GET_RE_DEBUG_FLAGS_DECL;
/*
PerlIO_printf( Perl_debug_log, "%*sChar : ",(int)depth * 2 + 2,"" );
for( charid = 0 ; charid < trie->uniquecharcount ; charid++ ) {
- SV ** const tmp = av_fetch( trie->revcharmap, charid, 0);
+ SV ** const tmp = av_fetch( revcharmap, charid, 0);
if ( tmp ) {
PerlIO_printf( Perl_debug_log, "%*s",
colwidth,
#define TRIE_STORE_REVCHAR \
STMT_START { \
- SV *tmp = newSVpvs(""); \
- if (UTF) SvUTF8_on(tmp); \
- Perl_sv_catpvf( aTHX_ tmp, "%c", (int)uvc ); \
- av_push( TRIE_REVCHARMAP(trie), tmp ); \
- } STMT_END
+ if (UTF) { \
+ SV *zlopp = newSV(2); \
+ unsigned char *flrbbbbb = (unsigned char *) SvPVX(zlopp); \
+ unsigned const char *const kapow = uvuni_to_utf8(flrbbbbb, uvc & 0xFF); \
+ SvCUR_set(zlopp, kapow - flrbbbbb); \
+ SvPOK_on(zlopp); \
+ SvUTF8_on(zlopp); \
+ av_push(revcharmap, zlopp); \
+ } else { \
+ char ooooff = (char)uvc; \
+ av_push(revcharmap, newSVpvn(&ooooff, 1)); \
+ } \
+ } STMT_END
#define TRIE_READ_CHAR STMT_START { \
wordlen++; \
/* store the word for dumping */ \
SV* tmp; \
if (OP(noper) != NOTHING) \
- tmp = newSVpvn(STRING(noper), STR_LEN(noper)); \
+ tmp = newSVpvn_utf8(STRING(noper), STR_LEN(noper), UTF); \
else \
- tmp = newSVpvn( "", 0 ); \
- if ( UTF ) SvUTF8_on( tmp ); \
- av_push( trie->words, tmp ); \
+ tmp = newSVpvn_utf8( "", 0, UTF ); \
+ av_push( trie_words, tmp ); \
}); \
\
curword++; \
\
if ( noper_next < tail ) { \
if (!trie->jump) \
- Newxz( trie->jump, word_count + 1, U16); \
+ trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, sizeof(U16) ); \
trie->jump[curword] = (U16)(noper_next - convert); \
if (!jumper) \
jumper = noper_next; \
/* we only allocate the nextword buffer when there */\
/* a dupe, so first time we have to do the allocation */\
if (!trie->nextword) \
- Newxz( trie->nextword, word_count + 1, U16); \
+ trie->nextword = (U16 *) \
+ PerlMemShared_calloc( word_count + 1, sizeof(U16)); \
while ( trie->nextword[dupe] ) \
dupe= trie->nextword[dupe]; \
trie->nextword[dupe]= curword; \
dVAR;
/* first pass, loop through and scan words */
reg_trie_data *trie;
+ HV *widecharmap = NULL;
+ AV *revcharmap = newAV();
regnode *cur;
const U32 uniflags = UTF8_ALLOW_DEFAULT;
STRLEN len = 0;
)
);
- const U32 data_slot = add_data( pRExC_state, 1, "t" );
- SV *re_trie_maxbuff;
-#ifndef DEBUGGING
- /* these are only used during construction but are useful during
- * debugging so we store them in the struct when debugging.
+#ifdef DEBUGGING
+ const U32 data_slot = add_data( pRExC_state, 4, "tuuu" );
+ AV *trie_words = NULL;
+ /* along with revcharmap, this only used during construction but both are
+ * useful during debugging so we store them in the struct when debugging.
*/
+#else
+ const U32 data_slot = add_data( pRExC_state, 2, "tu" );
STRLEN trie_charcount=0;
- AV *trie_revcharmap;
#endif
+ SV *re_trie_maxbuff;
GET_RE_DEBUG_FLAGS_DECL;
#ifndef DEBUGGING
PERL_UNUSED_ARG(depth);
#endif
- Newxz( trie, 1, reg_trie_data );
+ trie = (reg_trie_data *) PerlMemShared_calloc( 1, sizeof(reg_trie_data) );
trie->refcount = 1;
trie->startstate = 1;
trie->wordcount = word_count;
- RExC_rx->data->data[ data_slot ] = (void*)trie;
- Newxz( trie->charmap, 256, U16 );
+ RExC_rxi->data->data[ data_slot ] = (void*)trie;
+ trie->charmap = (U16 *) PerlMemShared_calloc( 256, sizeof(U16) );
if (!(UTF && folder))
- Newxz( trie->bitmap, ANYOF_BITMAP_SIZE, char );
+ trie->bitmap = (char *) PerlMemShared_calloc( ANYOF_BITMAP_SIZE, 1 );
DEBUG_r({
- trie->words = newAV();
+ trie_words = newAV();
});
- TRIE_REVCHARMAP(trie) = newAV();
re_trie_maxbuff = get_sv(RE_TRIE_MAXBUF_NAME, 1);
if (!SvIOK(re_trie_maxbuff)) {
have unique chars.
We use an array of integers to represent the character codes 0..255
- (trie->charmap) and we use a an HV* to store unicode characters. We use the
+ (trie->charmap) and we use a an HV* to store Unicode characters. We use the
native representation of the character value as the key and IV's for the
coded index.
U8 foldbuf[ UTF8_MAXBYTES_CASE + 1 ];
const U8 *scan = (U8*)NULL;
U32 wordlen = 0; /* required init */
- STRLEN chars=0;
+ STRLEN chars = 0;
+ bool set_bit = trie->bitmap ? 1 : 0; /*store the first char in the bitmap?*/
if (OP(noper) == NOTHING) {
trie->minlen= 0;
continue;
}
- if (trie->bitmap) {
- TRIE_BITMAP_SET(trie,*uc);
- if ( folder ) TRIE_BITMAP_SET(trie,folder[ *uc ]);
- }
+ if ( set_bit ) /* bitmap only alloced when !(UTF&&Folding) */
+ TRIE_BITMAP_SET(trie,*uc); /* store the raw first byte
+ regardless of encoding */
+
for ( ; uc < e ; uc += len ) {
TRIE_CHARCOUNT(trie)++;
TRIE_READ_CHAR;
trie->charmap[ folder[ uvc ] ] = trie->charmap[ uvc ];
TRIE_STORE_REVCHAR;
}
+ if ( set_bit ) {
+ /* store the codepoint in the bitmap, and if its ascii
+ also store its folded equivelent. */
+ TRIE_BITMAP_SET(trie,uvc);
+
+ /* store the folded codepoint */
+ if ( folder ) TRIE_BITMAP_SET(trie,folder[ uvc ]);
+
+ if ( !UTF ) {
+ /* store first byte of utf8 representation of
+ codepoints in the 127 < uvc < 256 range */
+ if (127 < uvc && uvc < 192) {
+ TRIE_BITMAP_SET(trie,194);
+ } else if (191 < uvc ) {
+ TRIE_BITMAP_SET(trie,195);
+ /* && uvc < 256 -- we know uvc is < 256 already */
+ }
+ }
+ set_bit = 0; /* We've done our bit :-) */
+ }
} else {
SV** svpp;
- if ( !trie->widecharmap )
- trie->widecharmap = newHV();
+ if ( !widecharmap )
+ widecharmap = newHV();
- svpp = hv_fetch( trie->widecharmap, (char*)&uvc, sizeof( UV ), 1 );
+ svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 1 );
if ( !svpp )
Perl_croak( aTHX_ "error creating/fetching widecharmap entry for 0x%"UVXf, uvc );
DEBUG_TRIE_COMPILE_r(
PerlIO_printf( Perl_debug_log, "%*sTRIE(%s): W:%d C:%d Uq:%d Min:%d Max:%d\n",
(int)depth * 2 + 2,"",
- ( trie->widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
+ ( widecharmap ? "UTF8" : "NATIVE" ), (int)word_count,
(int)TRIE_CHARCOUNT(trie), trie->uniquecharcount,
(int)trie->minlen, (int)trie->maxlen )
);
- Newxz( trie->wordlen, word_count, U32 );
+ trie->wordlen = (U32 *) PerlMemShared_calloc( word_count, sizeof(U32) );
/*
We now know what we are dealing with in terms of unique chars and
DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
"%*sCompiling trie using list compiler\n",
(int)depth * 2 + 2, ""));
-
- Newxz( trie->states, TRIE_CHARCOUNT(trie) + 2, reg_trie_state );
+
+ trie->states = (reg_trie_state *)
+ PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
+ sizeof(reg_trie_state) );
TRIE_LIST_NEW(1);
next_alloc = 2;
if ( uvc < 256 ) {
charid = trie->charmap[ uvc ];
} else {
- SV** const svpp = hv_fetch( trie->widecharmap, (char*)&uvc, sizeof( UV ), 0);
+ SV** const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
if ( !svpp ) {
charid = 0;
} else {
/* next alloc is the NEXT state to be allocated */
trie->statecount = next_alloc;
- Renew( trie->states, next_alloc, reg_trie_state );
+ trie->states = (reg_trie_state *)
+ PerlMemShared_realloc( trie->states,
+ next_alloc
+ * sizeof(reg_trie_state) );
/* and now dump it out before we compress it */
- DEBUG_TRIE_COMPILE_MORE_r(
- dump_trie_interim_list(trie,next_alloc,depth+1)
+ DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_list(trie, widecharmap,
+ revcharmap, next_alloc,
+ depth+1)
);
- Newxz( trie->trans, transcount ,reg_trie_trans );
+ trie->trans = (reg_trie_trans *)
+ PerlMemShared_calloc( transcount, sizeof(reg_trie_trans) );
{
U32 state;
U32 tp = 0;
}
if ( transcount < tp + maxid - minid + 1) {
transcount *= 2;
- Renew( trie->trans, transcount, reg_trie_trans );
+ trie->trans = (reg_trie_trans *)
+ PerlMemShared_realloc( trie->trans,
+ transcount
+ * sizeof(reg_trie_trans) );
Zero( trie->trans + (transcount / 2), transcount / 2 , reg_trie_trans );
}
base = trie->uniquecharcount + tp - minid;
"%*sCompiling trie using table compiler\n",
(int)depth * 2 + 2, ""));
- Newxz( trie->trans, ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1,
- reg_trie_trans );
- Newxz( trie->states, TRIE_CHARCOUNT(trie) + 2, reg_trie_state );
+ trie->trans = (reg_trie_trans *)
+ PerlMemShared_calloc( ( TRIE_CHARCOUNT(trie) + 1 )
+ * trie->uniquecharcount + 1,
+ sizeof(reg_trie_trans) );
+ trie->states = (reg_trie_state *)
+ PerlMemShared_calloc( TRIE_CHARCOUNT(trie) + 2,
+ sizeof(reg_trie_state) );
next_alloc = trie->uniquecharcount + 1;
if ( uvc < 256 ) {
charid = trie->charmap[ uvc ];
} else {
- SV* const * const svpp = hv_fetch( trie->widecharmap, (char*)&uvc, sizeof( UV ), 0);
+ SV* const * const svpp = hv_fetch( widecharmap, (char*)&uvc, sizeof( UV ), 0);
charid = svpp ? (U16)SvIV(*svpp) : 0;
}
if ( charid ) {
} /* end second pass */
/* and now dump it out before we compress it */
- DEBUG_TRIE_COMPILE_MORE_r(
- dump_trie_interim_table(trie,next_alloc,depth+1)
- );
+ DEBUG_TRIE_COMPILE_MORE_r(dump_trie_interim_table(trie, widecharmap,
+ revcharmap,
+ next_alloc, depth+1));
{
/*
}
}
trie->lasttrans = pos + 1;
- Renew( trie->states, laststate, reg_trie_state);
+ trie->states = (reg_trie_state *)
+ PerlMemShared_realloc( trie->states, laststate
+ * sizeof(reg_trie_state) );
DEBUG_TRIE_COMPILE_MORE_r(
PerlIO_printf( Perl_debug_log,
"%*sAlloc: %d Orig: %"IVdf" elements, Final:%"IVdf". Savings of %%%5.2f\n",
(UV)trie->lasttrans)
);
/* resize the trans array to remove unused space */
- Renew( trie->trans, trie->lasttrans, reg_trie_trans);
+ trie->trans = (reg_trie_trans *)
+ PerlMemShared_realloc( trie->trans, trie->lasttrans
+ * sizeof(reg_trie_trans) );
/* and now dump out the compressed format */
- DEBUG_TRIE_COMPILE_r(
- dump_trie(trie,depth+1)
- );
+ DEBUG_TRIE_COMPILE_r(dump_trie(trie, widecharmap, revcharmap, depth+1));
{ /* Modify the program and insert the new TRIE node*/
U8 nodetype =(U8)(flags & 0xFF);
#ifdef DEBUGGING
regnode *optimize = NULL;
+#ifdef RE_TRACK_PATTERN_OFFSETS
+
U32 mjd_offset = 0;
U32 mjd_nodelen = 0;
-#endif
+#endif /* RE_TRACK_PATTERN_OFFSETS */
+#endif /* DEBUGGING */
/*
This means we convert either the first branch or the first Exact,
depending on whether the thing following (in 'last') is a branch
if ( first != startbranch || OP( last ) == BRANCH ) {
/* branch sub-chain */
NEXT_OFF( first ) = (U16)(last - first);
+#ifdef RE_TRACK_PATTERN_OFFSETS
DEBUG_r({
mjd_offset= Node_Offset((convert));
mjd_nodelen= Node_Length((convert));
});
+#endif
/* whole branch chain */
- } else {
+ }
+#ifdef RE_TRACK_PATTERN_OFFSETS
+ else {
DEBUG_r({
const regnode *nop = NEXTOPER( convert );
mjd_offset= Node_Offset((nop));
mjd_nodelen= Node_Length((nop));
});
}
-
DEBUG_OPTIMISE_r(
PerlIO_printf(Perl_debug_log, "%*sMJD offset:%"UVuf" MJD length:%"UVuf"\n",
(int)depth * 2 + 2, "",
(UV)mjd_offset, (UV)mjd_nodelen)
);
-
+#endif
/* But first we check to see if there is a common prefix we can
split out as an EXACT and put in front of the TRIE node. */
trie->startstate= 1;
- if ( trie->bitmap && !trie->widecharmap && !trie->jump ) {
+ if ( trie->bitmap && !widecharmap && !trie->jump ) {
U32 state;
for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
U32 ofs = 0;
trie->trans[ base + ofs - trie->uniquecharcount ].check == state )
{
if ( ++count > 1 ) {
- SV **tmp = av_fetch( TRIE_REVCHARMAP(trie), ofs, 0);
+ SV **tmp = av_fetch( revcharmap, ofs, 0);
const U8 *ch = (U8*)SvPV_nolen_const( *tmp );
if ( state == 1 ) break;
if ( count == 2 ) {
(int)depth * 2 + 2, "",
(UV)state));
if (idx >= 0) {
- SV ** const tmp = av_fetch( TRIE_REVCHARMAP(trie), idx, 0);
+ SV ** const tmp = av_fetch( revcharmap, idx, 0);
const U8 * const ch = (U8*)SvPV_nolen_const( *tmp );
TRIE_BITMAP_SET(trie,*ch);
}
}
if ( count == 1 ) {
- SV **tmp = av_fetch( TRIE_REVCHARMAP(trie), idx, 0);
- char *ch = SvPV_nolen( *tmp );
+ SV **tmp = av_fetch( revcharmap, idx, 0);
+ STRLEN len;
+ char *ch = SvPV( *tmp, len );
DEBUG_OPTIMISE_r({
SV *sv=sv_newmortal();
PerlIO_printf( Perl_debug_log,
str=STRING(convert);
STR_LEN(convert)=0;
}
- while (*ch) {
+ STR_LEN(convert) += len;
+ while (len--)
*str++ = *ch++;
- STR_LEN(convert)++;
- }
-
} else {
#ifdef DEBUGGING
if (state>1)
trie->startstate = state;
trie->minlen -= (state - 1);
trie->maxlen -= (state - 1);
- DEBUG_r({
- regnode *fix = convert;
- U32 word = trie->wordcount;
- mjd_nodelen++;
- Set_Node_Offset_Length(convert, mjd_offset, state - 1);
- while( ++fix < n ) {
- Set_Node_Offset_Length(fix, 0, 0);
- }
- while (word--) {
- SV ** const tmp = av_fetch( trie->words, word, 0 );
- if (tmp) {
- if ( STR_LEN(convert) <= SvCUR(*tmp) )
- sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
- else
- sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
- }
- }
- });
+#ifdef DEBUGGING
+ /* At least the UNICOS C compiler choked on this
+ * being argument to DEBUG_r(), so let's just have
+ * it right here. */
+ if (
+#ifdef PERL_EXT_RE_BUILD
+ 1
+#else
+ DEBUG_r_TEST
+#endif
+ ) {
+ regnode *fix = convert;
+ U32 word = trie->wordcount;
+ mjd_nodelen++;
+ Set_Node_Offset_Length(convert, mjd_offset, state - 1);
+ while( ++fix < n ) {
+ Set_Node_Offset_Length(fix, 0, 0);
+ }
+ while (word--) {
+ SV ** const tmp = av_fetch( trie_words, word, 0 );
+ if (tmp) {
+ if ( STR_LEN(convert) <= SvCUR(*tmp) )
+ sv_chop(*tmp, SvPV_nolen(*tmp) + STR_LEN(convert));
+ else
+ sv_chop(*tmp, SvPV_nolen(*tmp) + SvCUR(*tmp));
+ }
+ }
+ }
+#endif
if (trie->maxlen) {
convert = n;
} else {
{
OP( convert ) = TRIEC;
Copy(trie->bitmap, ((struct regnode_charclass *)convert)->bitmap, ANYOF_BITMAP_SIZE, char);
- Safefree(trie->bitmap);
+ PerlMemShared_free(trie->bitmap);
trie->bitmap= NULL;
} else
OP( convert ) = TRIE;
/* needed for dumping*/
DEBUG_r(if (optimize) {
regnode *opt = convert;
+
while ( ++opt < optimize) {
Set_Node_Offset_Length(opt,0,0);
}
Set_Node_Offset_Length(convert,mjd_offset,mjd_nodelen);
});
} /* end node insert */
-#ifndef DEBUGGING
- SvREFCNT_dec(TRIE_REVCHARMAP(trie));
+ RExC_rxi->data->data[ data_slot + 1 ] = (void*)widecharmap;
+#ifdef DEBUGGING
+ RExC_rxi->data->data[ data_slot + TRIE_WORDS_OFFSET ] = (void*)trie_words;
+ RExC_rxi->data->data[ data_slot + 3 ] = (void*)revcharmap;
+#else
+ SvREFCNT_dec(revcharmap);
#endif
return trie->jump
? MADE_JUMP_TRIE
try 'g' and succeed, prodceding to match 'cdgu'.
