**** 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, ...? */
U32 seen;
I32 size; /* Code size. */
- I32 npar; /* () count. */
+ I32 npar; /* Capture buffer count, (OPEN). */
+ I32 cpar; /* Capture buffer count, (CLOSE). */
+ I32 nestroot; /* root parens we are in - used by accept */
I32 extralen;
I32 seen_zerolen;
I32 seen_evals;
- I32 utf8;
+ regnode **open_parens; /* pointers to open parens */
+ regnode **close_parens; /* pointers to close parens */
+ regnode *opend; /* END node in program */
+ 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
char *starttry; /* -Dr: where regtry was called. */
#define RExC_starttry (pRExC_state->starttry)
#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_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_opend (pRExC_state->opend)
+#define RExC_paren_names (pRExC_state->paren_names)
+#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) == '?' || \
* 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)
#endif
+
+#define PBYTE(u8str,paren) ((U8*)(u8str))[(paren) >> 3]
+#define PBITVAL(paren) (1 << ((paren) & 7))
+#define PAREN_TEST(u8str,paren) ( PBYTE(u8str,paren) & PBITVAL(paren))
+#define PAREN_SET(u8str,paren) PBYTE(u8str,paren) |= PBITVAL(paren)
+#define PAREN_UNSET(u8str,paren) PBYTE(u8str,paren) &= (~PBITVAL(paren))
+
+
/* About scan_data_t.
During optimisation we recurse through the regexp program performing
#define SCF_WHILEM_VISITED_POS 0x2000
#define SCF_TRIE_RESTUDY 0x4000 /* Do restudy? */
-
+#define SCF_SEEN_ACCEPT 0x8000
#define UTF (RExC_utf8 != 0)
-#define LOC ((RExC_flags & PMf_LOCALE) != 0)
-#define FOLD ((RExC_flags & PMf_FOLD) != 0)
+#define LOC ((RExC_flags & RXf_PMf_LOCALE) != 0)
+#define FOLD ((RExC_flags & RXf_PMf_FOLD) != 0)
#define OOB_UNICODE 12345678
#define OOB_NAMEDCLASS -1
* arg. Show regex, up to a maximum length. If it's too long, chop and add
* "...".
*/
-#define FAIL(msg) STMT_START { \
+#define _FAIL(code) STMT_START { \
const char *ellipses = ""; \
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; \
ellipses = "..."; \
} \
- Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
- msg, (int)len, RExC_precomp, ellipses); \
+ code; \
} STMT_END
+#define FAIL(msg) _FAIL( \
+ Perl_croak(aTHX_ "%s in regex m/%.*s%s/", \
+ msg, (int)len, RExC_precomp, ellipses))
+
+#define FAIL2(msg,arg) _FAIL( \
+ Perl_croak(aTHX_ msg " in regex m/%.*s%s/", \
+ arg, (int)len, RExC_precomp, ellipses))
+
/*
* Simple_vFAIL -- like FAIL, but marks the current location in the scan
*/
*/
#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", \
- __LINE__, (node), (int)(byte))); \
+ __LINE__, (int)(node), (int)(byte))); \
if((node) < 0) { \
Perl_croak(aTHX_ "value of node is %d in Offset macro", (int)(node)); \
} else { \
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_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) */
S_cl_and(struct regnode_charclass_class *cl,
const struct regnode_charclass_class *and_with)
{
+
+ assert(and_with->type == ANYOF);
if (!(and_with->flags & ANYOF_CLASS)
&& !(cl->flags & ANYOF_CLASS)
&& (and_with->flags & ANYOF_LOCALE) == (cl->flags & ANYOF_LOCALE)
#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,
PerlIO_printf( Perl_debug_log, "%.*s", colwidth, "--------");
PerlIO_printf( Perl_debug_log, "\n");
- for( state = 1 ; state < trie->laststate ; state++ ) {
+ for( state = 1 ; state < trie->statecount ; state++ ) {
const U32 base = trie->states[ state ].trans.base;
PerlIO_printf( Perl_debug_log, "%*s#%4"UVXf"|", (int)depth * 2 + 2,"", (UV)state);
}
}
/*
- 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,
(SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
PERL_PV_ESCAPE_FIRSTCHAR
) ,
- TRIE_LIST_ITEM(state,charid).forid,
- (UV)TRIE_LIST_ITEM(state,charid).newstate
- );
- }
+ TRIE_LIST_ITEM(state,charid).forid,
+ (UV)TRIE_LIST_ITEM(state,charid).newstate
+ );
+ if (!(charid % 10))
+ PerlIO_printf(Perl_debug_log, "\n%*s| ",
+ (int)((depth * 2) + 14), "");
+ }
}
PerlIO_printf( Perl_debug_log, "\n");
}
}
/*
- 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,
*(d++) = uv;
*/
-#define TRIE_STORE_REVCHAR \
+#define TRIE_STORE_REVCHAR \
STMT_START { \
- SV *tmp = Perl_newSVpvf_nocontext( "%c", (int)uvc ); \
- if (UTF) SvUTF8_on(tmp); \
- 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++; \
#define TRIE_LIST_PUSH(state,fid,ns) STMT_START { \
if ( TRIE_LIST_CUR( state ) >=TRIE_LIST_LEN( state ) ) { \
- TRIE_LIST_LEN( state ) *= 2; \
- Renew( trie->states[ state ].trans.list, \
- TRIE_LIST_LEN( state ), reg_trie_trans_le ); \
+ U32 ging = TRIE_LIST_LEN( state ) *= 2; \
+ Renew( trie->states[ state ].trans.list, ging, reg_trie_trans_le ); \
} \
TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).forid = fid; \
TRIE_LIST_ITEM( state, TRIE_LIST_CUR( state ) ).newstate = ns; \
/* 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[curword] = (U16)(tail - noper_next); \
+ trie->jump = (U16 *) PerlMemShared_calloc( word_count + 1, sizeof(U16) ); \
+ trie->jump[curword] = (U16)(noper_next - convert); \
if (!jumper) \
jumper = noper_next; \
if (!nextbranch) \
/* 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;
U32 next_alloc = 0;
regnode *jumper = NULL;
regnode *nextbranch = NULL;
+ regnode *convert = NULL;
/* we just use folder as a flag in utf8 */
const U8 * const folder = ( flags == EXACTF
? PL_fold
)
);
- 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)) {
REG_NODE_NUM(last), REG_NODE_NUM(tail),
(int)depth);
});
+
+ /* Find the node we are going to overwrite */
+ if ( first == startbranch && OP( last ) != BRANCH ) {
+ /* whole branch chain */
+ convert = first;
+ } else {
+ /* branch sub-chain */
+ convert = NEXTOPER( first );
+ }
+
/* -- First loop and Setup --
We first traverse the branches and scan each word to determine if it
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
STRLEN transcount = 1;
- Newxz( trie->states, TRIE_CHARCOUNT(trie) + 2, reg_trie_state );
+ DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
+ "%*sCompiling trie using list compiler\n",
+ (int)depth * 2 + 2, ""));
+
+ 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 {
} /* end second pass */
- trie->laststate = next_alloc;
- Renew( trie->states, next_alloc, reg_trie_state );
+ /* next alloc is the NEXT state to be allocated */
+ trie->statecount = next_alloc;
+ 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;
use TRIE_NODENUM() to convert.
*/
-
-
- Newxz( trie->trans, ( TRIE_CHARCOUNT(trie) + 1 ) * trie->uniquecharcount + 1,
- reg_trie_trans );
- Newxz( trie->states, TRIE_CHARCOUNT(trie) + 2, reg_trie_state );
+ DEBUG_TRIE_COMPILE_MORE_r( PerlIO_printf( Perl_debug_log,
+ "%*sCompiling trie using table compiler\n",
+ (int)depth * 2 + 2, ""));
+
+ 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));
{
/*
const U32 laststate = TRIE_NODENUM( next_alloc );
U32 state, charid;
U32 pos = 0, zp=0;
- trie->laststate = laststate;
+ trie->statecount = laststate;
for ( state = 1 ; state < laststate ; state++ ) {
U8 flag = 0;
}
}
trie->lasttrans = pos + 1;
- Renew( trie->states, laststate + 1, 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",
} /* end table compress */
}
+ DEBUG_TRIE_COMPILE_MORE_r(
+ PerlIO_printf(Perl_debug_log, "%*sStatecount:%"UVxf" Lasttrans:%"UVxf"\n",
+ (int)depth * 2 + 2, "",
+ (UV)trie->statecount,
+ (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*/
- regnode *convert;
U8 nodetype =(U8)(flags & 0xFF);
char *str=NULL;
#ifdef DEBUGGING
-
- U32 mjd_offset;
- U32 mjd_nodelen;
-#endif
+ regnode *optimize = NULL;
+#ifdef RE_TRACK_PATTERN_OFFSETS
+
+ U32 mjd_offset = 0;
+ U32 mjd_nodelen = 0;
+#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
the whole branch sequence, including the first.