*/
/* add a fail transition */
- reg_trie_data *trie=(reg_trie_data *)RExC_rx->data->data[ARG(source)];
+ const U32 trie_offset = ARG(source);
+ reg_trie_data *trie=(reg_trie_data *)RExC_rxi->data->data[trie_offset];
U32 *q;
const U32 ucharcount = trie->uniquecharcount;
const U32 numstates = trie->statecount;
ARG_SET( stclass, data_slot );
- Newxz( aho, 1, reg_ac_data );
- RExC_rx->data->data[ data_slot ] = (void*)aho;
- aho->trie=trie;
- aho->states=(reg_trie_state *)savepvn((const char*)trie->states,
- numstates * sizeof(reg_trie_state));
+ aho = (reg_ac_data *) PerlMemShared_calloc( 1, sizeof(reg_ac_data) );
+ RExC_rxi->data->data[ data_slot ] = (void*)aho;
+ aho->trie=trie_offset;
+ aho->states=(reg_trie_state *)PerlMemShared_malloc( numstates * sizeof(reg_trie_state) );
+ Copy( trie->states, aho->states, numstates, reg_trie_state );
Newxz( q, numstates, U32);
- Newxz( aho->fail, numstates, U32 );
+ aho->fail = (U32 *) PerlMemShared_calloc( numstates, sizeof(U32) );
aho->refcount = 1;
fail = aho->fail;
/* initialize fail[0..1] to be 1 so that we always have
I32 stop; /* what stopparen do we use */
} scan_frame;
+
+#define SCAN_COMMIT(s, data, m) scan_commit(s, data, m, is_inf)
+
+#define CASE_SYNST_FNC(nAmE) \
+case nAmE: \
+ if (flags & SCF_DO_STCLASS_AND) { \
+ for (value = 0; value < 256; value++) \
+ if (!is_ ## nAmE ## _cp(value)) \
+ ANYOF_BITMAP_CLEAR(data->start_class, value); \
+ } \
+ else { \
+ for (value = 0; value < 256; value++) \
+ if (is_ ## nAmE ## _cp(value)) \
+ ANYOF_BITMAP_SET(data->start_class, value); \
+ } \
+ break; \
+case N ## nAmE: \
+ if (flags & SCF_DO_STCLASS_AND) { \
+ for (value = 0; value < 256; value++) \
+ if (is_ ## nAmE ## _cp(value)) \
+ ANYOF_BITMAP_CLEAR(data->start_class, value); \
+ } \
+ else { \
+ for (value = 0; value < 256; value++) \
+ if (!is_ ## nAmE ## _cp(value)) \
+ ANYOF_BITMAP_SET(data->start_class, value); \
+ } \
+ break
+
+
+
STATIC I32
S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
I32 *minlenp, I32 *deltap,
fake_study_recurse:
while ( scan && OP(scan) != END && scan < last ){
/* Peephole optimizer: */
- DEBUG_STUDYDATA(data,depth);
+ DEBUG_STUDYDATA("Peep:", data,depth);
DEBUG_PEEP("Peep",scan,depth);
JOIN_EXACT(scan,&min,0);
regnode * const startbranch=scan;
if (flags & SCF_DO_SUBSTR)
- scan_commit(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
+ SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
if (flags & SCF_DO_STCLASS)
cl_init_zero(pRExC_state, &accum);
last = cur;
}
} else {
- if ( last ) {
+/*
+ Currently we assume that the trie can handle unicode and ascii
+ matches fold cased matches. If this proves true then the following
+ define will prevent tries in this situation.
+
+ #define TRIE_TYPE_IS_SAFE (UTF || optype==EXACT)
+*/
+#define TRIE_TYPE_IS_SAFE 1
+ if ( last && TRIE_TYPE_IS_SAFE ) {
make_trie( pRExC_state,
startbranch, first, cur, tail, count,
optype, depth+1 );
"", SvPV_nolen_const( mysv ),REG_NODE_NUM(cur));
});
- if ( last ) {
+
+ if ( last && TRIE_TYPE_IS_SAFE ) {
made= make_trie( pRExC_state, startbranch, first, scan, tail, count, optype, depth+1 );
#ifdef TRIE_STUDY_OPT
if ( ((made == MADE_EXACT_TRIE &&
end = RExC_close_parens[paren-1];
} else {
paren = 0;
- start = RExC_rx->program + 1;
+ start = RExC_rxi->program + 1;
end = RExC_opend;
}
if (!recursed) {
Newx(newframe,1,scan_frame);
} else {
if (flags & SCF_DO_SUBSTR) {
- scan_commit(pRExC_state,data,minlenp);
+ SCAN_COMMIT(pRExC_state,data,minlenp);
data->longest = &(data->longest_float);
}
is_inf = is_inf_internal = 1;
/* Search for fixed substrings supports EXACT only. */
if (flags & SCF_DO_SUBSTR) {
assert(data);
- scan_commit(pRExC_state, data, minlenp);
+ SCAN_COMMIT(pRExC_state, data, minlenp);
}
if (UTF) {
const U8 * const s = (U8 *)STRING(scan);
is_inf = is_inf_internal = 1;
scan = regnext(scan);
if (flags & SCF_DO_SUBSTR) {
- scan_commit(pRExC_state, data, minlenp); /* Cannot extend fixed substrings */
+ SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot extend fixed substrings */
data->longest = &(data->longest_float);
}
goto optimize_curly_tail;
next_is_eval = (OP(scan) == EVAL);
do_curly:
if (flags & SCF_DO_SUBSTR) {
- if (mincount == 0) scan_commit(pRExC_state,data,minlenp); /* Cannot extend fixed substrings */
+ if (mincount == 0) SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot extend fixed substrings */
pos_before = data->pos_min;
}
if (data) {
l -= old;
/* Get the added string: */
- last_str = newSVpvn(s + old, l);
- if (UTF)
- SvUTF8_on(last_str);
+ last_str = newSVpvn_utf8(s + old, l, UTF);
if (deltanext == 0 && pos_before == b) {
/* What was added is a constant string */
if (mincount > 1) {
SvUTF8(sv) && SvMAGICAL(sv) ?
mg_find(sv, PERL_MAGIC_utf8) : NULL;
if (mg && mg->mg_len >= 0)
- mg->mg_len += CHR_SVLEN(last_str);
+ mg->mg_len += CHR_SVLEN(last_str) - l;
}
data->last_end += l * (mincount - 1);
}
if (mincount != maxcount) {
/* Cannot extend fixed substrings found inside
the group. */
- scan_commit(pRExC_state,data,minlenp);
+ SCAN_COMMIT(pRExC_state,data,minlenp);
if (mincount && last_str) {
SV * const sv = data->last_found;
MAGIC * const mg = SvUTF8(sv) && SvMAGICAL(sv) ?
continue;
default: /* REF and CLUMP only? */
if (flags & SCF_DO_SUBSTR) {
- scan_commit(pRExC_state,data,minlenp); /* Cannot expect anything... */
+ SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
data->longest = &(data->longest_float);
}
is_inf = is_inf_internal = 1;
break;
}
}
+ else if (OP(scan) == LNBREAK) {
+ if (flags & SCF_DO_STCLASS) {
+ int value = 0;
+ data->start_class->flags &= ~ANYOF_EOS; /* No match on empty */
+ if (flags & SCF_DO_STCLASS_AND) {
+ for (value = 0; value < 256; value++)
+ if (!is_VERTWS_cp(value))
+ ANYOF_BITMAP_CLEAR(data->start_class, value);
+ }
+ else {
+ for (value = 0; value < 256; value++)
+ if (is_VERTWS_cp(value))
+ ANYOF_BITMAP_SET(data->start_class, value);
+ }
+ if (flags & SCF_DO_STCLASS_OR)
+ cl_and(data->start_class, and_withp);
+ flags &= ~SCF_DO_STCLASS;
+ }
+ min += 1;
+ delta += 1;
+ if (flags & SCF_DO_SUBSTR) {
+ SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
+ data->pos_min += 1;
+ data->pos_delta += 1;
+ data->longest = &(data->longest_float);
+ }
+
+ }
+ else if (OP(scan) == FOLDCHAR) {
+ int d = ARG(scan)==0xDF ? 1 : 2;
+ flags &= ~SCF_DO_STCLASS;
+ min += 1;
+ delta += d;
+ if (flags & SCF_DO_SUBSTR) {
+ SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
+ data->pos_min += 1;
+ data->pos_delta += d;
+ data->longest = &(data->longest_float);
+ }
+ }
else if (strchr((const char*)PL_simple,OP(scan))) {
int value = 0;
if (flags & SCF_DO_SUBSTR) {
- scan_commit(pRExC_state,data,minlenp);
+ SCAN_COMMIT(pRExC_state,data,minlenp);
data->pos_min++;
}
min++;
}
}
break;
+ CASE_SYNST_FNC(VERTWS);
+ CASE_SYNST_FNC(HORIZWS);
+
}
if (flags & SCF_DO_STCLASS_OR)
cl_and(data->start_class, and_withp);
if ((flags & SCF_DO_SUBSTR) && data->last_found) {
f |= SCF_DO_SUBSTR;
if (scan->flags)
- scan_commit(pRExC_state, &data_fake,minlenp);
+ SCAN_COMMIT(pRExC_state, &data_fake,minlenp);
data_fake.last_found=newSVsv(data->last_found);
}
}
if ((flags & SCF_DO_SUBSTR) && data_fake.last_found) {
if (RExC_rx->minlen<*minnextp)
RExC_rx->minlen=*minnextp;
- scan_commit(pRExC_state, &data_fake, minnextp);
+ SCAN_COMMIT(pRExC_state, &data_fake, minnextp);
SvREFCNT_dec(data_fake.last_found);
if ( data_fake.minlen_fixed != minlenp )
}
else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
if (flags & SCF_DO_SUBSTR) {
- scan_commit(pRExC_state,data,minlenp);
+ SCAN_COMMIT(pRExC_state,data,minlenp);
flags &= ~SCF_DO_SUBSTR;
}
if (data && OP(scan)==ACCEPT) {
else if (OP(scan) == LOGICAL && scan->flags == 2) /* Embedded follows */
{
if (flags & SCF_DO_SUBSTR) {
- scan_commit(pRExC_state,data,minlenp);
+ SCAN_COMMIT(pRExC_state,data,minlenp);
data->longest = &(data->longest_float);
}
is_inf = is_inf_internal = 1;
check there too. */
regnode *trie_node= scan;
regnode *tail= regnext(scan);
- reg_trie_data *trie = (reg_trie_data*)RExC_rx->data->data[ ARG(scan) ];
+ reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
I32 max1 = 0, min1 = I32_MAX;
struct regnode_charclass_class accum;
if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
- scan_commit(pRExC_state, data,minlenp); /* Cannot merge strings after this. */
+ SCAN_COMMIT(pRExC_state, data,minlenp); /* Cannot merge strings after this. */
if (flags & SCF_DO_STCLASS)
cl_init_zero(pRExC_state, &accum);
}
#else
else if (PL_regkind[OP(scan)] == TRIE) {
- reg_trie_data *trie = (reg_trie_data*)RExC_rx->data->data[ ARG(scan) ];
+ reg_trie_data *trie = (reg_trie_data*)RExC_rxi->data->data[ ARG(scan) ];
U8*bang=NULL;
min += trie->minlen;
delta += (trie->maxlen - trie->minlen);
flags &= ~SCF_DO_STCLASS; /* xxx */
if (flags & SCF_DO_SUBSTR) {
- scan_commit(pRExC_state,data,minlenp); /* Cannot expect anything... */
+ SCAN_COMMIT(pRExC_state,data,minlenp); /* Cannot expect anything... */
data->pos_min += trie->minlen;
data->pos_delta += (trie->maxlen - trie->minlen);
if (trie->maxlen != trie->minlen)
}
#endif /* old or new */
#endif /* TRIE_STUDY_OPT */
+
/* Else: zero-length, ignore. */
scan = regnext(scan);
}
finish:
assert(!frame);
+ DEBUG_STUDYDATA("pre-fin:",data,depth);
*scanp = scan;
*deltap = is_inf_internal ? I32_MAX : delta;
if (flags & SCF_TRIE_RESTUDY)
data->flags |= SCF_TRIE_RESTUDY;
- DEBUG_STUDYDATA(data,depth);
+ DEBUG_STUDYDATA("post-fin:",data,depth);
return min < stopmin ? min : stopmin;
}
-STATIC I32
-S_add_data(RExC_state_t *pRExC_state, I32 n, const char *s)
+STATIC U32
+S_add_data(RExC_state_t *pRExC_state, U32 n, const char *s)
{
- if (RExC_rx->data) {
- const U32 count = RExC_rx->data->count;
- Renewc(RExC_rx->data,
- sizeof(*RExC_rx->data) + sizeof(void*) * (count + n - 1),
- char, struct reg_data);
- Renew(RExC_rx->data->what, count + n, U8);
- RExC_rx->data->count += n;
- }
- else {
- Newxc(RExC_rx->data, sizeof(*RExC_rx->data) + sizeof(void*) * (n - 1),
- char, struct reg_data);
- Newx(RExC_rx->data->what, n, U8);
- RExC_rx->data->count = n;
- }
- Copy(s, RExC_rx->data->what + RExC_rx->data->count - n, n, U8);
- return RExC_rx->data->count - n;
+ U32 count = RExC_rxi->data ? RExC_rxi->data->count : 0;
+
+ Renewc(RExC_rxi->data,
+ sizeof(*RExC_rxi->data) + sizeof(void*) * (count + n - 1),
+ char, struct reg_data);
+ if(count)
+ Renew(RExC_rxi->data->what, count + n, U8);
+ else
+ Newx(RExC_rxi->data->what, n, U8);
+ RExC_rxi->data->count = count + n;
+ Copy(s, RExC_rxi->data->what + count, n, U8);
+ return count;
}
+/*XXX: todo make this not included in a non debugging perl */
#ifndef PERL_IN_XSUB_RE
void
Perl_reginitcolors(pTHX)
extern const struct regexp_engine my_reg_engine;
#define RE_ENGINE_PTR &my_reg_engine
#endif
-/* these make a few things look better, to avoid indentation */
-#define BEGIN_BLOCK {
-#define END_BLOCK }
-
-regexp *
-Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
+
+#ifndef PERL_IN_XSUB_RE
+REGEXP *
+Perl_pregcomp(pTHX_ const SV * const pattern, const U32 flags)
{
dVAR;
- GET_RE_DEBUG_FLAGS_DECL;
- DEBUG_r(if (!PL_colorset) reginitcolors());
-#ifndef PERL_IN_XSUB_RE
- BEGIN_BLOCK
+ HV * const table = GvHV(PL_hintgv);
/* Dispatch a request to compile a regexp to correct
regexp engine. */
- HV * const table = GvHV(PL_hintgv);
if (table) {
SV **ptr= hv_fetchs(table, "regcomp", FALSE);
+ GET_RE_DEBUG_FLAGS_DECL;
if (ptr && SvIOK(*ptr) && SvIV(*ptr)) {
const regexp_engine *eng=INT2PTR(regexp_engine*,SvIV(*ptr));
DEBUG_COMPILE_r({
PerlIO_printf(Perl_debug_log, "Using engine %"UVxf"\n",
SvIV(*ptr));
});
- return CALLREGCOMP_ENG(eng, exp, xend, pm);
+ return CALLREGCOMP_ENG(eng, pattern, flags);
}
}
- END_BLOCK
+ return Perl_re_compile(aTHX_ pattern, flags);
+}
#endif
- BEGIN_BLOCK
- register regexp *r;
+
+REGEXP *
+Perl_re_compile(pTHX_ const SV * const pattern, U32 pm_flags)
+{
+ dVAR;
+ REGEXP *rx;
+ struct regexp *r;
+ register regexp_internal *ri;
+ STRLEN plen;
+ char* exp = SvPV((SV*)pattern, plen);
+ char* xend = exp + plen;
regnode *scan;
- regnode *first;
I32 flags;
I32 minlen = 0;
I32 sawplus = 0;
int restudied= 0;
RExC_state_t copyRExC_state;
#endif
- if (exp == NULL)
- FAIL("NULL regexp argument");
+ GET_RE_DEBUG_FLAGS_DECL;
+ DEBUG_r(if (!PL_colorset) reginitcolors());
- RExC_utf8 = pm->op_pmdynflags & PMdf_CMP_UTF8;
+ RExC_utf8 = RExC_orig_utf8 = SvUTF8(pattern);
- RExC_precomp = exp;
DEBUG_COMPILE_r({
SV *dsv= sv_newmortal();
RE_PV_QUOTED_DECL(s, RExC_utf8,
- dsv, RExC_precomp, (xend - exp), 60);
+ dsv, exp, plen, 60);
PerlIO_printf(Perl_debug_log, "%sCompiling REx%s %s\n",
PL_colors[4],PL_colors[5],s);
});
- RExC_flags = pm->op_pmflags;
+
+redo_first_pass:
+ RExC_precomp = exp;
+ RExC_flags = pm_flags;
RExC_sawback = 0;
RExC_seen = 0;
RExC_end = xend;
RExC_naughty = 0;
RExC_npar = 1;
- RExC_cpar = 1;
RExC_nestroot = 0;
RExC_size = 0L;
RExC_emit = &PL_regdummy;
RExC_close_parens = NULL;
RExC_opend = NULL;
RExC_paren_names = NULL;
+#ifdef DEBUGGING
+ RExC_paren_name_list = NULL;
+#endif
RExC_recurse = NULL;
RExC_recurse_count = 0;
RExC_precomp = NULL;
return(NULL);
}
+ if (RExC_utf8 && !RExC_orig_utf8) {
+ /* It's possible to write a regexp in ascii that represents Unicode
+ codepoints outside of the byte range, such as via \x{100}. If we
+ detect such a sequence we have to convert the entire pattern to utf8
+ and then recompile, as our sizing calculation will have been based
+ on 1 byte == 1 character, but we will need to use utf8 to encode
+ at least some part of the pattern, and therefore must convert the whole
+ thing.
+ XXX: somehow figure out how to make this less expensive...
+ -- dmq */
+ STRLEN len = plen;
+ DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log,
+ "UTF8 mismatch! Converting to utf8 for resizing and compile\n"));
+ exp = (char*)Perl_bytes_to_utf8(aTHX_ (U8*)exp, &len);
+ xend = exp + len;
+ RExC_orig_utf8 = RExC_utf8;
+ SAVEFREEPV(exp);
+ goto redo_first_pass;
+ }
DEBUG_PARSE_r({
PerlIO_printf(Perl_debug_log,
"Required size %"IVdf" nodes\n"
if (RExC_whilem_seen > 15)
RExC_whilem_seen = 15;
-#ifdef DEBUGGING
- /* Make room for a sentinel value at the end of the program */
- RExC_size++;
-#endif
-
/* Allocate space and zero-initialize. Note, the two step process
of zeroing when in debug mode, thus anything assigned has to
happen after that */
- Newxc(r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode),
- char, regexp);
- if (r == NULL)
+ rx = (REGEXP*) newSV_type(SVt_REGEXP);
+ r = (struct regexp*)SvANY(rx);
+ Newxc(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode),
+ char, regexp_internal);
+ if ( r == NULL || ri == NULL )
FAIL("Regexp out of space");
#ifdef DEBUGGING
/* avoid reading uninitialized memory in DEBUGGING code in study_chunk() */
- Zero(r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode), char);
+ Zero(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode), char);
#else
- /* bulk initialize fields with 0. */
- Zero(r, sizeof(regexp), char);
+ /* bulk initialize base fields with 0. */
+ Zero(ri, sizeof(regexp_internal), char);
#endif
/* non-zero initialization begins here */
+ RXi_SET( r, ri );
r->engine= RE_ENGINE_PTR;
- r->refcnt = 1;
- r->prelen = xend - exp;
- r->precomp = savepvn(RExC_precomp, r->prelen);
- r->extflags = pm->op_pmflags & RXf_PMf_COMPILETIME;
+ r->extflags = pm_flags;
+ {
+ bool has_p = ((r->extflags & RXf_PMf_KEEPCOPY) == RXf_PMf_KEEPCOPY);
+ bool has_minus = ((r->extflags & RXf_PMf_STD_PMMOD) != RXf_PMf_STD_PMMOD);
+ bool has_runon = ((RExC_seen & REG_SEEN_RUN_ON_COMMENT)==REG_SEEN_RUN_ON_COMMENT);
+ U16 reganch = (U16)((r->extflags & RXf_PMf_STD_PMMOD)
+ >> RXf_PMf_STD_PMMOD_SHIFT);
+ const char *fptr = STD_PAT_MODS; /*"msix"*/
+ char *p;
+ const STRLEN wraplen = plen + has_minus + has_p + has_runon
+ + (sizeof(STD_PAT_MODS) - 1)
+ + (sizeof("(?:)") - 1);
+
+ p = sv_grow((SV *)rx, wraplen + 1);
+ SvCUR_set(rx, wraplen);
+ SvPOK_on(rx);
+ SvFLAGS(rx) |= SvUTF8(pattern);
+ *p++='('; *p++='?';
+ if (has_p)
+ *p++ = KEEPCOPY_PAT_MOD; /*'p'*/
+ {
+ char *r = p + (sizeof(STD_PAT_MODS) - 1) + has_minus - 1;
+ char *colon = r + 1;
+ char ch;
+
+ while((ch = *fptr++)) {
+ if(reganch & 1)
+ *p++ = ch;
+ else
+ *r-- = ch;
+ reganch >>= 1;
+ }
+ if(has_minus) {
+ *r = '-';
+ p = colon;
+ }
+ }
+
+ *p++ = ':';
+ Copy(RExC_precomp, p, plen, char);
+ assert ((RX_WRAPPED(rx) - p) < 16);
+ r->pre_prefix = p - RX_WRAPPED(rx);
+ p += plen;
+ if (has_runon)
+ *p++ = '\n';
+ *p++ = ')';
+ *p = 0;
+ }
+
r->intflags = 0;
r->nparens = RExC_npar - 1; /* set early to validate backrefs */
}
/* Useful during FAIL. */
- Newxz(r->offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
- if (r->offsets) {
- r->offsets[0] = RExC_size;
- }
+#ifdef RE_TRACK_PATTERN_OFFSETS
+ Newxz(ri->u.offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
DEBUG_OFFSETS_r(PerlIO_printf(Perl_debug_log,
"%s %"UVuf" bytes for offset annotations.\n",
- r->offsets ? "Got" : "Couldn't get",
+ ri->u.offsets ? "Got" : "Couldn't get",
(UV)((2*RExC_size+1) * sizeof(U32))));
-
+#endif
+ SetProgLen(ri,RExC_size);
+ RExC_rx_sv = rx;
RExC_rx = r;
+ RExC_rxi = ri;
/* Second pass: emit code. */
- RExC_flags = pm->op_pmflags; /* don't let top level (?i) bleed */
+ RExC_flags = pm_flags; /* don't let top level (?i) bleed */
RExC_parse = exp;
RExC_end = xend;
RExC_naughty = 0;
RExC_npar = 1;
- RExC_cpar = 1;
- RExC_emit_start = r->program;
- RExC_emit = r->program;
-#ifdef DEBUGGING
- /* put a sentinal on the end of the program so we can check for
- overwrites */
- r->program[RExC_size].type = 255;
-#endif
+ RExC_emit_start = ri->program;
+ RExC_emit = ri->program;
+ RExC_emit_bound = ri->program + RExC_size + 1;
+
/* Store the count of eval-groups for security checks: */
- RExC_emit->next_off = (RExC_seen_evals > (I32)U16_MAX) ? U16_MAX : (U16)RExC_seen_evals;
+ RExC_rx->seen_evals = RExC_seen_evals;
REGC((U8)REG_MAGIC, (char*) RExC_emit++);
- r->data = 0;
- if (reg(pRExC_state, 0, &flags,1) == NULL)
+ if (reg(pRExC_state, 0, &flags,1) == NULL) {
+ ReREFCNT_dec(rx);
return(NULL);
-
+ }
/* XXXX To minimize changes to RE engine we always allocate
3-units-long substrs field. */
Newx(r->substrs, 1, struct reg_substr_data);
#endif
/* Dig out information for optimizations. */
- r->extflags = pm->op_pmflags & RXf_PMf_COMPILETIME; /* Again? */
- pm->op_pmflags = RExC_flags;
+ r->extflags = RExC_flags; /* was pm_op */
+ /*dmq: removed as part of de-PMOP: pm->op_pmflags = RExC_flags; */
+
if (UTF)
- r->extflags |= RXf_UTF8; /* Unicode in it? */
- r->regstclass = NULL;
+ SvUTF8_on(rx); /* Unicode in it? */
+ ri->regstclass = NULL;
if (RExC_naughty >= 10) /* Probably an expensive pattern. */
r->intflags |= PREGf_NAUGHTY;
- scan = r->program + 1; /* First BRANCH. */
+ scan = ri->program + 1; /* First BRANCH. */
/* testing for BRANCH here tells us whether there is "must appear"
data in the pattern. If there is then we can use it for optimisations */
struct regnode_charclass_class ch_class; /* pointed to by data */
int stclass_flag;
I32 last_close = 0; /* pointed to by data */
-
- first = scan;
+ regnode *first= scan;
+ regnode *first_next= regnext(first);
+
/* Skip introductions and multiplicators >= 1. */
while ((OP(first) == OPEN && (sawopen = 1)) ||
/* An OR of *one* alternative - should not happen now. */
- (OP(first) == BRANCH && OP(regnext(first)) != BRANCH) ||
+ (OP(first) == BRANCH && OP(first_next) != BRANCH) ||
/* for now we can't handle lookbehind IFMATCH*/
(OP(first) == IFMATCH && !first->flags) ||
(OP(first) == PLUS) ||
(OP(first) == MINMOD) ||
/* An {n,m} with n>0 */
- (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) )
+ (PL_regkind[OP(first)] == CURLY && ARG1(first) > 0) ||
+ (OP(first) == NOTHING && PL_regkind[OP(first_next)] != END ))
{
if (OP(first) == PLUS)
first += EXTRA_STEP_2ARGS;
} else /* XXX possible optimisation for /(?=)/ */
first = NEXTOPER(first);
+ first_next= regnext(first);
}
/* Starting-point info. */
if (OP(first) == EXACT)
NOOP; /* Empty, get anchored substr later. */
else if ((OP(first) == EXACTF || OP(first) == EXACTFL))
- r->regstclass = first;
+ ri->regstclass = first;
}
#ifdef TRIE_STCLASS
else if (PL_regkind[OP(first)] == TRIE &&
- ((reg_trie_data *)r->data->data[ ARG(first) ])->minlen>0)
+ ((reg_trie_data *)ri->data->data[ ARG(first) ])->minlen>0)
{
regnode *trie_op;
/* this can happen only on restudy */
if ( OP(first) == TRIE ) {
- struct regnode_1 *trieop;
- Newxz(trieop,1,struct regnode_1);
+ struct regnode_1 *trieop = (struct regnode_1 *)
+ PerlMemShared_calloc(1, sizeof(struct regnode_1));
StructCopy(first,trieop,struct regnode_1);
trie_op=(regnode *)trieop;
} else {
- struct regnode_charclass *trieop;
- Newxz(trieop,1,struct regnode_charclass);
+ struct regnode_charclass *trieop = (struct regnode_charclass *)
+ PerlMemShared_calloc(1, sizeof(struct regnode_charclass));
StructCopy(first,trieop,struct regnode_charclass);
trie_op=(regnode *)trieop;
}
OP(trie_op)+=2;
make_trie_failtable(pRExC_state, (regnode *)first, trie_op, 0);
- r->regstclass = trie_op;
+ ri->regstclass = trie_op;
}
#endif
else if (strchr((const char*)PL_simple,OP(first)))
- r->regstclass = first;
+ ri->regstclass = first;
else if (PL_regkind[OP(first)] == BOUND ||
PL_regkind[OP(first)] == NBOUND)
- r->regstclass = first;
+ ri->regstclass = first;
else if (PL_regkind[OP(first)] == BOL) {
r->extflags |= (OP(first) == MBOL
? RXf_ANCH_MBOL
data.last_found = newSVpvs("");
data.longest = &(data.longest_fixed);
first = scan;
- if (!r->regstclass) {
+ if (!ri->regstclass) {
cl_init(pRExC_state, &ch_class);
data.start_class = &ch_class;
stclass_flag = SCF_DO_STCLASS_AND;
if ( RExC_npar == 1 && data.longest == &(data.longest_fixed)
&& data.last_start_min == 0 && data.last_end > 0
&& !RExC_seen_zerolen
+ && !(RExC_seen & REG_SEEN_VERBARG)
&& (!(RExC_seen & REG_SEEN_GPOS) || (r->extflags & RXf_ANCH_GPOS)))
r->extflags |= RXf_CHECK_ALL;
- scan_commit(pRExC_state, &data,&minlen);
+ scan_commit(pRExC_state, &data,&minlen,0);
SvREFCNT_dec(data.last_found);
/* Note that code very similar to this but for anchored string
SvREFCNT_dec(data.longest_fixed);
longest_fixed_length = 0;
}
- if (r->regstclass
- && (OP(r->regstclass) == REG_ANY || OP(r->regstclass) == SANY))
- r->regstclass = NULL;
+ if (ri->regstclass
+ && (OP(ri->regstclass) == REG_ANY || OP(ri->regstclass) == SANY))
+ ri->regstclass = NULL;
if ((!(r->anchored_substr || r->anchored_utf8) || r->anchored_offset)
&& stclass_flag
&& !(data.start_class->flags & ANYOF_EOS)
&& !cl_is_anything(data.start_class))
{
- const I32 n = add_data(pRExC_state, 1, "f");
+ const U32 n = add_data(pRExC_state, 1, "f");
- Newx(RExC_rx->data->data[n], 1,
+ Newx(RExC_rxi->data->data[n], 1,
struct regnode_charclass_class);
StructCopy(data.start_class,
- (struct regnode_charclass_class*)RExC_rx->data->data[n],
+ (struct regnode_charclass_class*)RExC_rxi->data->data[n],
struct regnode_charclass_class);
- r->regstclass = (regnode*)RExC_rx->data->data[n];
+ ri->regstclass = (regnode*)RExC_rxi->data->data[n];
r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
DEBUG_COMPILE_r({ SV *sv = sv_newmortal();
regprop(r, sv, (regnode*)data.start_class);
DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "\nMulti Top Level\n"));
- scan = r->program + 1;
+ scan = ri->program + 1;
cl_init(pRExC_state, &ch_class);
data.start_class = &ch_class;
data.last_closep = &last_close;
if (!(data.start_class->flags & ANYOF_EOS)
&& !cl_is_anything(data.start_class))
{
- const I32 n = add_data(pRExC_state, 1, "f");
+ const U32 n = add_data(pRExC_state, 1, "f");
- Newx(RExC_rx->data->data[n], 1,
+ Newx(RExC_rxi->data->data[n], 1,
struct regnode_charclass_class);
StructCopy(data.start_class,
- (struct regnode_charclass_class*)RExC_rx->data->data[n],
+ (struct regnode_charclass_class*)RExC_rxi->data->data[n],
struct regnode_charclass_class);
- r->regstclass = (regnode*)RExC_rx->data->data[n];
+ ri->regstclass = (regnode*)RExC_rxi->data->data[n];
r->intflags &= ~PREGf_SKIP; /* Used in find_byclass(). */
DEBUG_COMPILE_r({ SV* sv = sv_newmortal();
regprop(r, sv, (regnode*)data.start_class);
the "real" pattern. */
DEBUG_OPTIMISE_r({
PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf"\n",
- minlen, r->minlen);
+ (IV)minlen, (IV)r->minlen);
});
r->minlenret = minlen;
if (r->minlen < minlen)
if (RExC_seen & REG_SEEN_CUTGROUP)
r->intflags |= PREGf_CUTGROUP_SEEN;
if (RExC_paren_names)
- r->paren_names = (HV*)SvREFCNT_inc(RExC_paren_names);
+ RXp_PAREN_NAMES(r) = (HV*)SvREFCNT_inc(RExC_paren_names);
else
- r->paren_names = NULL;
-
+ RXp_PAREN_NAMES(r) = NULL;
+
+#ifdef STUPID_PATTERN_CHECKS
+ if (RX_PRELEN(rx) == 0)
+ r->extflags |= RXf_NULL;
+ if (r->extflags & RXf_SPLIT && RX_PRELEN(rx) == 1 && RX_PRECOMP(rx)[0] == ' ')
+ /* XXX: this should happen BEFORE we compile */
+ r->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
+ else if (RX_PRELEN(rx) == 3 && memEQ("\\s+", RX_PRECOMP(rx), 3))
+ r->extflags |= RXf_WHITE;
+ else if (RX_PRELEN(rx) == 1 && RXp_PRECOMP(rx)[0] == '^')
+ r->extflags |= RXf_START_ONLY;
+#else
+ if (r->extflags & RXf_SPLIT && RX_PRELEN(rx) == 1 && RX_PRECOMP(rx)[0] == ' ')
+ /* XXX: this should happen BEFORE we compile */
+ r->extflags |= (RXf_SKIPWHITE|RXf_WHITE);
+ else {
+ regnode *first = ri->program + 1;
+ U8 fop = OP(first);
+ U8 nop = OP(NEXTOPER(first));
+
+ if (PL_regkind[fop] == NOTHING && nop == END)
+ r->extflags |= RXf_NULL;
+ else if (PL_regkind[fop] == BOL && nop == END)
+ r->extflags |= RXf_START_ONLY;
+ else if (fop == PLUS && nop ==SPACE && OP(regnext(first))==END)
+ r->extflags |= RXf_WHITE;
+ }
+#endif
+#ifdef DEBUGGING
+ if (RExC_paren_names) {
+ ri->name_list_idx = add_data( pRExC_state, 1, "p" );
+ ri->data->data[ri->name_list_idx] = (void*)SvREFCNT_inc(RExC_paren_name_list);
+ } else
+#endif
+ ri->name_list_idx = 0;
+
if (RExC_recurse_count) {
for ( ; RExC_recurse_count ; RExC_recurse_count-- ) {
const regnode *scan = RExC_recurse[RExC_recurse_count-1];
ARG2L_SET( scan, RExC_open_parens[ARG(scan)-1] - scan );
}
}
- Newxz(r->startp, RExC_npar, I32);
- Newxz(r->endp, RExC_npar, I32);
+ Newxz(r->offs, RExC_npar, regexp_paren_pair);
/* assume we don't need to swap parens around before we match */
DEBUG_DUMP_r({
PerlIO_printf(Perl_debug_log,"Final program:\n");
regdump(r);
});
- DEBUG_OFFSETS_r(if (r->offsets) {
- const U32 len = r->offsets[0];
+#ifdef RE_TRACK_PATTERN_OFFSETS
+ DEBUG_OFFSETS_r(if (ri->u.offsets) {
+ const U32 len = ri->u.offsets[0];
U32 i;
GET_RE_DEBUG_FLAGS_DECL;
- PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)r->offsets[0]);
+ PerlIO_printf(Perl_debug_log, "Offsets: [%"UVuf"]\n\t", (UV)ri->u.offsets[0]);
for (i = 1; i <= len; i++) {
- if (r->offsets[i*2-1] || r->offsets[i*2])
+ if (ri->u.offsets[i*2-1] || ri->u.offsets[i*2])
PerlIO_printf(Perl_debug_log, "%"UVuf":%"UVuf"[%"UVuf"] ",
- (UV)i, (UV)r->offsets[i*2-1], (UV)r->offsets[i*2]);
+ (UV)i, (UV)ri->u.offsets[i*2-1], (UV)ri->u.offsets[i*2]);
}
PerlIO_printf(Perl_debug_log, "\n");
});
- return(r);
- END_BLOCK
+#endif
+ return rx;
}
-#undef CORE_ONLY_BLOCK
-#undef END_BLOCK
#undef RE_ENGINE_PTR
-#ifndef PERL_IN_XSUB_RE
+
SV*
-Perl_reg_named_buff_sv(pTHX_ SV* namesv)
+Perl_reg_named_buff(pTHX_ REGEXP * const rx, SV * const key, SV * const value,
+ const U32 flags)
{
- I32 parno = 0; /* no match */
- if (PL_curpm) {
- const REGEXP * const rx = PM_GETRE(PL_curpm);
- if (rx && rx->paren_names) {
- HE *he_str = hv_fetch_ent( rx->paren_names, namesv, 0, 0 );
- if (he_str) {
- IV i;
- SV* sv_dat=HeVAL(he_str);
- I32 *nums=(I32*)SvPVX(sv_dat);
- for ( i=0; i<SvIVX(sv_dat); i++ ) {
- if ((I32)(rx->lastparen) >= nums[i] &&
- rx->endp[nums[i]] != -1)
- {
- parno = nums[i];
- break;
- }
+ PERL_UNUSED_ARG(value);
+
+ if (flags & RXapif_FETCH) {
+ return reg_named_buff_fetch(rx, key, flags);
+ } else if (flags & (RXapif_STORE | RXapif_DELETE | RXapif_CLEAR)) {
+ Perl_croak(aTHX_ PL_no_modify);
+ return NULL;
+ } else if (flags & RXapif_EXISTS) {
+ return reg_named_buff_exists(rx, key, flags)
+ ? &PL_sv_yes
+ : &PL_sv_no;
+ } else if (flags & RXapif_REGNAMES) {
+ return reg_named_buff_all(rx, flags);
+ } else if (flags & (RXapif_SCALAR | RXapif_REGNAMES_COUNT)) {
+ return reg_named_buff_scalar(rx, flags);
+ } else {
+ Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff", (int)flags);
+ return NULL;
+ }
+}
+
+SV*
+Perl_reg_named_buff_iter(pTHX_ REGEXP * const rx, const SV * const lastkey,
+ const U32 flags)
+{
+ PERL_UNUSED_ARG(lastkey);
+
+ if (flags & RXapif_FIRSTKEY)
+ return reg_named_buff_firstkey(rx, flags);
+ else if (flags & RXapif_NEXTKEY)
+ return reg_named_buff_nextkey(rx, flags);
+ else {
+ Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_iter", (int)flags);
+ return NULL;
+ }
+}
+
+SV*
+Perl_reg_named_buff_fetch(pTHX_ REGEXP * const r, SV * const namesv,
+ const U32 flags)
+{
+ AV *retarray = NULL;
+ SV *ret;
+ struct regexp *const rx = (struct regexp *)SvANY(r);
+ if (flags & RXapif_ALL)
+ retarray=newAV();
+
+ if (rx && RXp_PAREN_NAMES(rx)) {
+ HE *he_str = hv_fetch_ent( RXp_PAREN_NAMES(rx), namesv, 0, 0 );
+ if (he_str) {
+ IV i;
+ SV* sv_dat=HeVAL(he_str);
+ I32 *nums=(I32*)SvPVX(sv_dat);
+ for ( i=0; i<SvIVX(sv_dat); i++ ) {
+ if ((I32)(rx->nparens) >= nums[i]
+ && rx->offs[nums[i]].start != -1
+ && rx->offs[nums[i]].end != -1)
+ {