*/
/* Find the node we are going to overwrite */
- if ( first == startbranch && OP( last ) != BRANCH ) {
- /* whole branch chain */
- convert = first;
- DEBUG_r({
- const regnode *nop = NEXTOPER( convert );
- mjd_offset= Node_Offset((nop));
- mjd_nodelen= Node_Length((nop));
- });
- } else {
+ if ( first != startbranch || OP( last ) == BRANCH ) {
/* branch sub-chain */
- convert = NEXTOPER( first );
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 */
+ }
+#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;
- DEBUG_OPTIMISE_r(
- PerlIO_printf(Perl_debug_log, "%*sLaststate:%"UVuf"\n",
- (int)depth * 2 + 2, "",
- (UV)trie->laststate)
- );
- for ( state = 1 ; state < trie->laststate-1 ; state++ ) {
+ for ( state = 1 ; state < trie->statecount-1 ; state++ ) {
U32 ofs = 0;
I32 idx = -1;
U32 count = 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);
- const char *ch = SvPV_nolen_const( *tmp );
- DEBUG_OPTIMISE_r(
+ 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,
"%*sPrefix State: %"UVuf" Idx:%"UVuf" Char='%s'\n",
(int)depth * 2 + 2, "",
- (UV)state, (UV)idx, ch)
- );
+ (UV)state, (UV)idx,
+ pv_pretty(sv, SvPV_nolen_const(*tmp), SvCUR(*tmp), 6,
+ PL_colors[0], PL_colors[1],
+ (SvUTF8(*tmp) ? PERL_PV_ESCAPE_UNI : 0) |
+ PERL_PV_ESCAPE_FIRSTCHAR
+ )
+ );
+ });
if ( state==1 ) {
OP( convert ) = nodetype;
str=STRING(convert);
STR_LEN(convert)=0;
}
- *str++=*ch;
- STR_LEN(convert)++;
-
+ STR_LEN(convert) += len;
+ while (len--)
+ *str++ = *ch++;
} else {
#ifdef DEBUGGING
if (state>1)
trie->startstate = state;
trie->minlen -= (state - 1);
trie->maxlen -= (state - 1);
- DEBUG_r({
- regnode *fix = convert;
- mjd_nodelen++;
- Set_Node_Offset_Length(convert, mjd_offset, state - 1);
- while( ++fix < n ) {
- Set_Node_Offset_Length(fix, 0, 0);
- }
- });
+#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 {
NEXT_OFF(convert) = (U16)(tail - convert);
+ DEBUG_r(optimize= n);
}
}
}
+ if (!jumper)
+ jumper = last;
if ( trie->maxlen ) {
NEXT_OFF( convert ) = (U16)(tail - convert);
ARG_SET( convert, data_slot );
jump[0], which is otherwise unused by the jump logic.
We use this when dumping a trie and during optimisation. */
if (trie->jump)
- trie->jump[0] = (U16)(tail - nextbranch);
- if (!jumper)
- jumper = last;
+ trie->jump[0] = (U16)(nextbranch - convert);
+
/* XXXX */
if ( !trie->states[trie->startstate].wordnum && trie->bitmap &&
( (char *)jumper - (char *)convert) >= (int)sizeof(struct regnode_charclass) )
{
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;
/* store the type in the flags */
convert->flags = nodetype;
- /* XXX We really should free up the resource in trie now, as we wont use them */
+ DEBUG_r({
+ optimize = convert
+ + NODE_STEP_REGNODE
+ + regarglen[ OP( convert ) ];
+ });
+ /* XXX We really should free up the resource in trie now,
+ as we won't use them - (which resources?) dmq */
}
/* needed for dumping*/
- DEBUG_r({
- regnode *optimize = convert
- + NODE_STEP_REGNODE
- + regarglen[ OP( convert ) ];
+ DEBUG_r(if (optimize) {
regnode *opt = convert;
- while (++opt<optimize) {
+
+ 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->laststate;
+ const U32 numstates = trie->statecount;
const U32 ubound = trie->lasttrans + ucharcount;
U32 q_read = 0;
U32 q_write = 0;
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,
- (trie->laststate+1)*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
*/
fail[ 0 ] = fail[ 1 ] = 0;
DEBUG_TRIE_COMPILE_r({