+ ret = newSVpvs("");
+ CALLREG_NUMBUF_FETCH(r,nums[i],ret);
+ if (!retarray)
+ return ret;
+ } else {
+ ret = newSVsv(&PL_sv_undef);
+ }
+ if (retarray) {
+ SvREFCNT_inc_simple_void(ret);
+ av_push(retarray, ret);
}
}
+ if (retarray)
+ return newRV((SV*)retarray);
}
}
- if ( !parno ) {
- return 0;
+ return NULL;
+}
+
+bool
+Perl_reg_named_buff_exists(pTHX_ REGEXP * const r, SV * const key,
+ const U32 flags)
+{
+ struct regexp *const rx = (struct regexp *)SvANY(r);
+ if (rx && RXp_PAREN_NAMES(rx)) {
+ if (flags & RXapif_ALL) {
+ return hv_exists_ent(RXp_PAREN_NAMES(rx), key, 0);
+ } else {
+ SV *sv = CALLREG_NAMED_BUFF_FETCH(r, key, flags);
+ if (sv) {
+ SvREFCNT_dec(sv);
+ return TRUE;
+ } else {
+ return FALSE;
+ }
+ }
} else {
- GV *gv_paren;
- SV *sv= sv_newmortal();
- Perl_sv_setpvf(aTHX_ sv, "%"IVdf,(IV)parno);
- gv_paren= Perl_gv_fetchsv(aTHX_ sv, GV_ADD, SVt_PVGV);
- return GvSVn(gv_paren);
+ return FALSE;
}
}
-#endif
+
+SV*
+Perl_reg_named_buff_firstkey(pTHX_ REGEXP * const r, const U32 flags)
+{
+ struct regexp *const rx = (struct regexp *)SvANY(r);
+ if ( rx && RXp_PAREN_NAMES(rx) ) {
+ (void)hv_iterinit(RXp_PAREN_NAMES(rx));
+
+ return CALLREG_NAMED_BUFF_NEXTKEY(r, NULL, flags & ~RXapif_FIRSTKEY);
+ } else {
+ return FALSE;
+ }
+}
+
+SV*
+Perl_reg_named_buff_nextkey(pTHX_ REGEXP * const r, const U32 flags)
+{
+ struct regexp *const rx = (struct regexp *)SvANY(r);
+ if (rx && RXp_PAREN_NAMES(rx)) {
+ HV *hv = RXp_PAREN_NAMES(rx);
+ HE *temphe;
+ while ( (temphe = hv_iternext_flags(hv,0)) ) {
+ IV i;
+ IV parno = 0;
+ SV* sv_dat = HeVAL(temphe);
+ I32 *nums = (I32*)SvPVX(sv_dat);
+ for ( i = 0; i < SvIVX(sv_dat); i++ ) {
+ if ((I32)(rx->lastcloseparen) >= nums[i] &&
+ rx->offs[nums[i]].start != -1 &&
+ rx->offs[nums[i]].end != -1)
+ {
+ parno = nums[i];
+ break;
+ }
+ }
+ if (parno || flags & RXapif_ALL) {
+ return newSVhek(HeKEY_hek(temphe));
+ }
+ }
+ }
+ return NULL;
+}
+
+SV*
+Perl_reg_named_buff_scalar(pTHX_ REGEXP * const r, const U32 flags)
+{
+ SV *ret;
+ AV *av;
+ I32 length;
+ struct regexp *const rx = (struct regexp *)SvANY(r);
+
+ if (rx && RXp_PAREN_NAMES(rx)) {
+ if (flags & (RXapif_ALL | RXapif_REGNAMES_COUNT)) {
+ return newSViv(HvTOTALKEYS(RXp_PAREN_NAMES(rx)));
+ } else if (flags & RXapif_ONE) {
+ ret = CALLREG_NAMED_BUFF_ALL(r, (flags | RXapif_REGNAMES));
+ av = (AV*)SvRV(ret);
+ length = av_len(av);
+ return newSViv(length + 1);
+ } else {
+ Perl_croak(aTHX_ "panic: Unknown flags %d in named_buff_scalar", (int)flags);
+ return NULL;
+ }
+ }
+ return &PL_sv_undef;
+}
+
+SV*
+Perl_reg_named_buff_all(pTHX_ REGEXP * const r, const U32 flags)
+{
+ struct regexp *const rx = (struct regexp *)SvANY(r);
+ AV *av = newAV();
+
+ if (rx && RXp_PAREN_NAMES(rx)) {
+ HV *hv= RXp_PAREN_NAMES(rx);
+ HE *temphe;
+ (void)hv_iterinit(hv);
+ while ( (temphe = hv_iternext_flags(hv,0)) ) {
+ IV i;
+ IV parno = 0;
+ SV* sv_dat = HeVAL(temphe);
+ I32 *nums = (I32*)SvPVX(sv_dat);
+ for ( i = 0; i < SvIVX(sv_dat); i++ ) {
+ if ((I32)(rx->lastcloseparen) >= nums[i] &&
+ rx->offs[nums[i]].start != -1 &&
+ rx->offs[nums[i]].end != -1)
+ {
+ parno = nums[i];
+ break;
+ }
+ }
+ if (parno || flags & RXapif_ALL) {
+ av_push(av, newSVhek(HeKEY_hek(temphe)));
+ }
+ }
+ }
+
+ return newRV((SV*)av);
+}
+
+void
+Perl_reg_numbered_buff_fetch(pTHX_ REGEXP * const r, const I32 paren,
+ SV * const sv)
+{
+ struct regexp *const rx = (struct regexp *)SvANY(r);
+ char *s = NULL;
+ I32 i = 0;
+ I32 s1, t1;
+
+ if (!rx->subbeg) {
+ sv_setsv(sv,&PL_sv_undef);
+ return;
+ }
+ else
+ if (paren == RX_BUFF_IDX_PREMATCH && rx->offs[0].start != -1) {
+ /* $` */
+ i = rx->offs[0].start;
+ s = rx->subbeg;
+ }
+ else
+ if (paren == RX_BUFF_IDX_POSTMATCH && rx->offs[0].end != -1) {
+ /* $' */
+ s = rx->subbeg + rx->offs[0].end;
+ i = rx->sublen - rx->offs[0].end;
+ }
+ else
+ if ( 0 <= paren && paren <= (I32)rx->nparens &&
+ (s1 = rx->offs[paren].start) != -1 &&
+ (t1 = rx->offs[paren].end) != -1)
+ {
+ /* $& $1 ... */
+ i = t1 - s1;
+ s = rx->subbeg + s1;
+ } else {
+ sv_setsv(sv,&PL_sv_undef);
+ return;
+ }
+ assert(rx->sublen >= (s - rx->subbeg) + i );
+ if (i >= 0) {
+ const int oldtainted = PL_tainted;
+ TAINT_NOT;
+ sv_setpvn(sv, s, i);
+ PL_tainted = oldtainted;
+ if ( (rx->extflags & RXf_CANY_SEEN)
+ ? (RXp_MATCH_UTF8(rx)
+ && (!i || is_utf8_string((U8*)s, i)))
+ : (RXp_MATCH_UTF8(rx)) )
+ {
+ SvUTF8_on(sv);
+ }
+ else
+ SvUTF8_off(sv);
+ if (PL_tainting) {
+ if (RXp_MATCH_TAINTED(rx)) {
+ if (SvTYPE(sv) >= SVt_PVMG) {
+ MAGIC* const mg = SvMAGIC(sv);
+ MAGIC* mgt;
+ PL_tainted = 1;
+ SvMAGIC_set(sv, mg->mg_moremagic);
+ SvTAINT(sv);
+ if ((mgt = SvMAGIC(sv))) {
+ mg->mg_moremagic = mgt;
+ SvMAGIC_set(sv, mg);
+ }
+ } else {
+ PL_tainted = 1;
+ SvTAINT(sv);
+ }
+ } else
+ SvTAINTED_off(sv);
+ }
+ } else {
+ sv_setsv(sv,&PL_sv_undef);
+ return;
+ }
+}
+
+void
+Perl_reg_numbered_buff_store(pTHX_ REGEXP * const rx, const I32 paren,
+ SV const * const value)
+{
+ PERL_UNUSED_ARG(rx);
+ PERL_UNUSED_ARG(paren);
+ PERL_UNUSED_ARG(value);
+
+ if (!PL_localizing)
+ Perl_croak(aTHX_ PL_no_modify);
+}
+
+I32
+Perl_reg_numbered_buff_length(pTHX_ REGEXP * const r, const SV * const sv,
+ const I32 paren)
+{
+ struct regexp *const rx = (struct regexp *)SvANY(r);
+ I32 i;
+ I32 s1, t1;
+
+ /* Some of this code was originally in C<Perl_magic_len> in F<mg.c> */
+ switch (paren) {
+ /* $` / ${^PREMATCH} */
+ case RX_BUFF_IDX_PREMATCH:
+ if (rx->offs[0].start != -1) {
+ i = rx->offs[0].start;
+ if (i > 0) {
+ s1 = 0;
+ t1 = i;
+ goto getlen;
+ }
+ }
+ return 0;
+ /* $' / ${^POSTMATCH} */
+ case RX_BUFF_IDX_POSTMATCH:
+ if (rx->offs[0].end != -1) {
+ i = rx->sublen - rx->offs[0].end;
+ if (i > 0) {
+ s1 = rx->offs[0].end;
+ t1 = rx->sublen;
+ goto getlen;
+ }
+ }
+ return 0;
+ /* $& / ${^MATCH}, $1, $2, ... */
+ default:
+ if (paren <= (I32)rx->nparens &&
+ (s1 = rx->offs[paren].start) != -1 &&
+ (t1 = rx->offs[paren].end) != -1)
+ {
+ i = t1 - s1;
+ goto getlen;
+ } else {
+ if (ckWARN(WARN_UNINITIALIZED))
+ report_uninit((SV*)sv);
+ return 0;
+ }
+ }
+ getlen:
+ if (i > 0 && RXp_MATCH_UTF8(rx)) {
+ const char * const s = rx->subbeg + s1;
+ const U8 *ep;
+ STRLEN el;
+
+ i = t1 - s1;
+ if (is_utf8_string_loclen((U8*)s, i, &ep, &el))
+ i = el;
+ }
+ return i;
+}
+
+SV*
+Perl_reg_qr_package(pTHX_ REGEXP * const rx)
+{
+ PERL_UNUSED_ARG(rx);
+ if (0)
+ return NULL;
+ else
+ return newSVpvs("Regexp");
+}
/* Scans the name of a named buffer from the pattern.
* If flags is REG_RSN_RETURN_NULL returns null.
STATIC SV*
S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags) {
char *name_start = RExC_parse;
- if ( UTF ) {
- STRLEN numlen;
- while( isIDFIRST_uni(utf8n_to_uvchr((U8*)RExC_parse,
- RExC_end - RExC_parse, &numlen, UTF8_ALLOW_DEFAULT)))
- {
- RExC_parse += numlen;
- }
- } else {
- while( isIDFIRST(*RExC_parse) )
- RExC_parse++;
+
+ if (isIDFIRST_lazy_if(RExC_parse, UTF)) {
+ /* skip IDFIRST by using do...while */
+ if (UTF)
+ do {
+ RExC_parse += UTF8SKIP(RExC_parse);
+ } while (isALNUM_utf8((U8*)RExC_parse));
+ else
+ do {
+ RExC_parse++;
+ } while (isALNUM(*RExC_parse));
}
+
if ( flags ) {
- SV* sv_name = sv_2mortal(Perl_newSVpvn(aTHX_ name_start,
- (int)(RExC_parse - name_start)));
- if (UTF)
- SvUTF8_on(sv_name);
+ SV* sv_name
+ = newSVpvn_flags(name_start, (int)(RExC_parse - name_start),
+ SVs_TEMP | (UTF ? SVf_UTF8 : 0));
if ( flags == REG_RSN_RETURN_NAME)
return sv_name;
else if (flags==REG_RSN_RETURN_DATA) {
PerlIO_printf(Perl_debug_log,"%16s",""); \
\
if (SIZE_ONLY) \
- num=RExC_size; \
+ num = RExC_size + 1; \
else \
num=REG_NODE_NUM(RExC_emit); \
if (RExC_lastnum!=num) \
#define REGTAIL_STUDY(x,y,z) regtail((x),(y),(z),depth+1)
#endif
-/* this idea is borrowed from STR_WITH_LEN in handy.h */
-#define CHECK_WORD(s,v,l) \
- (((sizeof(s)-1)==(l)) && (strnEQ(start_verb, (s ""), (sizeof(s)-1))))
-
STATIC regnode *
S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
/* paren: Parenthesized? 0=top, 1=(, inside: changed to letter. */
register regnode *ender = NULL;
register I32 parno = 0;
I32 flags;
- const I32 oregflags = RExC_flags;
+ U32 oregflags = RExC_flags;
bool have_branch = 0;
bool is_open = 0;
+ I32 freeze_paren = 0;
+ I32 after_freeze = 0;
/* for (?g), (?gc), and (?o) warnings; warning
about (?c) will warn about (?g) -- japhy */
GET_RE_DEBUG_FLAGS_DECL;
DEBUG_PARSE("reg ");
-
*flagp = 0; /* Tentatively. */
switch ( *start_verb ) {
case 'A': /* (*ACCEPT) */
- if ( CHECK_WORD("ACCEPT",start_verb,verb_len) ) {
+ if ( memEQs(start_verb,verb_len,"ACCEPT") ) {
op = ACCEPT;
internal_argval = RExC_nestroot;
}
break;
case 'C': /* (*COMMIT) */
- if ( CHECK_WORD("COMMIT",start_verb,verb_len) )
+ if ( memEQs(start_verb,verb_len,"COMMIT") )
op = COMMIT;
break;
case 'F': /* (*FAIL) */
- if ( verb_len==1 || CHECK_WORD("FAIL",start_verb,verb_len) ) {
+ if ( verb_len==1 || memEQs(start_verb,verb_len,"FAIL") ) {
op = OPFAIL;
argok = 0;
}
break;
case ':': /* (*:NAME) */
case 'M': /* (*MARK:NAME) */
- if ( verb_len==0 || CHECK_WORD("MARK",start_verb,verb_len) ) {
+ if ( verb_len==0 || memEQs(start_verb,verb_len,"MARK") ) {
op = MARKPOINT;
argok = -1;
}
break;
case 'P': /* (*PRUNE) */
- if ( CHECK_WORD("PRUNE",start_verb,verb_len) )
+ if ( memEQs(start_verb,verb_len,"PRUNE") )
op = PRUNE;
break;
case 'S': /* (*SKIP) */
- if ( CHECK_WORD("SKIP",start_verb,verb_len) )
+ if ( memEQs(start_verb,verb_len,"SKIP") )
op = SKIP;
break;
case 'T': /* (*THEN) */
/* [19:06] <TimToady> :: is then */
- if ( CHECK_WORD("THEN",start_verb,verb_len) ) {
+ if ( memEQs(start_verb,verb_len,"THEN") ) {
op = CUTGROUP;
RExC_seen |= REG_SEEN_CUTGROUP;
}
if (start_arg) {
SV *sv = newSVpvn( start_arg, RExC_parse - start_arg);
ARG(ret) = add_data( pRExC_state, 1, "S" );
- RExC_rx->data->data[ARG(ret)]=(void*)sv;
+ RExC_rxi->data->data[ARG(ret)]=(void*)sv;
ret->flags = 0;
} else {
ret->flags = 1;
return ret;
} else
if (*RExC_parse == '?') { /* (?...) */
- U32 posflags = 0, negflags = 0;
- U32 *flagsp = &posflags;
bool is_logical = 0;
const char * const seqstart = RExC_parse;
ret = NULL; /* For look-ahead/behind. */
switch (paren) {
- case '<': /* (?<...) */
+ case 'P': /* (?P...) variants for those used to PCRE/Python */
+ paren = *RExC_parse++;
+ if ( paren == '<') /* (?P<...>) named capture */
+ goto named_capture;
+ else if (paren == '>') { /* (?P>name) named recursion */
+ goto named_recursion;
+ }
+ else if (paren == '=') { /* (?P=...) named backref */
+ /* this pretty much dupes the code for \k<NAME> in regatom(), if
+ you change this make sure you change that */
+ char* name_start = RExC_parse;
+ U32 num = 0;
+ SV *sv_dat = reg_scan_name(pRExC_state,
+ SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
+ if (RExC_parse == name_start || *RExC_parse != ')')
+ vFAIL2("Sequence %.3s... not terminated",parse_start);
+
+ if (!SIZE_ONLY) {
+ num = add_data( pRExC_state, 1, "S" );
+ RExC_rxi->data->data[num]=(void*)sv_dat;
+ SvREFCNT_inc_simple_void(sv_dat);
+ }
+ RExC_sawback = 1;
+ ret = reganode(pRExC_state,
+ (U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF),
+ num);
+ *flagp |= HASWIDTH;
+
+ Set_Node_Offset(ret, parse_start+1);
+ Set_Node_Cur_Length(ret); /* MJD */
+
+ nextchar(pRExC_state);
+ return ret;
+ }
+ RExC_parse++;
+ vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
+ /*NOTREACHED*/
+ case '<': /* (?<...) */
if (*RExC_parse == '!')
paren = ',';
else if (*RExC_parse != '=')
+ named_capture:
{ /* (?<...>) */
char *name_start;
SV *svname;
SIZE_ONLY ? /* reverse test from the others */
REG_RSN_RETURN_NAME :
REG_RSN_RETURN_NULL);
- if (RExC_parse == name_start)
- goto unknown;
+ if (RExC_parse == name_start) {
+ RExC_parse++;
+ vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
+ /*NOTREACHED*/
+ }
if (*RExC_parse != paren)
vFAIL2("Sequence (?%c... not terminated",
paren=='>' ? '<' : paren);
if (!RExC_paren_names) {
RExC_paren_names= newHV();
sv_2mortal((SV*)RExC_paren_names);
+#ifdef DEBUGGING
+ RExC_paren_name_list= newAV();
+ sv_2mortal((SV*)RExC_paren_name_list);
+#endif
}
he_str = hv_fetch_ent( RExC_paren_names, svname, 1, 0 );
if ( he_str )
Perl_croak(aTHX_
"panic: paren_name hash element allocation failed");
} else if ( SvPOK(sv_dat) ) {
- IV count=SvIV(sv_dat);
- I32 *pv=(I32*)SvGROW(sv_dat,SvCUR(sv_dat)+sizeof(I32)+1);
- SvCUR_set(sv_dat,SvCUR(sv_dat)+sizeof(I32));
- pv[count]=RExC_npar;
- SvIVX(sv_dat)++;
+ /* (?|...) can mean we have dupes so scan to check
+ its already been stored. Maybe a flag indicating
+ we are inside such a construct would be useful,
+ but the arrays are likely to be quite small, so
+ for now we punt -- dmq */
+ IV count = SvIV(sv_dat);
+ I32 *pv = (I32*)SvPVX(sv_dat);
+ IV i;
+ for ( i = 0 ; i < count ; i++ ) {
+ if ( pv[i] == RExC_npar ) {
+ count = 0;
+ break;
+ }
+ }
+ if ( count ) {
+ pv = (I32*)SvGROW(sv_dat, SvCUR(sv_dat) + sizeof(I32)+1);
+ SvCUR_set(sv_dat, SvCUR(sv_dat) + sizeof(I32));
+ pv[count] = RExC_npar;
+ SvIVX(sv_dat)++;
+ }
} else {
(void)SvUPGRADE(sv_dat,SVt_PVNV);
sv_setpvn(sv_dat, (char *)&(RExC_npar), sizeof(I32));
SvIOK_on(sv_dat);
SvIVX(sv_dat)= 1;
}
+#ifdef DEBUGGING
+ if (!av_store(RExC_paren_name_list, RExC_npar, SvREFCNT_inc(svname)))
+ SvREFCNT_dec(svname);
+#endif
/*sv_dump(sv_dat);*/
}
nextchar(pRExC_state);
return ret;
}
+ break;
+ case '|': /* (?|...) */
+ /* branch reset, behave like a (?:...) except that
+ buffers in alternations share the same numbers */
+ paren = ':';
+ after_freeze = freeze_paren = RExC_npar;
+ break;
case ':': /* (?:...) */
case '>': /* (?>...) */
break;
if (*RExC_parse != ')')
FAIL("Sequence (?R) not terminated");
ret = reg_node(pRExC_state, GOSTART);
+ *flagp |= POSTPONED;
nextchar(pRExC_state);
return ret;
/*notreached*/
{ /* named and numeric backreferences */
I32 num;
- char * parse_start;
case '&': /* (?&NAME) */
parse_start = RExC_parse - 1;
+ named_recursion:
{
SV *sv_dat = reg_scan_name(pRExC_state,
SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
Set_Node_Length(ret, 1 + regarglen[OP(ret)]); /* MJD */
Set_Node_Offset(ret, parse_start); /* MJD */
+ *flagp |= POSTPONED;
nextchar(pRExC_state);
return ret;
} /* named and numeric backreferences */
/* NOT REACHED */
- case 'p': /* (?p...) */
- if (SIZE_ONLY && ckWARN2(WARN_DEPRECATED, WARN_REGEXP))
- vWARNdep(RExC_parse, "(?p{}) is deprecated - use (??{})");
- /* FALL THROUGH*/
case '?': /* (??...) */
is_logical = 1;
- if (*RExC_parse != '{')
- goto unknown;
+ if (*RExC_parse != '{') {
+ RExC_parse++;
+ vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
+ /*NOTREACHED*/
+ }
+ *flagp |= POSTPONED;
paren = *RExC_parse++;
/* FALL THROUGH */
case '{': /* (?{...}) */
{
- I32 count = 1, n = 0;
+ I32 count = 1;
+ U32 n = 0;
char c;
char *s = RExC_parse;
LEAVE;
n = add_data(pRExC_state, 3, "nop");
- RExC_rx->data->data[n] = (void*)rop;
- RExC_rx->data->data[n+1] = (void*)sop;
- RExC_rx->data->data[n+2] = (void*)pad;
+ RExC_rxi->data->data[n] = (void*)rop;
+ RExC_rxi->data->data[n+1] = (void*)sop;
+ RExC_rxi->data->data[n+2] = (void*)pad;
SvREFCNT_dec(sv);
}
else { /* First pass */
{
char ch = RExC_parse[0] == '<' ? '>' : '\'';
char *name_start= RExC_parse++;
- I32 num = 0;
+ U32 num = 0;
SV *sv_dat=reg_scan_name(pRExC_state,
SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
if (RExC_parse == name_start || *RExC_parse != ch)
RExC_parse++;
if (!SIZE_ONLY) {
num = add_data( pRExC_state, 1, "S" );
- RExC_rx->data->data[num]=(void*)sv_dat;
- SvREFCNT_inc(sv_dat);
+ RExC_rxi->data->data[num]=(void*)sv_dat;
+ SvREFCNT_inc_simple_void(sv_dat);
}
ret = reganode(pRExC_state,NGROUPP,num);
goto insert_if_check_paren;
}
else
REGTAIL(pRExC_state, ret, ender);
+ RExC_size++; /* XXX WHY do we need this?!!