- PerlIO_printf(Perl_debug_log, "%*sStclass Failtable: 0", (int)(depth * 2), "");
+ PerlIO_printf(Perl_debug_log,
+ "%*sStclass Failtable (%"UVuf" states): 0",
+ (int)(depth * 2), "", (UV)numstates
+ );
for( q_read=1; q_read<numstates; q_read++ ) {
PerlIO_printf(Perl_debug_log, ", %"UVuf, (UV)fail[q_read]);
}
#endif
#define DEBUG_PEEP(str,scan,depth) \
- DEBUG_OPTIMISE_r({ \
+ DEBUG_OPTIMISE_r({if (scan){ \
SV * const mysv=sv_newmortal(); \
regnode *Next = regnext(scan); \
regprop(RExC_rx, mysv, scan); \
- PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s [%d]\n", \
+ PerlIO_printf(Perl_debug_log, "%*s" str ">%3d: %s (%d)\n", \
(int)depth*2, "", REG_NODE_NUM(scan), SvPV_nolen_const(mysv),\
Next ? (REG_NODE_NUM(Next)) : 0 ); \
- });
+ }});
/* Stops at toplevel WHILEM as well as at "last". At end *scanp is set
to the position after last scanned or to NULL. */
+#define INIT_AND_WITHP \
+ assert(!and_withp); \
+ Newx(and_withp,1,struct regnode_charclass_class); \
+ SAVEFREEPV(and_withp)
+
+/* this is a chain of data about sub patterns we are processing that
+ need to be handled seperately/specially in study_chunk. Its so
+ we can simulate recursion without losing state. */
+struct scan_frame;
+typedef struct scan_frame {
+ regnode *last; /* last node to process in this frame */
+ regnode *next; /* next node to process when last is reached */
+ struct scan_frame *prev; /*previous frame*/
+ 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,
+S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
I32 *minlenp, I32 *deltap,
- regnode *last, scan_data_t *data, U32 flags, U32 depth)
+ regnode *last,
+ scan_data_t *data,
+ I32 stopparen,
+ U8* recursed,
+ struct regnode_charclass_class *and_withp,
+ U32 flags, U32 depth)
/* scanp: Start here (read-write). */
/* deltap: Write maxlen-minlen here. */
/* last: Stop before this one. */
+ /* data: string data about the pattern */
+ /* stopparen: treat close N as END */
+ /* recursed: which subroutines have we recursed into */
+ /* and_withp: Valid if flags & SCF_DO_STCLASS_OR */
{
dVAR;
I32 min = 0, pars = 0, code;
int is_inf_internal = 0; /* The studied chunk is infinite */
I32 is_par = OP(scan) == OPEN ? ARG(scan) : 0;
scan_data_t data_fake;
- struct regnode_charclass_class and_with; /* Valid if flags & SCF_DO_STCLASS_OR */
SV *re_trie_maxbuff = NULL;
regnode *first_non_open = scan;
-
+ I32 stopmin = I32_MAX;
+ scan_frame *frame = NULL;
GET_RE_DEBUG_FLAGS_DECL;
+
#ifdef DEBUGGING
- StructCopy(&zero_scan_data, &data_fake, scan_data_t);
+ StructCopy(&zero_scan_data, &data_fake, scan_data_t);
#endif
+
if ( depth == 0 ) {
- while (first_non_open && OP(first_non_open) == OPEN)
+ while (first_non_open && OP(first_non_open) == OPEN)
first_non_open=regnext(first_non_open);
}
- while (scan && OP(scan) != END && scan < last) {
+ 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);
/* The principal pseudo-switch. Cannot be a switch, since we
look into several different things. */
if (OP(scan) == BRANCH || OP(scan) == BRANCHJ
- || OP(scan) == IFTHEN || OP(scan) == SUSPEND) {
+ || OP(scan) == IFTHEN) {
next = regnext(scan);
code = OP(scan);
/* demq: the op(next)==code check is to see if we have "branch-branch" AFAICT */
- if (OP(next) == code || code == IFTHEN || code == SUSPEND) {
+ if (OP(next) == code || code == IFTHEN) {
/* NOTE - There is similar code to this block below for handling
TRIE nodes on a re-study. If you change stuff here check there
too. */