+ For large programs it seems to be required
+ but I can't figure out why. -- dmq*/
return ret;
}
else {
vFAIL("Sequence (? incomplete");
break;
default:
- --RExC_parse;
- parse_flags: /* (?i) */
- while (*RExC_parse && strchr("iogcmsx", *RExC_parse)) {
+ --RExC_parse;
+ parse_flags: /* (?i) */
+ {
+ U32 posflags = 0, negflags = 0;
+ U32 *flagsp = &posflags;
+
+ while (*RExC_parse) {
+ /* && strchr("iogcmsx", *RExC_parse) */
/* (?g), (?gc) and (?o) are useless here
and must be globally applied -- japhy */
-
- if (*RExC_parse == 'o' || *RExC_parse == 'g') {
+ switch (*RExC_parse) {
+ CASE_STD_PMMOD_FLAGS_PARSE_SET(flagsp);
+ case ONCE_PAT_MOD: /* 'o' */
+ case GLOBAL_PAT_MOD: /* 'g' */
if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
const I32 wflagbit = *RExC_parse == 'o' ? WASTED_O : WASTED_G;
if (! (wastedflags & wflagbit) ) {
);
}
}
- }
- else if (*RExC_parse == 'c') {
+ break;
+
+ case CONTINUE_PAT_MOD: /* 'c' */
if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
if (! (wastedflags & WASTED_C) ) {
wastedflags |= WASTED_GC;
);
}
}
- }
- else { pmflag(flagsp, *RExC_parse); }
-
- ++RExC_parse;
- }
- if (*RExC_parse == '-') {
- flagsp = &negflags;
- wastedflags = 0; /* reset so (?g-c) warns twice */
+ break;
+ case KEEPCOPY_PAT_MOD: /* 'p' */
+ if (flagsp == &negflags) {
+ if (SIZE_ONLY && ckWARN(WARN_REGEXP))
+ vWARN(RExC_parse + 1,"Useless use of (?-p)");
+ } else {
+ *flagsp |= RXf_PMf_KEEPCOPY;
+ }
+ break;
+ case '-':
+ if (flagsp == &negflags) {
+ RExC_parse++;
+ vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
+ /*NOTREACHED*/
+ }
+ flagsp = &negflags;
+ wastedflags = 0; /* reset so (?g-c) warns twice */
+ break;
+ case ':':
+ paren = ':';
+ /*FALLTHROUGH*/
+ case ')':
+ RExC_flags |= posflags;
+ RExC_flags &= ~negflags;
+ if (paren != ':') {
+ oregflags |= posflags;
+ oregflags &= ~negflags;
+ }
+ nextchar(pRExC_state);
+ if (paren != ':') {
+ *flagp = TRYAGAIN;
+ return NULL;
+ } else {
+ ret = NULL;
+ goto parse_rest;
+ }
+ /*NOTREACHED*/
+ default:
+ RExC_parse++;
+ vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
+ /*NOTREACHED*/
+ }
++RExC_parse;
- goto parse_flags;
- }
- RExC_flags |= posflags;
- RExC_flags &= ~negflags;
- if (*RExC_parse == ':') {
- RExC_parse++;
- paren = ':';
- break;
- }
- unknown:
- if (*RExC_parse != ')') {
- RExC_parse++;
- vFAIL3("Sequence (%.*s...) not recognized", RExC_parse-seqstart, seqstart);
}
- nextchar(pRExC_state);
- *flagp = TRYAGAIN;
- return NULL;
- }
+ }} /* one for the default block, one for the switch */
}
else { /* (...) */
capturing_parens:
if (!SIZE_ONLY ){
if (!RExC_nestroot)
RExC_nestroot = parno;
- if (RExC_seen & REG_SEEN_RECURSE) {
+ if (RExC_seen & REG_SEEN_RECURSE
+ && !RExC_open_parens[parno-1])
+ {
DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
"Setting open paren #%"IVdf" to %d\n",
(IV)parno, REG_NODE_NUM(ret)));
}
else /* ! paren */
ret = NULL;
-
+
+ parse_rest:
/* Pick up the branches, linking them together. */
parse_start = RExC_parse; /* MJD */
br = regbranch(pRExC_state, &flags, 1,depth+1);
}
else if (paren != '?') /* Not Conditional */
ret = br;
- *flagp |= flags & (SPSTART | HASWIDTH);
+ *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
lastbr = br;
while (*RExC_parse == '|') {
if (!SIZE_ONLY && RExC_extralen) {
if (SIZE_ONLY)
RExC_extralen += 2; /* Account for LONGJMP. */
nextchar(pRExC_state);
+ if (freeze_paren) {
+ if (RExC_npar > after_freeze)
+ after_freeze = RExC_npar;
+ RExC_npar = freeze_paren;
+ }
br = regbranch(pRExC_state, &flags, 0, depth+1);
if (br == NULL)
return(NULL);
REGTAIL(pRExC_state, lastbr, br); /* BRANCH -> BRANCH. */
lastbr = br;
- if (flags&HASWIDTH)
- *flagp |= HASWIDTH;
- *flagp |= flags&SPSTART;
+ *flagp |= flags & (SPSTART | HASWIDTH | POSTPONED);
}
if (have_branch || paren != ':') {
ender = reg_node(pRExC_state, TAIL);
break;
case 1:
- RExC_cpar++;
ender = reganode(pRExC_state, CLOSE, parno);
if (!SIZE_ONLY && RExC_seen & REG_SEEN_RECURSE) {
DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
FAIL("Junk on end of regexp"); /* "Can't happen". */
/* NOTREACHED */
}
-
+ if (after_freeze)
+ RExC_npar = after_freeze;
return(ret);
}
I32 flags = 0, c = 0;
GET_RE_DEBUG_FLAGS_DECL;
DEBUG_PARSE("brnc");
+
if (first)
ret = NULL;
else {
}
else if (ret == NULL)
ret = latest;
- *flagp |= flags&HASWIDTH;
+ *flagp |= flags&(HASWIDTH|POSTPONED);
if (chain == NULL) /* First piece. */
*flagp |= flags&SPSTART;
else {
goto do_curly;
}
nest_check:
- if (!SIZE_ONLY && !(flags&HASWIDTH) && max > REG_INFTY/3 && ckWARN(WARN_REGEXP)) {
+ if (!SIZE_ONLY && !(flags&(HASWIDTH|POSTPONED)) && max > REG_INFTY/3 && ckWARN(WARN_REGEXP)) {
vWARN3(RExC_parse,
"%.*s matches null string many times",
(int)(RExC_parse >= origparse ? RExC_parse - origparse : 0),
/* RExC_parse points at the beginning brace,
endbrace points at the last */
if ( name[0]=='U' && name[1]=='+' ) {
- /* its a "unicode hex" notation {U+89AB} */
+ /* its a "Unicode hex" notation {U+89AB} */
I32 fl = PERL_SCAN_ALLOW_UNDERSCORES
| PERL_SCAN_DISALLOW_PREFIX
| (SIZE_ONLY ? PERL_SCAN_SILENT_ILLDIGIT : 0);
UV cp;
+ char string;
len = (STRLEN)(endbrace - name - 2);
cp = grok_hex(name + 2, &len, &fl, NULL);
if ( len != (STRLEN)(endbrace - name - 2) ) {
*valuep = cp;
return NULL;
}
- sv_str= Perl_newSVpvf_nocontext("%c",(int)cp);
+ string = (char)cp;
+ sv_str= newSVpvn(&string, 1);
} else {
/* fetch the charnames handler for this scope */
HV * const table = GvHV(PL_hintgv);
char *s;
char *p, *pend;
STRLEN charlen = 1;
+#ifdef DEBUGGING
char * parse_start = name-3; /* needed for the offsets */
+#endif
GET_RE_DEBUG_FLAGS_DECL; /* needed for the offsets */
ret = reg_node(pRExC_state,
S_reg_recode(pTHX_ const char value, SV **encp)
{
STRLEN numlen = 1;
- SV * const sv = sv_2mortal(newSVpvn(&value, numlen));
- const char * const s = encp && *encp ? sv_recode_to_utf8(sv, *encp)
- : SvPVX(sv);
+ SV * const sv = newSVpvn_flags(&value, numlen, SVs_TEMP);
+ const char * const s = *encp ? sv_recode_to_utf8(sv, *encp) : SvPVX(sv);
const STRLEN newlen = SvCUR(sv);
UV uv = UNICODE_REPLACEMENT;
if (!newlen || numlen != newlen) {
uv = UNICODE_REPLACEMENT;
- if (encp)
- *encp = NULL;
+ *encp = NULL;
}
return uv;
}
/*
- regatom - the lowest level
- *
- * Optimization: gobbles an entire sequence of ordinary characters so that
- * it can turn them into a single node, which is smaller to store and
- * faster to run. Backslashed characters are exceptions, each becoming a
- * separate node; the code is simpler that way and it's not worth fixing.
- *
- * [Yes, it is worth fixing, some scripts can run twice the speed.]
- * [It looks like its ok, as in S_study_chunk we merge adjacent EXACT nodes]
- */
+
+ Try to identify anything special at the start of the pattern. If there
+ is, then handle it as required. This may involve generating a single regop,
+ such as for an assertion; or it may involve recursing, such as to
+ handle a () structure.
+
+ If the string doesn't start with something special then we gobble up
+ as much literal text as we can.
+
+ Once we have been able to handle whatever type of thing started the
+ sequence, we return.
+
+ Note: we have to be careful with escapes, as they can be both literal
+ and special, and in the case of \10 and friends can either, depending
+ on context. Specifically there are two seperate switches for handling
+ escape sequences, with the one for handling literal escapes requiring
+ a dummy entry for all of the special escapes that are actually handled
+ by the other.
+*/
+
STATIC regnode *
S_regatom(pTHX_ RExC_state_t *pRExC_state, I32 *flagp, U32 depth)
{
DEBUG_PARSE("atom");
*flagp = WORST; /* Tentatively. */
+
tryagain:
- switch (*RExC_parse) {
+ switch ((U8)*RExC_parse) {
case '^':
RExC_seen_zerolen++;
nextchar(pRExC_state);
}
return(NULL);
}
- *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE);
+ *flagp |= flags&(HASWIDTH|SPSTART|SIMPLE|POSTPONED);
break;
case '|':
case ')':
RExC_parse++;
vFAIL("Quantifier follows nothing");
break;
+ case 0xDF:
+ case 0xC3:
+ case 0xCE:
+ do_foldchar:
+ if (!LOC && FOLD) {
+ U32 len,cp;
+ len=0; /* silence a spurious compiler warning */
+ if ((cp = what_len_TRICKYFOLD_safe(RExC_parse,RExC_end,UTF,len))) {
+ *flagp |= HASWIDTH; /* could be SIMPLE too, but needs a handler in regexec.regrepeat */
+ RExC_parse+=len-1; /* we get one from nextchar() as well. :-( */
+ ret = reganode(pRExC_state, FOLDCHAR, cp);
+ Set_Node_Length(ret, 1); /* MJD */
+ nextchar(pRExC_state); /* kill whitespace under /x */
+ return ret;
+ }
+ }
+ goto outer_default;
case '\\':
- switch (*++RExC_parse) {
+ /* Special Escapes
+
+ This switch handles escape sequences that resolve to some kind
+ of special regop and not to literal text. Escape sequnces that
+ resolve to literal text are handled below in the switch marked
+ "Literal Escapes".
+
+ Every entry in this switch *must* have a corresponding entry
+ in the literal escape switch. However, the opposite is not
+ required, as the default for this switch is to jump to the
+ literal text handling code.
+ */
+ switch ((U8)*++RExC_parse) {
+ case 0xDF:
+ case 0xC3:
+ case 0xCE:
+ goto do_foldchar;
+ /* Special Escapes */
case 'A':
RExC_seen_zerolen++;
ret = reg_node(pRExC_state, SBOL);
*flagp |= SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'G':
ret = reg_node(pRExC_state, GPOS);
RExC_seen |= REG_SEEN_GPOS;
*flagp |= SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
+ case 'K':
+ RExC_seen_zerolen++;
+ ret = reg_node(pRExC_state, KEEPS);
+ *flagp |= SIMPLE;
+ /* XXX:dmq : disabling in-place substitution seems to
+ * be necessary here to avoid cases of memory corruption, as
+ * with: C<$_="x" x 80; s/x\K/y/> -- rgs
+ */
+ RExC_seen |= REG_SEEN_LOOKBEHIND;
+ goto finish_meta_pat;
case 'Z':
ret = reg_node(pRExC_state, SEOL);
*flagp |= SIMPLE;
RExC_seen_zerolen++; /* Do not optimize RE away */
- nextchar(pRExC_state);
- break;
+ goto finish_meta_pat;
case 'z':
ret = reg_node(pRExC_state, EOS);
*flagp |= SIMPLE;
RExC_seen_zerolen++; /* Do not optimize RE away */
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'C':
ret = reg_node(pRExC_state, CANY);
RExC_seen |= REG_SEEN_CANY;
*flagp |= HASWIDTH|SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'X':
ret = reg_node(pRExC_state, CLUMP);
*flagp |= HASWIDTH;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'w':
ret = reg_node(pRExC_state, (U8)(LOC ? ALNUML : ALNUM));
*flagp |= HASWIDTH|SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'W':
ret = reg_node(pRExC_state, (U8)(LOC ? NALNUML : NALNUM));
*flagp |= HASWIDTH|SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'b':
RExC_seen_zerolen++;
RExC_seen |= REG_SEEN_LOOKBEHIND;
ret = reg_node(pRExC_state, (U8)(LOC ? BOUNDL : BOUND));
*flagp |= SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'B':
RExC_seen_zerolen++;
RExC_seen |= REG_SEEN_LOOKBEHIND;
ret = reg_node(pRExC_state, (U8)(LOC ? NBOUNDL : NBOUND));
*flagp |= SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 's':
ret = reg_node(pRExC_state, (U8)(LOC ? SPACEL : SPACE));
*flagp |= HASWIDTH|SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'S':
ret = reg_node(pRExC_state, (U8)(LOC ? NSPACEL : NSPACE));
*flagp |= HASWIDTH|SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'd':
ret = reg_node(pRExC_state, DIGIT);
*flagp |= HASWIDTH|SIMPLE;
- nextchar(pRExC_state);
- Set_Node_Length(ret, 2); /* MJD */
- break;
+ goto finish_meta_pat;
case 'D':
ret = reg_node(pRExC_state, NDIGIT);
*flagp |= HASWIDTH|SIMPLE;
+ goto finish_meta_pat;
+ case 'R':
+ ret = reg_node(pRExC_state, LNBREAK);
+ *flagp |= HASWIDTH|SIMPLE;
+ goto finish_meta_pat;
+ case 'h':
+ ret = reg_node(pRExC_state, HORIZWS);
+ *flagp |= HASWIDTH|SIMPLE;
+ goto finish_meta_pat;
+ case 'H':
+ ret = reg_node(pRExC_state, NHORIZWS);
+ *flagp |= HASWIDTH|SIMPLE;
+ goto finish_meta_pat;
+ case 'v':
+ ret = reg_node(pRExC_state, VERTWS);
+ *flagp |= HASWIDTH|SIMPLE;
+ goto finish_meta_pat;
+ case 'V':
+ ret = reg_node(pRExC_state, NVERTWS);
+ *flagp |= HASWIDTH|SIMPLE;
+ finish_meta_pat:
nextchar(pRExC_state);
Set_Node_Length(ret, 2); /* MJD */
- break;
+ break;
case 'p':
case 'P':
{
char* const oldregxend = RExC_end;
+#ifdef DEBUGGING
char* parse_start = RExC_parse - 2;
+#endif
if (RExC_parse[1] == '{') {
/* a lovely hack--pretend we saw [\pX] instead */
ret= reg_namedseq(pRExC_state, NULL);
break;
case 'k': /* Handle \k<NAME> and \k'NAME' */
+ parse_named_seq:
{
char ch= RExC_parse[1];
- if (ch != '<' && ch != '\'') {
- if (SIZE_ONLY)
- vWARN( RExC_parse + 1,
- "Possible broken named back reference treated as literal k");
- parse_start--;
- goto defchar;
+ if (ch != '<' && ch != '\'' && ch != '{') {
+ RExC_parse++;
+ vFAIL2("Sequence %.2s... not terminated",parse_start);
} else {
+ /* this pretty much dupes the code for (?P=...) in reg(), if
+ you change this make sure you change that */
char* name_start = (RExC_parse += 2);
- I32 num = 0;
+ U32 num = 0;
SV *sv_dat = reg_scan_name(pRExC_state,
SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
- ch= (ch == '<') ? '>' : '\'';
-
+ ch= (ch == '<') ? '>' : (ch == '{') ? '}' : '\'';
if (RExC_parse == name_start || *RExC_parse != ch)
- vFAIL2("Sequence \\k%c... not terminated",
- (ch == '>' ? '<' : ch));
-
+ vFAIL2("Sequence %.3s... not terminated",parse_start);
+
+ if (!SIZE_ONLY) {
+ num = add_data( pRExC_state, 1, "S" );
+ RExC_rxi->data->data[num]=(void*)sv_dat;
+ SvREFCNT_inc_simple_void(sv_dat);
+ }
+
RExC_sawback = 1;
ret = reganode(pRExC_state,
(U8)(FOLD ? (LOC ? NREFFL : NREFF) : NREF),
num);
*flagp |= HASWIDTH;
-
-
- if (!SIZE_ONLY) {
- num = add_data( pRExC_state, 1, "S" );
- ARG_SET(ret,num);
- RExC_rx->data->data[num]=(void*)sv_dat;
- SvREFCNT_inc(sv_dat);
- }
+
/* override incorrect value set in reganode MJD */
Set_Node_Offset(ret, parse_start+1);
Set_Node_Cur_Length(ret); /* MJD */
nextchar(pRExC_state);
-
+
}
break;
- }
- case 'n':
- case 'r':
- case 't':
- case 'f':
- case 'e':
- case 'a':
- case 'x':
- case 'c':
- case '0':
- goto defchar;
- case 'R':
+ }
+ case 'g':
case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9':
{
I32 num;
- bool isrel=(*RExC_parse=='R');
- if (isrel)
+ bool isg = *RExC_parse == 'g';
+ bool isrel = 0;
+ bool hasbrace = 0;
+ if (isg) {
RExC_parse++;
+ if (*RExC_parse == '{') {
+ RExC_parse++;
+ hasbrace = 1;
+ }
+ if (*RExC_parse == '-') {
+ RExC_parse++;
+ isrel = 1;
+ }
+ if (hasbrace && !isDIGIT(*RExC_parse)) {
+ if (isrel) RExC_parse--;
+ RExC_parse -= 2;
+ goto parse_named_seq;
+ } }
num = atoi(RExC_parse);
+ if (isg && num == 0)
+ vFAIL("Reference to invalid group 0");
if (isrel) {
- num = RExC_cpar - num;
+ num = RExC_npar - num;
if (num < 1)
vFAIL("Reference to nonexistent or unclosed group");
}
- if (num > 9 && num >= RExC_npar)
+ if (!isg && num > 9 && num >= RExC_npar)
goto defchar;
else {
char * const parse_start = RExC_parse - 1; /* MJD */
while (isDIGIT(*RExC_parse))
RExC_parse++;
-
+ if (parse_start == RExC_parse - 1)
+ vFAIL("Unterminated \\g... pattern");
+ if (hasbrace) {
+ if (*RExC_parse != '}')
+ vFAIL("Unterminated \\g{...} pattern");
+ RExC_parse++;
+ }
if (!SIZE_ONLY) {
if (num > (I32)RExC_rx->nparens)
vFAIL("Reference to nonexistent group");
- /* People make this error all the time apparently.
- So we cant fail on it, even though we should
-
- else if (num >= RExC_cpar)
- vFAIL("Reference to unclosed group will always match");
- */
}
RExC_sawback = 1;
ret = reganode(pRExC_state,
case '#':
if (RExC_flags & RXf_PMf_EXTENDED) {
- while (RExC_parse < RExC_end && *RExC_parse != '\n')
- RExC_parse++;
- if (RExC_parse < RExC_end)
+ if ( reg_skipcomment( pRExC_state ) )
goto tryagain;
}
/* FALL THROUGH */
- default: {
+ default:
+ outer_default:{
register STRLEN len;
register UV ender;
register char *p;
char * const oldp = p;
if (RExC_flags & RXf_PMf_EXTENDED)
- p = regwhite(p, RExC_end);
- switch (*p) {
+ p = regwhite( pRExC_state, p );
+ switch ((U8)*p) {
+ case 0xDF:
+ case 0xC3:
+ case 0xCE:
+ if (LOC || !FOLD || !is_TRICKYFOLD_safe(p,RExC_end,UTF))
+ goto normal_default;
case '^':
case '$':
case '.':
case '|':
goto loopdone;
case '\\':
- switch (*++p) {
- case 'A':
- case 'C':
- case 'X':
- case 'G':
- case 'Z':
- case 'z':
- case 'w':
- case 'W':
- case 'b':
- case 'B':
- case 's':
- case 'S':
- case 'd':
- case 'D':
- case 'p':
- case 'P':
- case 'N':
- case 'R':
+ /* Literal Escapes Switch
+
+ This switch is meant to handle escape sequences that
+ resolve to a literal character.
+
+ Every escape sequence that represents something
+ else, like an assertion or a char class, is handled
+ in the switch marked 'Special Escapes' above in this
+ routine, but also has an entry here as anything that
+ isn't explicitly mentioned here will be treated as
+ an unescaped equivalent literal.
+ */
+
+ switch ((U8)*++p) {
+ /* These are all the special escapes. */
+ case 0xDF:
+ case 0xC3:
+ case 0xCE:
+ if (LOC || !FOLD || !is_TRICKYFOLD_safe(p,RExC_end,UTF))
+ goto normal_default;
+ case 'A': /* Start assertion */
+ case 'b': case 'B': /* Word-boundary assertion*/
+ case 'C': /* Single char !DANGEROUS! */
+ case 'd': case 'D': /* digit class */
+ case 'g': case 'G': /* generic-backref, pos assertion */
+ case 'h': case 'H': /* HORIZWS */
+ case 'k': case 'K': /* named backref, keep marker */
+ case 'N': /* named char sequence */
+ case 'p': case 'P': /* Unicode property */
+ case 'R': /* LNBREAK */
+ case 's': case 'S': /* space class */
+ case 'v': case 'V': /* VERTWS */
+ case 'w': case 'W': /* word class */
+ case 'X': /* eXtended Unicode "combining character sequence" */
+ case 'z': case 'Z': /* End of line/string assertion */
--p;
goto loopdone;
+
+ /* Anything after here is an escape that resolves to a
+ literal. (Except digits, which may or may not)
+ */
case 'n':
ender = '\n';
p++;
ender = *p++;
break;
}
- if (RExC_flags & RXf_PMf_EXTENDED)
- p = regwhite(p, RExC_end);
+ if ( RExC_flags & RXf_PMf_EXTENDED)
+ p = regwhite( pRExC_state, p );
if (UTF && FOLD) {
/* Prime the casefolded buffer. */
ender = toFOLD_uni(ender, tmpbuf, &foldlen);
}
- if (ISMULT2(p)) { /* Back off on ?+*. */
+ if (p < RExC_end && ISMULT2(p)) { /* Back off on ?+*. */
if (len)
p = oldp;
else if (UTF) {
}
STATIC char *
-S_regwhite(char *p, const char *e)
+S_regwhite( RExC_state_t *pRExC_state, char *p )
{
+ const char *e = RExC_end;
while (p < e) {
if (isSPACE(*p))
++p;
else if (*p == '#') {
+ bool ended = 0;
do {
- p++;
- } while (p < e && *p != '\n');
+ if (*p++ == '\n') {
+ ended = 1;
+ break;
+ }
+ } while (p < e);
+ if (!ended)
+ RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
}
else
break;
}
+#define _C_C_T_(NAME,TEST,WORD) \
+ANYOF_##NAME: \
+ if (LOC) \
+ ANYOF_CLASS_SET(ret, ANYOF_##NAME); \
+ else { \
+ for (value = 0; value < 256; value++) \
+ if (TEST) \
+ ANYOF_BITMAP_SET(ret, value); \
+ } \
+ yesno = '+'; \
+ what = WORD; \
+ break; \
+case ANYOF_N##NAME: \
+ if (LOC) \
+ ANYOF_CLASS_SET(ret, ANYOF_N##NAME); \
+ else { \
+ for (value = 0; value < 256; value++) \
+ if (!TEST) \
+ ANYOF_BITMAP_SET(ret, value); \
+ } \
+ yesno = '!'; \
+ what = WORD; \
+ break
+
+#define _C_C_T_NOLOC_(NAME,TEST,WORD) \
+ANYOF_##NAME: \
+ for (value = 0; value < 256; value++) \
+ if (TEST) \
+ ANYOF_BITMAP_SET(ret, value); \
+ yesno = '+'; \
+ what = WORD; \
+ break; \
+case ANYOF_N##NAME: \
+ for (value = 0; value < 256; value++) \
+ if (!TEST) \
+ ANYOF_BITMAP_SET(ret, value); \
+ yesno = '!'; \
+ what = WORD; \
+ break
+
/*
parse a class specification and produce either an ANYOF node that
- matches the pattern. If the pattern matches a single char only and
- that char is < 256 then we produce an EXACT node instead.
+ matches the pattern or if the pattern matches a single char only and
+ that char is < 256 and we are case insensitive then we produce an
+ EXACT node instead.
*/
+
STATIC regnode *
S_regclass(pTHX_ RExC_state_t *pRExC_state, U32 depth)
{
dVAR;
- register UV value = 0;
register UV nextvalue;
register IV prevvalue = OOB_UNICODE;
register IV range = 0;
+ UV value = 0; /* XXX:dmq: needs to be referenceable (unfortunately) */
register regnode *ret;
STRLEN numlen;
IV namedclass;
case 'S': namedclass = ANYOF_NSPACE; break;
case 'd': namedclass = ANYOF_DIGIT; break;
case 'D': namedclass = ANYOF_NDIGIT; break;
+ case 'v': namedclass = ANYOF_VERTWS; break;
+ case 'V': namedclass = ANYOF_NVERTWS; break;
+ case 'h': namedclass = ANYOF_HORIZWS; break;
+ case 'H': namedclass = ANYOF_NHORIZWS; break;
case 'N': /* Handle \N{NAME} in class */
{
/* We only pay attention to the first char of
range = 0; /* this was not a true range */
}
+
+
if (!SIZE_ONLY) {
const char *what = NULL;
char yesno = 0;
* A similar issue a little earlier when switching on value.
* --jhi */
switch ((I32)namedclass) {
- case ANYOF_ALNUM:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_ALNUM);
- else {
- for (value = 0; value < 256; value++)
- if (isALNUM(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Word";
- break;
- case ANYOF_NALNUM:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NALNUM);
- else {
- for (value = 0; value < 256; value++)
- if (!isALNUM(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Word";
- break;
- case ANYOF_ALNUMC:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_ALNUMC);
- else {
- for (value = 0; value < 256; value++)
- if (isALNUMC(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Alnum";
- break;
- case ANYOF_NALNUMC:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NALNUMC);
- else {
- for (value = 0; value < 256; value++)
- if (!isALNUMC(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Alnum";
- break;
- case ANYOF_ALPHA:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_ALPHA);
- else {
- for (value = 0; value < 256; value++)
- if (isALPHA(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Alpha";
- break;
- case ANYOF_NALPHA:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NALPHA);
- else {
- for (value = 0; value < 256; value++)
- if (!isALPHA(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Alpha";
- break;
+ case _C_C_T_(ALNUM, isALNUM(value), "Word");
+ case _C_C_T_(ALNUMC, isALNUMC(value), "Alnum");
+ case _C_C_T_(ALPHA, isALPHA(value), "Alpha");
+ case _C_C_T_(BLANK, isBLANK(value), "Blank");
+ case _C_C_T_(CNTRL, isCNTRL(value), "Cntrl");
+ case _C_C_T_(GRAPH, isGRAPH(value), "Graph");
+ case _C_C_T_(LOWER, isLOWER(value), "Lower");
+ case _C_C_T_(PRINT, isPRINT(value), "Print");
+ case _C_C_T_(PSXSPC, isPSXSPC(value), "Space");
+ case _C_C_T_(PUNCT, isPUNCT(value), "Punct");
+ case _C_C_T_(SPACE, isSPACE(value), "SpacePerl");
+ case _C_C_T_(UPPER, isUPPER(value), "Upper");
+ case _C_C_T_(XDIGIT, isXDIGIT(value), "XDigit");
+ case _C_C_T_NOLOC_(VERTWS, is_VERTWS_latin1(&value), "VertSpace");
+ case _C_C_T_NOLOC_(HORIZWS, is_HORIZWS_latin1(&value), "HorizSpace");
case ANYOF_ASCII:
if (LOC)
ANYOF_CLASS_SET(ret, ANYOF_ASCII);
}
yesno = '!';
what = "ASCII";
- break;
- case ANYOF_BLANK:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_BLANK);
- else {
- for (value = 0; value < 256; value++)
- if (isBLANK(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Blank";
- break;
- case ANYOF_NBLANK:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NBLANK);
- else {
- for (value = 0; value < 256; value++)
- if (!isBLANK(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Blank";
- break;
- case ANYOF_CNTRL:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_CNTRL);
- else {
- for (value = 0; value < 256; value++)
- if (isCNTRL(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Cntrl";
- break;
- case ANYOF_NCNTRL:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NCNTRL);
- else {
- for (value = 0; value < 256; value++)
- if (!isCNTRL(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Cntrl";
- break;
+ break;
case ANYOF_DIGIT:
if (LOC)
ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
}
yesno = '!';
what = "Digit";
+ break;
+ case ANYOF_MAX:
+ /* this is to handle \p and \P */
break;
- case ANYOF_GRAPH:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_GRAPH);
- else {
- for (value = 0; value < 256; value++)
- if (isGRAPH(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Graph";
- break;
- case ANYOF_NGRAPH:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NGRAPH);
- else {
- for (value = 0; value < 256; value++)
- if (!isGRAPH(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Graph";
- break;
- case ANYOF_LOWER:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_LOWER);
- else {
- for (value = 0; value < 256; value++)
- if (isLOWER(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Lower";
- break;
- case ANYOF_NLOWER:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NLOWER);
- else {
- for (value = 0; value < 256; value++)
- if (!isLOWER(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Lower";
- break;
- case ANYOF_PRINT:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_PRINT);
- else {
- for (value = 0; value < 256; value++)
- if (isPRINT(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Print";
- break;
- case ANYOF_NPRINT:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NPRINT);
- else {
- for (value = 0; value < 256; value++)
- if (!isPRINT(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Print";
- break;
- case ANYOF_PSXSPC:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_PSXSPC);
- else {
- for (value = 0; value < 256; value++)
- if (isPSXSPC(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Space";
- break;
- case ANYOF_NPSXSPC:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NPSXSPC);
- else {
- for (value = 0; value < 256; value++)
- if (!isPSXSPC(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Space";
- break;
- case ANYOF_PUNCT:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_PUNCT);
- else {
- for (value = 0; value < 256; value++)
- if (isPUNCT(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Punct";
- break;
- case ANYOF_NPUNCT:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NPUNCT);
- else {
- for (value = 0; value < 256; value++)
- if (!isPUNCT(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Punct";
- break;
- case ANYOF_SPACE:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_SPACE);
- else {
- for (value = 0; value < 256; value++)
- if (isSPACE(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "SpacePerl";
- break;
- case ANYOF_NSPACE:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NSPACE);
- else {
- for (value = 0; value < 256; value++)
- if (!isSPACE(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "SpacePerl";
- break;
- case ANYOF_UPPER:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_UPPER);
- else {
- for (value = 0; value < 256; value++)
- if (isUPPER(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Upper";
- break;
- case ANYOF_NUPPER:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NUPPER);
- else {
- for (value = 0; value < 256; value++)
- if (!isUPPER(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Upper";
- break;
- case ANYOF_XDIGIT:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_XDIGIT);
- else {
- for (value = 0; value < 256; value++)
- if (isXDIGIT(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "XDigit";
- break;
- case ANYOF_NXDIGIT:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NXDIGIT);
- else {
- for (value = 0; value < 256; value++)
- if (!isXDIGIT(value))
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "XDigit";
- break;
- case ANYOF_MAX:
- /* this is to handle \p and \P */
- break;
- default:
- vFAIL("Invalid [::] class");
+ default:
+ vFAIL("Invalid [::] class");
break;
}
if (what) {
{
if (isLOWER(prevvalue)) {
for (i = prevvalue; i <= ceilvalue; i++)
- if (isLOWER(i))
+ if (isLOWER(i) && !ANYOF_BITMAP_TEST(ret,i)) {
+ stored++;
ANYOF_BITMAP_SET(ret, i);
+ }
} else {
for (i = prevvalue; i <= ceilvalue; i++)
- if (isUPPER(i))
+ if (isUPPER(i) && !ANYOF_BITMAP_TEST(ret,i)) {
+ stored++;
ANYOF_BITMAP_SET(ret, i);
+ }
}
}
else
if (!unicode_alternate)
unicode_alternate = newAV();
- sv = newSVpvn((char*)foldbuf, foldlen);
- SvUTF8_on(sv);
+ sv = newSVpvn_utf8((char*)foldbuf, foldlen,
+ TRUE);
av_push(unicode_alternate, sv);
}
}
return ret;
/****** !SIZE_ONLY AFTER HERE *********/
- if( stored == 1 && value < 256
+ if( stored == 1 && (value < 128 || (value < 256 && !UTF))
&& !( ANYOF_FLAGS(ret) & ( ANYOF_FLAGS_ALL ^ ANYOF_FOLD ) )
) {
/* optimize single char class to an EXACT node
av_store(av, 2, (SV*)unicode_alternate);
rv = newRV_noinc((SV*)av);
n = add_data(pRExC_state, 1, "s");
- RExC_rx->data->data[n] = (void*)rv;
+ RExC_rxi->data->data[n] = (void*)rv;
ARG_SET(ret, n);
}
return ret;
}
+#undef _C_C_T_
+
+
+/* reg_skipcomment()
+
+ Absorbs an /x style # comments from the input stream.
+ Returns true if there is more text remaining in the stream.
+ Will set the REG_SEEN_RUN_ON_COMMENT flag if the comment
+ terminates the pattern without including a newline.
+
+ Note its the callers responsibility to ensure that we are
+ actually in /x mode
+
+*/
+
+STATIC bool
+S_reg_skipcomment(pTHX_ RExC_state_t *pRExC_state)
+{
+ bool ended = 0;
+ while (RExC_parse < RExC_end)
+ if (*RExC_parse++ == '\n') {
+ ended = 1;
+ break;
+ }
+ if (!ended) {
+ /* we ran off the end of the pattern without ending
+ the comment, so we have to add an \n when wrapping */
+ RExC_seen |= REG_SEEN_RUN_ON_COMMENT;
+ return 0;
+ } else
+ return 1;
+}
+
+/* nextchar()
+
+ Advance that parse position, and optionall absorbs
+ "whitespace" from the inputstream.
+
+ Without /x "whitespace" means (?#...) style comments only,
+ with /x this means (?#...) and # comments and whitespace proper.
+
+ Returns the RExC_parse point from BEFORE the scan occurs.
+
+ This is the /x friendly way of saying RExC_parse++.
+*/
STATIC char*
S_nextchar(pTHX_ RExC_state_t *pRExC_state)
continue;
}
else if (*RExC_parse == '#') {
- while (RExC_parse < RExC_end)
- if (*RExC_parse++ == '\n') break;
- continue;
+ if ( reg_skipcomment( pRExC_state ) )
+ continue;
}
}
return retval;
RExC_size += 1;
return(ret);
}
-#ifdef DEBUGGING
- if (OP(RExC_emit) == 255)
- Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %s: %d ",
- reg_name[op], OP(RExC_emit));
-#endif
+ if (RExC_emit >= RExC_emit_bound)
+ Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d", op);
+
NODE_ALIGN_FILL(ret);
ptr = ret;
FILL_ADVANCE_NODE(ptr, op);
+#ifdef RE_TRACK_PATTERN_OFFSETS
if (RExC_offsets) { /* MJD */
MJD_OFFSET_DEBUG(("%s:%d: (op %s) %s %"UVuf" (len %"UVuf") (max %"UVuf").\n",
"reg_node", __LINE__,
- reg_name[op],
+ PL_reg_name[op],
(UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0]
? "Overwriting end of array!\n" : "OK",
(UV)(RExC_emit - RExC_emit_start),
(UV)RExC_offsets[0]));
Set_Node_Offset(RExC_emit, RExC_parse + (op == END));
}
-
+#endif
RExC_emit = ptr;
return(ret);
}
*/
return(ret);
}
-#ifdef DEBUGGING
- if (OP(RExC_emit) == 255)
- Perl_croak(aTHX_ "panic: reganode overwriting end of allocated program space");
-#endif
+ if (RExC_emit >= RExC_emit_bound)
+ Perl_croak(aTHX_ "panic: reg_node overrun trying to emit %d", op);
+
NODE_ALIGN_FILL(ret);
ptr = ret;
FILL_ADVANCE_NODE_ARG(ptr, op, arg);
+#ifdef RE_TRACK_PATTERN_OFFSETS
if (RExC_offsets) { /* MJD */
MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
"reganode",
__LINE__,
- reg_name[op],
+ PL_reg_name[op],
(UV)(RExC_emit - RExC_emit_start) > RExC_offsets[0] ?
"Overwriting end of array!\n" : "OK",
(UV)(RExC_emit - RExC_emit_start),
(UV)RExC_offsets[0]));
Set_Cur_Node_Offset;
}
-
+#endif
RExC_emit = ptr;
return(ret);
}
const int offset = regarglen[(U8)op];
const int size = NODE_STEP_REGNODE + offset;
GET_RE_DEBUG_FLAGS_DECL;
+ PERL_UNUSED_ARG(depth);
/* (PL_regkind[(U8)op] == CURLY ? EXTRA_STEP_2ARGS : 0); */
- DEBUG_PARSE_FMT("inst"," - %s",reg_name[op]);
+ DEBUG_PARSE_FMT("inst"," - %s",PL_reg_name[op]);
if (SIZE_ONLY) {
RExC_size += size;
return;
dst = RExC_emit;
if (RExC_open_parens) {
int paren;
- DEBUG_PARSE_FMT("inst"," - %"IVdf, (IV)RExC_npar);
+ /*DEBUG_PARSE_FMT("inst"," - %"IVdf, (IV)RExC_npar);*/
for ( paren=0 ; paren < RExC_npar ; paren++ ) {
if ( RExC_open_parens[paren] >= opnd ) {
- DEBUG_PARSE_FMT("open"," - %d",size);
+ /*DEBUG_PARSE_FMT("open"," - %d",size);*/
RExC_open_parens[paren] += size;
} else {
- DEBUG_PARSE_FMT("open"," - %s","ok");
+ /*DEBUG_PARSE_FMT("open"," - %s","ok");*/
}
if ( RExC_close_parens[paren] >= opnd ) {
- DEBUG_PARSE_FMT("close"," - %d",size);
+ /*DEBUG_PARSE_FMT("close"," - %d",size);*/
RExC_close_parens[paren] += size;
} else {
- DEBUG_PARSE_FMT("close"," - %s","ok");
+ /*DEBUG_PARSE_FMT("close"," - %s","ok");*/
}
}
}
while (src > opnd) {
StructCopy(--src, --dst, regnode);
+#ifdef RE_TRACK_PATTERN_OFFSETS
if (RExC_offsets) { /* MJD 20010112 */
MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s copy %"UVuf" -> %"UVuf" (max %"UVuf").\n",
"reg_insert",
__LINE__,
- reg_name[op],
+ PL_reg_name[op],
(UV)(dst - RExC_emit_start) > RExC_offsets[0]
? "Overwriting end of array!\n" : "OK",
(UV)(src - RExC_emit_start),
Set_Node_Offset_To_R(dst-RExC_emit_start, Node_Offset(src));
Set_Node_Length_To_R(dst-RExC_emit_start, Node_Length(src));
}
+#endif
}
place = opnd; /* Op node, where operand used to be. */
+#ifdef RE_TRACK_PATTERN_OFFSETS
if (RExC_offsets) { /* MJD */
MJD_OFFSET_DEBUG(("%s(%d): (op %s) %s %"UVuf" <- %"UVuf" (max %"UVuf").\n",
"reginsert",
__LINE__,
- reg_name[op],
+ PL_reg_name[op],
(UV)(place - RExC_emit_start) > RExC_offsets[0]
? "Overwriting end of array!\n" : "OK",
(UV)(place - RExC_emit_start),
Set_Node_Offset(place, RExC_parse);
Set_Node_Length(place, 1);
}
+#endif
src = NEXTOPER(place);
FILL_ADVANCE_NODE(place, op);
Zero(src, offset, regnode);
PerlIO_printf(Perl_debug_log, "~ %s (%d) %s %s\n",
SvPV_nolen_const(mysv), REG_NODE_NUM(scan),
(temp == NULL ? "->" : ""),
- (temp == NULL ? reg_name[OP(val)] : "")
+ (temp == NULL ? PL_reg_name[OP(val)] : "")
);
});
if (temp == NULL)
PerlIO_printf(Perl_debug_log, "~ %s (%d) -> %s\n",
SvPV_nolen_const(mysv),
REG_NODE_NUM(scan),
- reg_name[exact]);
+ PL_reg_name[exact]);
});
if (temp == NULL)
break;
SV * const mysv_val=sv_newmortal();
DEBUG_PARSE_MSG("");
regprop(RExC_rx, mysv_val, val);
- PerlIO_printf(Perl_debug_log, "~ attach to %s (%d) offset to %d\n",
- SvPV_nolen_const(mysv_val),
- REG_NODE_NUM(val),
- val - scan
+ PerlIO_printf(Perl_debug_log, "~ attach to %s (%"IVdf") offset to %"IVdf"\n",
+ SvPV_nolen_const(mysv_val),
+ (IV)REG_NODE_NUM(val),
+ (IV)(val - scan)
);
});
if (reg_off_by_arg[OP(scan)]) {
/*
- regdump - dump a regexp onto Perl_debug_log in vaguely comprehensible form
*/
+#ifdef DEBUGGING
+void
+S_regdump_extflags(pTHX_ const char *lead, const U32 flags) {
+ int bit;
+ int set=0;
+ for (bit=0; bit<32; bit++) {
+ if (flags & (1<<bit)) {
+ if (!set++ && lead)
+ PerlIO_printf(Perl_debug_log, "%s",lead);
+ PerlIO_printf(Perl_debug_log, "%s ",PL_reg_extflags_name[bit]);
+ }
+ }
+ if (lead) {
+ if (set)
+ PerlIO_printf(Perl_debug_log, "\n");
+ else
+ PerlIO_printf(Perl_debug_log, "%s[none-set]\n",lead);
+ }
+}
+#endif
+
void
Perl_regdump(pTHX_ const regexp *r)
{
dVAR;
SV * const sv = sv_newmortal();
SV *dsv= sv_newmortal();
+ RXi_GET_DECL(r,ri);
+ GET_RE_DEBUG_FLAGS_DECL;
- (void)dumpuntil(r, r->program, r->program + 1, NULL, NULL, sv, 0, 0);
+ (void)dumpuntil(r, ri->program, ri->program + 1, NULL, NULL, sv, 0, 0);
/* Header fields of interest. */
if (r->anchored_substr) {
if (r->check_substr || r->check_utf8)
PerlIO_printf(Perl_debug_log, ") ");
- if (r->regstclass) {
- regprop(r, sv, r->regstclass);
+ if (ri->regstclass) {
+ regprop(r, sv, ri->regstclass);
PerlIO_printf(Perl_debug_log, "stclass %s ", SvPVX_const(sv));
}
if (r->extflags & RXf_ANCH) {
PerlIO_putc(Perl_debug_log, ' ');
}
if (r->extflags & RXf_GPOS_SEEN)
- PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", r->gofs);
+ PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", (UV)r->gofs);
if (r->intflags & PREGf_SKIP)
PerlIO_printf(Perl_debug_log, "plus ");
if (r->intflags & PREGf_IMPLICIT)
PerlIO_printf(Perl_debug_log, "implicit ");
- PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
+ PerlIO_printf(Perl_debug_log, "minlen %"IVdf" ", (IV)r->minlen);
if (r->extflags & RXf_EVAL_SEEN)
PerlIO_printf(Perl_debug_log, "with eval ");
PerlIO_printf(Perl_debug_log, "\n");
+ DEBUG_FLAGS_r(regdump_extflags("r->extflags: ",r->extflags));
#else
PERL_UNUSED_CONTEXT;
PERL_UNUSED_ARG(r);
#ifdef DEBUGGING
dVAR;
register int k;
+ RXi_GET_DECL(prog,progi);
GET_RE_DEBUG_FLAGS_DECL;
+
sv_setpvn(sv, "", 0);
/* It would be nice to FAIL() here, but this may be called from
regexec.c, and it would be hard to supply pRExC_state. */
Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d", (int)OP(o), (int)REGNODE_MAX);
- sv_catpv(sv, reg_name[OP(o)]); /* Take off const! */
+ sv_catpv(sv, PL_reg_name[OP(o)]); /* Take off const! */
k = PL_regkind[OP(o)];
if (k == EXACT) {
- SV * const dsv = sv_2mortal(newSVpvs(""));
+ sv_catpvs(sv, " ");
/* Using is_utf8_string() (via PERL_PV_UNI_DETECT)
* is a crude hack but it may be the best for now since
* we have no flag "this EXACTish node was UTF-8"
* --jhi */
- const char * const s =
- pv_pretty(dsv, STRING(o), STR_LEN(o), 60,
- PL_colors[0], PL_colors[1],
- PERL_PV_ESCAPE_UNI_DETECT |
- PERL_PV_PRETTY_ELIPSES |
- PERL_PV_PRETTY_LTGT
- );
- Perl_sv_catpvf(aTHX_ sv, " %s", s );
+ pv_pretty(sv, STRING(o), STR_LEN(o), 60, PL_colors[0], PL_colors[1],
+ PERL_PV_ESCAPE_UNI_DETECT |
+ PERL_PV_PRETTY_ELLIPSES |
+ PERL_PV_PRETTY_LTGT |
+ PERL_PV_PRETTY_NOCLEAR
+ );
} else if (k == TRIE) {
/* print the details of the trie in dumpuntil instead, as
- * prog->data isn't available here */
+ * progi->data isn't available here */
const char op = OP(o);
- const I32 n = ARG(o);
+ const U32 n = ARG(o);
const reg_ac_data * const ac = IS_TRIE_AC(op) ?
- (reg_ac_data *)prog->data->data[n] :
+ (reg_ac_data *)progi->data->data[n] :
NULL;
- const reg_trie_data * const trie = !IS_TRIE_AC(op) ?
- (reg_trie_data*)prog->data->data[n] :
- ac->trie;
+ const reg_trie_data * const trie
+ = (reg_trie_data*)progi->data->data[!IS_TRIE_AC(op) ? n : ac->trie];
- Perl_sv_catpvf(aTHX_ sv, "-%s",reg_name[o->flags]);
+ Perl_sv_catpvf(aTHX_ sv, "-%s",PL_reg_name[o->flags]);
DEBUG_TRIE_COMPILE_r(
Perl_sv_catpvf(aTHX_ sv,
"<S:%"UVuf"/%"IVdf" W:%"UVuf" L:%"UVuf"/%"UVuf" C:%"UVuf"/%"UVuf">",
int i;
int rangestart = -1;
U8* bitmap = IS_ANYOF_TRIE(op) ? (U8*)ANYOF_BITMAP(o) : (U8*)TRIE_BITMAP(trie);
- Perl_sv_catpvf(aTHX_ sv, "[");
+ sv_catpvs(sv, "[");
for (i = 0; i <= 256; i++) {
if (i < 256 && BITMAP_TEST(bitmap,i)) {
if (rangestart == -1)
rangestart = -1;
}
}
- Perl_sv_catpvf(aTHX_ sv, "]");
+ sv_catpvs(sv, "]");
}
} else if (k == CURLY) {
}
else if (k == WHILEM && o->flags) /* Ordinal/of */
Perl_sv_catpvf(aTHX_ sv, "[%d/%d]", o->flags & 0xf, o->flags>>4);
- else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP || OP(o)==ACCEPT)
+ else if (k == REF || k == OPEN || k == CLOSE || k == GROUPP || OP(o)==ACCEPT) {
Perl_sv_catpvf(aTHX_ sv, "%d", (int)ARG(o)); /* Parenth number */
- else if (k == GOSUB)
+ if ( RXp_PAREN_NAMES(prog) ) {
+ if ( k != REF || OP(o) < NREF) {
+ AV *list= (AV *)progi->data->data[progi->name_list_idx];
+ SV **name= av_fetch(list, ARG(o), 0 );
+ if (name)
+ Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
+ }
+ else {
+ AV *list= (AV *)progi->data->data[ progi->name_list_idx ];
+ SV *sv_dat=(SV*)progi->data->data[ ARG( o ) ];
+ I32 *nums=(I32*)SvPVX(sv_dat);
+ SV **name= av_fetch(list, nums[0], 0 );
+ I32 n;
+ if (name) {
+ for ( n=0; n<SvIVX(sv_dat); n++ ) {
+ Perl_sv_catpvf(aTHX_ sv, "%s%"IVdf,
+ (n ? "," : ""), (IV)nums[n]);
+ }
+ Perl_sv_catpvf(aTHX_ sv, " '%"SVf"'", SVfARG(*name));
+ }
+ }
+ }
+ } else if (k == GOSUB)
Perl_sv_catpvf(aTHX_ sv, "%d[%+d]", (int)ARG(o),(int)ARG2L(o)); /* Paren and offset */
else if (k == VERB) {
if (!o->flags)
Perl_sv_catpvf(aTHX_ sv, ":%"SVf,
- (SV*)prog->data->data[ ARG( o ) ]);
+ SVfARG((SV*)progi->data->data[ ARG( o ) ]));
} else if (k == LOGICAL)
Perl_sv_catpvf(aTHX_ sv, "[%d]", o->flags); /* 2: embedded, otherwise 1 */
+ else if (k == FOLDCHAR)
+ Perl_sv_catpvf(aTHX_ sv, "[0x%"UVXf"]", PTR2UV(ARG(o)) );
else if (k == ANYOF) {
int i, rangestart = -1;
const U8 flags = ANYOF_FLAGS(o);
}
SV *
-Perl_re_intuit_string(pTHX_ regexp *prog)
+Perl_re_intuit_string(pTHX_ REGEXP * const r)
{ /* Assume that RE_INTUIT is set */
dVAR;
+ struct regexp *const prog = (struct regexp *)SvANY(r);
GET_RE_DEBUG_FLAGS_DECL;
PERL_UNUSED_CONTEXT;
}
/*
- pregfree - free a regexp
+ pregfree()
- See regdupe below if you change anything here.
+ handles refcounting and freeing the perl core regexp structure. When
+ it is necessary to actually free the structure the first thing it
+ does is call the 'free' method of the regexp_engine associated to to
+ the regexp, allowing the handling of the void *pprivate; member
+ first. (This routine is not overridable by extensions, which is why
+ the extensions free is called first.)
+
+ See regdupe and regdupe_internal if you change anything here.
*/
+#ifndef PERL_IN_XSUB_RE
+void
+Perl_pregfree(pTHX_ REGEXP *r)
+{
+ SvREFCNT_dec(r);
+}
void
-Perl_pregfree(pTHX_ struct regexp *r)
+Perl_pregfree2(pTHX_ REGEXP *rx)
{
dVAR;
-
+ struct regexp *const r = (struct regexp *)SvANY(rx);
GET_RE_DEBUG_FLAGS_DECL;
- if (!r || (--r->refcnt > 0))
- return;
+ if (r->mother_re) {
+ ReREFCNT_dec(r->mother_re);
+ } else {
+ CALLREGFREE_PVT(rx); /* free the private data */
+ if (RXp_PAREN_NAMES(r))
+ SvREFCNT_dec(RXp_PAREN_NAMES(r));
+ }
+ if (r->substrs) {
+ if (r->anchored_substr)
+ SvREFCNT_dec(r->anchored_substr);
+ if (r->anchored_utf8)
+ SvREFCNT_dec(r->anchored_utf8);
+ if (r->float_substr)
+ SvREFCNT_dec(r->float_substr);
+ if (r->float_utf8)
+ SvREFCNT_dec(r->float_utf8);
+ Safefree(r->substrs);
+ }
+ RX_MATCH_COPY_FREE(rx);
+#ifdef PERL_OLD_COPY_ON_WRITE
+ if (r->saved_copy)
+ SvREFCNT_dec(r->saved_copy);
+#endif
+ Safefree(r->swap);
+ Safefree(r->offs);
+}
+
+/* reg_temp_copy()
+
+ This is a hacky workaround to the structural issue of match results
+ being stored in the regexp structure which is in turn stored in
+ PL_curpm/PL_reg_curpm. The problem is that due to qr// the pattern
+ could be PL_curpm in multiple contexts, and could require multiple
+ result sets being associated with the pattern simultaneously, such
+ as when doing a recursive match with (??{$qr})
+
+ The solution is to make a lightweight copy of the regexp structure
+ when a qr// is returned from the code executed by (??{$qr}) this
+ lightweight copy doesnt actually own any of its data except for
+ the starp/end and the actual regexp structure itself.
+
+*/
+
+
+REGEXP *
+Perl_reg_temp_copy (pTHX_ REGEXP *rx) {
+ REGEXP *ret_x = (REGEXP*) newSV_type(SVt_REGEXP);
+ struct regexp *ret = (struct regexp *)SvANY(ret_x);
+ struct regexp *const r = (struct regexp *)SvANY(rx);
+ register const I32 npar = r->nparens+1;
+ (void)ReREFCNT_inc(rx);
+ /* We can take advantage of the existing "copied buffer" mechanism in SVs
+ by pointing directly at the buffer, but flagging that the allocated
+ space in the copy is zero. As we've just done a struct copy, it's now
+ a case of zero-ing that, rather than copying the current length. */
+ SvPV_set(ret_x, RX_WRAPPED(rx));
+ SvFLAGS(ret_x) |= SvFLAGS(rx) & (SVf_POK|SVp_POK|SVf_UTF8);
+ StructCopy(&(r->xpv_cur), &(ret->xpv_cur), struct regexp_allocated);
+ SvLEN_set(ret_x, 0);
+ Newx(ret->offs, npar, regexp_paren_pair);
+ Copy(r->offs, ret->offs, npar, regexp_paren_pair);
+ if (r->substrs) {
+ Newx(ret->substrs, 1, struct reg_substr_data);
+ StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
+
+ SvREFCNT_inc_void(ret->anchored_substr);
+ SvREFCNT_inc_void(ret->anchored_utf8);
+ SvREFCNT_inc_void(ret->float_substr);
+ SvREFCNT_inc_void(ret->float_utf8);
+
+ /* check_substr and check_utf8, if non-NULL, point to either their
+ anchored or float namesakes, and don't hold a second reference. */
+ }
+ RX_MATCH_COPIED_off(ret_x);
+#ifdef PERL_OLD_COPY_ON_WRITE
+ ret->saved_copy = NULL;
+#endif
+ ret->mother_re = rx;
+ ret->swap = NULL;
+
+ return ret_x;
+}
+#endif
+
+/* regfree_internal()
+
+ Free the private data in a regexp. This is overloadable by
+ extensions. Perl takes care of the regexp structure in pregfree(),
+ this covers the *pprivate pointer which technically perldoesnt
+ know about, however of course we have to handle the
+ regexp_internal structure when no extension is in use.
+
+ Note this is called before freeing anything in the regexp
+ structure.
+ */
+
+void
+Perl_regfree_internal(pTHX_ REGEXP * const rx)
+{
+ dVAR;
+ struct regexp *const r = (struct regexp *)SvANY(rx);
+ RXi_GET_DECL(r,ri);
+ GET_RE_DEBUG_FLAGS_DECL;
+
DEBUG_COMPILE_r({
if (!PL_colorset)
reginitcolors();
{
SV *dsv= sv_newmortal();
- RE_PV_QUOTED_DECL(s, (r->extflags & RXf_UTF8),
- dsv, r->precomp, r->prelen, 60);
+ RE_PV_QUOTED_DECL(s, RX_UTF8(rx),
+ dsv, RX_PRECOMP(rx), RX_PRELEN(rx), 60);
PerlIO_printf(Perl_debug_log,"%sFreeing REx:%s %s\n",
PL_colors[4],PL_colors[5],s);
}
});
-
- /* gcov results gave these as non-null 100% of the time, so there's no
- optimisation in checking them before calling Safefree */
- Safefree(r->precomp);
- Safefree(r->offsets); /* 20010421 MJD */
- RX_MATCH_COPY_FREE(r);
-#ifdef PERL_OLD_COPY_ON_WRITE
- if (r->saved_copy)
- SvREFCNT_dec(r->saved_copy);
+#ifdef RE_TRACK_PATTERN_OFFSETS
+ if (ri->u.offsets)
+ Safefree(ri->u.offsets); /* 20010421 MJD */
#endif
- if (r->substrs) {
- if (r->anchored_substr)
- SvREFCNT_dec(r->anchored_substr);
- if (r->anchored_utf8)
- SvREFCNT_dec(r->anchored_utf8);
- if (r->float_substr)
- SvREFCNT_dec(r->float_substr);
- if (r->float_utf8)
- SvREFCNT_dec(r->float_utf8);
- Safefree(r->substrs);
- }
- if (r->paren_names)
- SvREFCNT_dec(r->paren_names);
- if (r->data) {
- int n = r->data->count;
+ if (ri->data) {
+ int n = ri->data->count;
PAD* new_comppad = NULL;
PAD* old_comppad;
PADOFFSET refcnt;
while (--n >= 0) {
/* If you add a ->what type here, update the comment in regcomp.h */
- switch (r->data->what[n]) {
+ switch (ri->data->what[n]) {
case 's':
case 'S':
- SvREFCNT_dec((SV*)r->data->data[n]);
+ case 'u':
+ SvREFCNT_dec((SV*)ri->data->data[n]);
break;
case 'f':
- Safefree(r->data->data[n]);
+ Safefree(ri->data->data[n]);
break;
case 'p':
- new_comppad = (AV*)r->data->data[n];
+ new_comppad = (AV*)ri->data->data[n];
break;
case 'o':
if (new_comppad == NULL)
(SvTYPE(new_comppad) == SVt_PVAV) ? new_comppad : NULL
);
OP_REFCNT_LOCK;
- refcnt = OpREFCNT_dec((OP_4tree*)r->data->data[n]);
+ refcnt = OpREFCNT_dec((OP_4tree*)ri->data->data[n]);
OP_REFCNT_UNLOCK;
if (!refcnt)
- op_free((OP_4tree*)r->data->data[n]);
+ op_free((OP_4tree*)ri->data->data[n]);
PAD_RESTORE_LOCAL(old_comppad);
SvREFCNT_dec((SV*)new_comppad);
{ /* Aho Corasick add-on structure for a trie node.
Used in stclass optimization only */
U32 refcount;
- reg_ac_data *aho=(reg_ac_data*)r->data->data[n];
+ reg_ac_data *aho=(reg_ac_data*)ri->data->data[n];
OP_REFCNT_LOCK;
refcount = --aho->refcount;
OP_REFCNT_UNLOCK;
if ( !refcount ) {
- Safefree(aho->states);
- Safefree(aho->fail);
- aho->trie=NULL; /* not necessary to free this as it is
- handled by the 't' case */
- Safefree(r->data->data[n]); /* do this last!!!! */
- Safefree(r->regstclass);
+ PerlMemShared_free(aho->states);
+ PerlMemShared_free(aho->fail);
+ /* do this last!!!! */
+ PerlMemShared_free(ri->data->data[n]);
+ PerlMemShared_free(ri->regstclass);
}
}
break;
{
/* trie structure. */
U32 refcount;
- reg_trie_data *trie=(reg_trie_data*)r->data->data[n];
+ reg_trie_data *trie=(reg_trie_data*)ri->data->data[n];
OP_REFCNT_LOCK;
refcount = --trie->refcount;
OP_REFCNT_UNLOCK;
if ( !refcount ) {
- Safefree(trie->charmap);
- if (trie->widecharmap)
- SvREFCNT_dec((SV*)trie->widecharmap);
- Safefree(trie->states);
- Safefree(trie->trans);
+ PerlMemShared_free(trie->charmap);
+ PerlMemShared_free(trie->states);
+ PerlMemShared_free(trie->trans);
if (trie->bitmap)
- Safefree(trie->bitmap);
+ PerlMemShared_free(trie->bitmap);
if (trie->wordlen)
- Safefree(trie->wordlen);
+ PerlMemShared_free(trie->wordlen);
if (trie->jump)
- Safefree(trie->jump);
+ PerlMemShared_free(trie->jump);
if (trie->nextword)
- Safefree(trie->nextword);
-#ifdef DEBUGGING
- if (trie->words)
- SvREFCNT_dec((SV*)trie->words);
- if (trie->revcharmap)
- SvREFCNT_dec((SV*)trie->revcharmap);
-#endif
- Safefree(r->data->data[n]); /* do this last!!!! */
+ PerlMemShared_free(trie->nextword);
+ /* do this last!!!! */
+ PerlMemShared_free(ri->data->data[n]);
}
}
break;
default:
- Perl_croak(aTHX_ "panic: regfree data code '%c'", r->data->what[n]);
+ Perl_croak(aTHX_ "panic: regfree data code '%c'", ri->data->what[n]);
}
}
- Safefree(r->data->what);
- Safefree(r->data);
- }
- Safefree(r->startp);
- Safefree(r->endp);
- if (r->swap) {
- Safefree(r->swap->startp);
- Safefree(r->swap->endp);
- Safefree(r->swap);
+ Safefree(ri->data->what);
+ Safefree(ri->data);
}
- Safefree(r);
+
+ Safefree(ri);
}
#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
/*
- regdupe - duplicate a regexp.
-
- This routine is called by sv.c's re_dup and is expected to clone a
- given regexp structure. It is a no-op when not under USE_ITHREADS.
- (Originally this *was* re_dup() for change history see sv.c)
+ re_dup - duplicate a regexp.
- See pregfree() above if you change anything here.
+ This routine is expected to clone a given regexp structure. It is not
+ compiler under USE_ITHREADS.
+
+ After all of the core data stored in struct regexp is duplicated
+ the regexp_engine.dupe method is used to copy any private data
+ stored in the *pprivate pointer. This allows extensions to handle
+ any duplication it needs to do.
+
+ See pregfree() and regfree_internal() if you change anything here.
*/
#if defined(USE_ITHREADS)
-regexp *
-Perl_regdupe(pTHX_ const regexp *r, CLONE_PARAMS *param)
+#ifndef PERL_IN_XSUB_RE
+void
+Perl_re_dup_guts(pTHX_ const REGEXP *sstr, REGEXP *dstr, CLONE_PARAMS *param)
{
dVAR;
- REGEXP *ret;
- int i, len, npar;
- struct reg_substr_datum *s;
+ I32 npar;
+ const struct regexp *r = (const struct regexp *)SvANY(sstr);
+ struct regexp *ret = (struct regexp *)SvANY(dstr);
+
+ npar = r->nparens+1;
+ Newx(ret->offs, npar, regexp_paren_pair);
+ Copy(r->offs, ret->offs, npar, regexp_paren_pair);
+ if(ret->swap) {
+ /* no need to copy these */
+ Newx(ret->swap, npar, regexp_paren_pair);
+ }
- if (!r)
- return (REGEXP *)NULL;
+ if (ret->substrs) {
+ /* Do it this way to avoid reading from *r after the StructCopy().
+ That way, if any of the sv_dup_inc()s dislodge *r from the L1
+ cache, it doesn't matter. */
+ const bool anchored = r->check_substr
+ ? r->check_substr == r->anchored_substr
+ : r->check_utf8 == r->anchored_utf8;
+ Newx(ret->substrs, 1, struct reg_substr_data);
+ StructCopy(r->substrs, ret->substrs, struct reg_substr_data);
+
+ ret->anchored_substr = sv_dup_inc(ret->anchored_substr, param);
+ ret->anchored_utf8 = sv_dup_inc(ret->anchored_utf8, param);
+ ret->float_substr = sv_dup_inc(ret->float_substr, param);
+ ret->float_utf8 = sv_dup_inc(ret->float_utf8, param);
+
+ /* check_substr and check_utf8, if non-NULL, point to either their
+ anchored or float namesakes, and don't hold a second reference. */
+
+ if (ret->check_substr) {
+ if (anchored) {
+ assert(r->check_utf8 == r->anchored_utf8);
+ ret->check_substr = ret->anchored_substr;
+ ret->check_utf8 = ret->anchored_utf8;
+ } else {
+ assert(r->check_substr == r->float_substr);
+ assert(r->check_utf8 == r->float_utf8);
+ ret->check_substr = ret->float_substr;
+ ret->check_utf8 = ret->float_utf8;
+ }
+ } else if (ret->check_utf8) {
+ if (anchored) {
+ ret->check_utf8 = ret->anchored_utf8;
+ } else {
+ ret->check_utf8 = ret->float_utf8;
+ }
+ }
+ }
- if ((ret = (REGEXP *)ptr_table_fetch(PL_ptr_table, r)))
- return ret;
+ RXp_PAREN_NAMES(ret) = hv_dup_inc(RXp_PAREN_NAMES(ret), param);
- len = r->offsets[0];
- npar = r->nparens+1;
+ if (ret->pprivate)
+ RXi_SET(ret,CALLREGDUPE_PVT(dstr,param));
- Newxc(ret, sizeof(regexp) + (len+1)*sizeof(regnode), char, regexp);
- Copy(r->program, ret->program, len+1, regnode);
+ if (RX_MATCH_COPIED(dstr))
+ ret->subbeg = SAVEPVN(ret->subbeg, ret->sublen);
+ else
+ ret->subbeg = NULL;
+#ifdef PERL_OLD_COPY_ON_WRITE
+ ret->saved_copy = NULL;
+#endif
- Newx(ret->startp, npar, I32);
- Copy(r->startp, ret->startp, npar, I32);
- Newx(ret->endp, npar, I32);
- Copy(r->startp, ret->startp, npar, I32);
- if(r->swap) {
- Newx(ret->swap, 1, regexp_paren_ofs);
- /* no need to copy these */
- Newx(ret->swap->startp, npar, I32);
- Newx(ret->swap->endp, npar, I32);
- } else {
- ret->swap = NULL;
- }
+ ret->mother_re = NULL;
+ ret->gofs = 0;
+}
+#endif /* PERL_IN_XSUB_RE */
- Newx(ret->substrs, 1, struct reg_substr_data);
- for (s = ret->substrs->data, i = 0; i < 3; i++, s++) {
- s->min_offset = r->substrs->data[i].min_offset;
- s->max_offset = r->substrs->data[i].max_offset;
- s->end_shift = r->substrs->data[i].end_shift;
- s->substr = sv_dup_inc(r->substrs->data[i].substr, param);
- s->utf8_substr = sv_dup_inc(r->substrs->data[i].utf8_substr, param);
- }
+/*
+ regdupe_internal()
+
+ This is the internal complement to regdupe() which is used to copy
+ the structure pointed to by the *pprivate pointer in the regexp.
+ This is the core version of the extension overridable cloning hook.
+ The regexp structure being duplicated will be copied by perl prior
+ to this and will be provided as the regexp *r argument, however
+ with the /old/ structures pprivate pointer value. Thus this routine
+ may override any copying normally done by perl.
+
+ It returns a pointer to the new regexp_internal structure.
+*/
+
+void *
+Perl_regdupe_internal(pTHX_ REGEXP * const rx, CLONE_PARAMS *param)
+{
+ dVAR;
+ struct regexp *const r = (struct regexp *)SvANY(rx);
+ regexp_internal *reti;
+ int len, npar;
+ RXi_GET_DECL(r,ri);
+
+ npar = r->nparens+1;
+ len = ProgLen(ri);
+
+ Newxc(reti, sizeof(regexp_internal) + (len+1)*sizeof(regnode), char, regexp_internal);
+ Copy(ri->program, reti->program, len+1, regnode);
+
+
+ reti->regstclass = NULL;
- ret->regstclass = NULL;
- if (r->data) {
+ if (ri->data) {
struct reg_data *d;
- const int count = r->data->count;
+ const int count = ri->data->count;
int i;
Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
d->count = count;
for (i = 0; i < count; i++) {
- d->what[i] = r->data->what[i];
+ d->what[i] = ri->data->what[i];
switch (d->what[i]) {
- /* legal options are one of: sSfpontT
+ /* legal options are one of: sSfpontTu
see also regcomp.h and pregfree() */
case 's':
case 'S':
- d->data[i] = sv_dup_inc((SV *)r->data->data[i], param);
- break;
- case 'p':
- d->data[i] = av_dup_inc((AV *)r->data->data[i], param);
+ case 'p': /* actually an AV, but the dup function is identical. */
+ case 'u': /* actually an HV, but the dup function is identical. */
+ d->data[i] = sv_dup_inc((SV *)ri->data->data[i], param);
break;
case 'f':
/* This is cheating. */
Newx(d->data[i], 1, struct regnode_charclass_class);
- StructCopy(r->data->data[i], d->data[i],
+ StructCopy(ri->data->data[i], d->data[i],
struct regnode_charclass_class);
- ret->regstclass = (regnode*)d->data[i];
+ reti->regstclass = (regnode*)d->data[i];
break;
case 'o':
/* Compiled op trees are readonly and in shared memory,
and can thus be shared without duplication. */
OP_REFCNT_LOCK;
- d->data[i] = (void*)OpREFCNT_inc((OP*)r->data->data[i]);
- OP_REFCNT_UNLOCK;
- break;
- case 'n':
- d->data[i] = r->data->data[i];
- break;
- case 't':
- d->data[i] = r->data->data[i];
- OP_REFCNT_LOCK;
- ((reg_trie_data*)d->data[i])->refcount++;
+ d->data[i] = (void*)OpREFCNT_inc((OP*)ri->data->data[i]);
OP_REFCNT_UNLOCK;
break;
case 'T':
- d->data[i] = r->data->data[i];
- OP_REFCNT_LOCK;
- ((reg_ac_data*)d->data[i])->refcount++;
- OP_REFCNT_UNLOCK;
/* Trie stclasses are readonly and can thus be shared
* without duplication. We free the stclass in pregfree
* when the corresponding reg_ac_data struct is freed.
*/
- ret->regstclass= r->regstclass;
+ reti->regstclass= ri->regstclass;
+ /* Fall through */
+ case 't':
+ OP_REFCNT_LOCK;
+ ((reg_trie_data*)ri->data->data[i])->refcount++;
+ OP_REFCNT_UNLOCK;
+ /* Fall through */
+ case 'n':
+ d->data[i] = ri->data->data[i];
break;
default:
- Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", r->data->what[i]);
+ Perl_croak(aTHX_ "panic: re_dup unknown data code '%c'", ri->data->what[i]);
}
}
- ret->data = d;
+ reti->data = d;
}
else
- ret->data = NULL;
-
- Newx(ret->offsets, 2*len+1, U32);
- Copy(r->offsets, ret->offsets, 2*len+1, U32);
-
- ret->precomp = SAVEPVN(r->precomp, r->prelen);
- ret->refcnt = r->refcnt;
- ret->minlen = r->minlen;
- ret->minlenret = r->minlenret;
- ret->prelen = r->prelen;
- ret->nparens = r->nparens;
- ret->lastparen = r->lastparen;
- ret->lastcloseparen = r->lastcloseparen;
- ret->intflags = r->intflags;
- ret->extflags = r->extflags;
+ reti->data = NULL;
- ret->sublen = r->sublen;
+ reti->name_list_idx = ri->name_list_idx;
- ret->engine = r->engine;
-
- ret->paren_names = hv_dup_inc(r->paren_names, param);
-
- if (RX_MATCH_COPIED(ret))
- ret->subbeg = SAVEPVN(r->subbeg, r->sublen);
- else
- ret->subbeg = NULL;
-#ifdef PERL_OLD_COPY_ON_WRITE
- ret->saved_copy = NULL;
+#ifdef RE_TRACK_PATTERN_OFFSETS
+ if (ri->u.offsets) {
+ Newx(reti->u.offsets, 2*len+1, U32);
+ Copy(ri->u.offsets, reti->u.offsets, 2*len+1, U32);
+ }
+#else
+ SetProgLen(reti,len);
#endif
- ptr_table_store(PL_ptr_table, r, ret);
- return ret;
-}
-#endif
-
-/*
- reg_stringify()
-
- converts a regexp embedded in a MAGIC struct to its stringified form,
- caching the converted form in the struct and returns the cached
- string.
-
- If lp is nonnull then it is used to return the length of the
- resulting string
-
- If flags is nonnull and the returned string contains UTF8 then
- (flags & 1) will be true.
-
- If haseval is nonnull then it is used to return whether the pattern
- contains evals.
-
- Normally called via macro:
-
- CALLREG_STRINGIFY(mg,0,0);
-
- And internally with
-
- CALLREG_AS_STR(mg,lp,flags,haseval)
-
- See sv_2pv_flags() in sv.c for an example of internal usage.
-
- */
-
-char *
-Perl_reg_stringify(pTHX_ MAGIC *mg, STRLEN *lp, U32 *flags, I32 *haseval ) {
- dVAR;
- const regexp * const re = (regexp *)mg->mg_obj;
-
- if (!mg->mg_ptr) {
- const char *fptr = "msix";
- char reflags[6];
- char ch;
- int left = 0;
- int right = 4;
- bool need_newline = 0;
- U16 reganch = (U16)((re->extflags & RXf_PMf_COMPILETIME) >> 12);
-
- while((ch = *fptr++)) {
- if(reganch & 1) {
- reflags[left++] = ch;
- }
- else {
- reflags[right--] = ch;
- }
- reganch >>= 1;
- }
- if(left != 4) {
- reflags[left] = '-';
- left = 5;
- }
-
- mg->mg_len = re->prelen + 4 + left;
- /*
- * If /x was used, we have to worry about a regex ending with a
- * comment later being embedded within another regex. If so, we don't
- * want this regex's "commentization" to leak out to the right part of
- * the enclosing regex, we must cap it with a newline.
- *
- * So, if /x was used, we scan backwards from the end of the regex. If
- * we find a '#' before we find a newline, we need to add a newline
- * ourself. If we find a '\n' first (or if we don't find '#' or '\n'),
- * we don't need to add anything. -jfriedl
- */
- if (PMf_EXTENDED & re->extflags) {
- const char *endptr = re->precomp + re->prelen;
- while (endptr >= re->precomp) {
- const char c = *(endptr--);
- if (c == '\n')
- break; /* don't need another */
- if (c == '#') {
- /* we end while in a comment, so we need a newline */
- mg->mg_len++; /* save space for it */
- need_newline = 1; /* note to add it */
- break;
- }
- }
- }
-
- Newx(mg->mg_ptr, mg->mg_len + 1 + left, char);
- mg->mg_ptr[0] = '(';
- mg->mg_ptr[1] = '?';
- Copy(reflags, mg->mg_ptr+2, left, char);
- *(mg->mg_ptr+left+2) = ':';
- Copy(re->precomp, mg->mg_ptr+3+left, re->prelen, char);
- if (need_newline)
- mg->mg_ptr[mg->mg_len - 2] = '\n';
- mg->mg_ptr[mg->mg_len - 1] = ')';
- mg->mg_ptr[mg->mg_len] = 0;
- }
- if (haseval)
- *haseval = re->program[0].next_off;
- if (flags)
- *flags = ((re->extflags & RXf_UTF8) ? 1 : 0);
-
- if (lp)
- *lp = mg->mg_len;
- return mg->mg_ptr;
+ return (void*)reti;
}
+#endif /* USE_ITHREADS */
#ifndef PERL_IN_XSUB_RE
+
/*
- regnext - dig the "next" pointer out of a node
*/
dVAR;
register I32 offset;
- if (p == &PL_regdummy)
+ if (!p)
return(NULL);
offset = (reg_off_by_arg[OP(p)] ? ARG(p) : NEXT_OFF(p));
const REGEXP * const rx = PM_GETRE(PL_curpm);
if (rx) {
U32 i;
- for (i = 1; i <= rx->nparens; i++) {
+ for (i = 1; i <= RX_NPARENS(rx); i++) {
char digits[TYPE_CHARS(long)];
const STRLEN len = my_snprintf(digits, sizeof(digits), "%lu", (long)i);
GV *const *const gvp
clear_re(pTHX_ void *r)
{
dVAR;
- ReREFCNT_dec((regexp *)r);
+ ReREFCNT_dec((REGEXP *)r);
}
#ifdef DEBUGGING
STATIC void
S_put_byte(pTHX_ SV *sv, int c)
{
- if (isCNTRL(c) || c == 255 || !isPRINT(c))
+ /* Our definition of isPRINT() ignores locales, so only bytes that are
+ not part of UTF-8 are considered printable. I assume that the same
+ holds for UTF-EBCDIC.
+ Also, code point 255 is not printable in either (it's E0 in EBCDIC,
+ which Wikipedia says:
+
+ EO, or Eight Ones, is an 8-bit EBCDIC character code represented as all
+ ones (binary 1111 1111, hexadecimal FF). It is similar, but not
+ identical, to the ASCII delete (DEL) or rubout control character.
+ ) So the old condition can be simplified to !isPRINT(c) */
+ if (!isPRINT(c))
Perl_sv_catpvf(aTHX_ sv, "\\%o", c);
- else if (c == '-' || c == ']' || c == '\\' || c == '^')
- Perl_sv_catpvf(aTHX_ sv, "\\%c", c);
- else
- Perl_sv_catpvf(aTHX_ sv, "%c", c);
+ else {
+ const char string = c;
+ if (c == '-' || c == ']' || c == '\\' || c == '^')
+ sv_catpvs(sv, "\\");
+ sv_catpvn(sv, &string, 1);
+ }
}
#define CLEAR_OPTSTART \
if (optstart) STMT_START { \
- DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log, " (%d nodes)\n", node - optstart)); \
+ DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log, " (%"IVdf" nodes)\n", (IV)(node - optstart))); \
optstart=NULL; \
} STMT_END
register U8 op = PSEUDO; /* Arbitrary non-END op. */
register const regnode *next;
const regnode *optstart= NULL;
+
+ RXi_GET_DECL(r,ri);
GET_RE_DEBUG_FLAGS_DECL;
-
+
#ifdef DEBUG_DUMPUNTIL
PerlIO_printf(Perl_debug_log, "--- %d : %d - %d - %d\n",indent,node-start,
last ? last-start : 0,plast ? plast-start : 0);
while (PL_regkind[op] != END && (!last || node < last)) {
/* While that wasn't END last time... */
-
NODE_ALIGN(node);
op = OP(node);
if (op == CLOSE || op == WHILEM)
indent--;
next = regnext((regnode *)node);
-
+
/* Where, what. */
if (OP(node) == OPTIMIZED) {
if (!optstart && RE_DEBUG_FLAG(RE_DEBUG_COMPILE_OPTIMISE))
goto after_print;
} else
CLEAR_OPTSTART;
-
+
regprop(r, sv, node);
PerlIO_printf(Perl_debug_log, "%4"IVdf":%*s%s", (IV)(node - start),
(int)(2*indent + 1), "", SvPVX_const(sv));
-
- if (OP(node) != OPTIMIZED) {
- if (next == NULL) /* Next ptr. */
- PerlIO_printf(Perl_debug_log, "(0)");
- else if (PL_regkind[(U8)op] == BRANCH && PL_regkind[OP(next)] != BRANCH )
- PerlIO_printf(Perl_debug_log, "(FAIL)");
- else
- PerlIO_printf(Perl_debug_log, "(%"IVdf")", (IV)(next - start));
-
- /*if (PL_regkind[(U8)op] != TRIE)*/
- (void)PerlIO_putc(Perl_debug_log, '\n');
- }
-
+
+ if (OP(node) != OPTIMIZED) {
+ if (next == NULL) /* Next ptr. */
+ PerlIO_printf(Perl_debug_log, " (0)");
+ else if (PL_regkind[(U8)op] == BRANCH && PL_regkind[OP(next)] != BRANCH )
+ PerlIO_printf(Perl_debug_log, " (FAIL)");
+ else
+ PerlIO_printf(Perl_debug_log, " (%"IVdf")", (IV)(next - start));
+ (void)PerlIO_putc(Perl_debug_log, '\n');
+ }
+
after_print:
if (PL_regkind[(U8)op] == BRANCHJ) {
assert(next);
else if ( PL_regkind[(U8)op] == TRIE ) {
const regnode *this_trie = node;
const char op = OP(node);
- const I32 n = ARG(node);
+ const U32 n = ARG(node);
const reg_ac_data * const ac = op>=AHOCORASICK ?
- (reg_ac_data *)r->data->data[n] :
+ (reg_ac_data *)ri->data->data[n] :
NULL;
- const reg_trie_data * const trie = op<AHOCORASICK ?
- (reg_trie_data*)r->data->data[n] :
- ac->trie;
+ const reg_trie_data * const trie =
+ (reg_trie_data*)ri->data->data[op<AHOCORASICK ? n : ac->trie];
+#ifdef DEBUGGING
+ AV *const trie_words = (AV *) ri->data->data[n + TRIE_WORDS_OFFSET];
+#endif
const regnode *nextbranch= NULL;
I32 word_idx;
sv_setpvn(sv, "", 0);
for (word_idx= 0; word_idx < (I32)trie->wordcount; word_idx++) {
- SV ** const elem_ptr = av_fetch(trie->words,word_idx,0);
+ SV ** const elem_ptr = av_fetch(trie_words,word_idx,0);
PerlIO_printf(Perl_debug_log, "%*s%s ",
(int)(2*(indent+3)), "",
elem_ptr ? pv_pretty(sv, SvPV_nolen_const(*elem_ptr), SvCUR(*elem_ptr), 60,
PL_colors[0], PL_colors[1],
(SvUTF8(*elem_ptr) ? PERL_PV_ESCAPE_UNI : 0) |
- PERL_PV_PRETTY_ELIPSES |
+ PERL_PV_PRETTY_ELLIPSES |
PERL_PV_PRETTY_LTGT
)
: "???"
);
if (trie->jump) {
U16 dist= trie->jump[word_idx+1];
- PerlIO_printf(Perl_debug_log, "(%u)\n",
- (dist ? this_trie + dist : next) - start);
+ PerlIO_printf(Perl_debug_log, "(%"UVuf")\n",
+ (UV)((dist ? this_trie + dist : next) - start));
if (dist) {
if (!nextbranch)
nextbranch= this_trie + trie->jump[0];
}
CLEAR_OPTSTART;
#ifdef DEBUG_DUMPUNTIL
- PerlIO_printf(Perl_debug_log, "--- %d\n",indent);
+ PerlIO_printf(Perl_debug_log, "--- %d\n", (int)indent);
#endif
return node;
}