struct regnode_charclass_class accum;
regnode * const startbranch=scan;
- if (flags & SCF_DO_SUBSTR) /* XXXX Add !SUSPEND? */
- scan_commit(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
+ if (flags & SCF_DO_SUBSTR)
+ SCAN_COMMIT(pRExC_state, data, minlenp); /* Cannot merge strings after this. */
if (flags & SCF_DO_STCLASS)
cl_init_zero(pRExC_state, &accum);
num++;
data_fake.flags = 0;
- if (data) {
+ if (data) {
data_fake.whilem_c = data->whilem_c;
data_fake.last_closep = data->last_closep;
}
else
data_fake.last_closep = &fake;
+
+ data_fake.pos_delta = delta;
next = regnext(scan);
scan = NEXTOPER(scan);
if (code != BRANCH)
cl_init(pRExC_state, &this_class);
data_fake.start_class = &this_class;
f = SCF_DO_STCLASS_AND;
- }
+ }
if (flags & SCF_WHILEM_VISITED_POS)
f |= SCF_WHILEM_VISITED_POS;
/* we suppose the run is continuous, last=next...*/
minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
- next, &data_fake, f,depth+1);
+ next, &data_fake,
+ stopparen, recursed, NULL, f,depth+1);
if (min1 > minnext)
min1 = minnext;
if (max1 < minnext + deltanext)
scan = next;
if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
pars++;
+ if (data_fake.flags & SCF_SEEN_ACCEPT) {
+ if ( stopmin > minnext)
+ stopmin = min + min1;
+ flags &= ~SCF_DO_SUBSTR;
+ if (data)
+ data->flags |= SCF_SEEN_ACCEPT;
+ }
if (data) {
if (data_fake.flags & SF_HAS_EVAL)
data->flags |= SF_HAS_EVAL;
}
if (flags & SCF_DO_STCLASS)
cl_or(pRExC_state, &accum, &this_class);
- if (code == SUSPEND)
- break;
}
if (code == IFTHEN && num < 2) /* Empty ELSE branch */
min1 = 0;
if (flags & SCF_DO_STCLASS_OR) {
cl_or(pRExC_state, data->start_class, &accum);
if (min1) {
- cl_and(data->start_class, &and_with);
+ cl_and(data->start_class, and_withp);
flags &= ~SCF_DO_STCLASS;
}
}
else {
/* Switch to OR mode: cache the old value of
* data->start_class */
- StructCopy(data->start_class, &and_with,
+ INIT_AND_WITHP;
+ StructCopy(data->start_class, and_withp,
struct regnode_charclass_class);
flags &= ~SCF_DO_STCLASS_AND;
StructCopy(&accum, data->start_class,
/* dont use tail as the end marker for this traverse */
for ( cur = startbranch ; cur != scan ; cur = regnext( cur ) ) {
regnode * const noper = NEXTOPER( cur );
+#if defined(DEBUGGING) || defined(NOJUMPTRIE)
regnode * const noper_next = regnext( noper );
+#endif
DEBUG_OPTIMISE_r({
regprop(RExC_rx, mysv, cur);
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 &&
startbranch == first)
|| ( first_non_open == first )) &&
- depth==0 )
+ depth==0 ) {
flags |= SCF_TRIE_RESTUDY;
+ if ( startbranch == first
+ && scan == tail )
+ {
+ RExC_seen &=~REG_TOP_LEVEL_BRANCHES;
+ }
+ }
#endif
}
}
} else /* single branch is optimized. */
scan = NEXTOPER(scan);
continue;
+ } else if (OP(scan) == SUSPEND || OP(scan) == GOSUB || OP(scan) == GOSTART) {
+ scan_frame *newframe = NULL;
+ I32 paren;
+ regnode *start;
+ regnode *end;
+
+ if (OP(scan) != SUSPEND) {
+ /* set the pointer */
+ if (OP(scan) == GOSUB) {
+ paren = ARG(scan);
+ RExC_recurse[ARG2L(scan)] = scan;
+ start = RExC_open_parens[paren-1];
+ end = RExC_close_parens[paren-1];
+ } else {
+ paren = 0;
+ start = RExC_rxi->program + 1;
+ end = RExC_opend;
+ }
+ if (!recursed) {
+ Newxz(recursed, (((RExC_npar)>>3) +1), U8);
+ SAVEFREEPV(recursed);
+ }
+ if (!PAREN_TEST(recursed,paren+1)) {
+ PAREN_SET(recursed,paren+1);
+ Newx(newframe,1,scan_frame);
+ } else {
+ if (flags & SCF_DO_SUBSTR) {
+ SCAN_COMMIT(pRExC_state,data,minlenp);
+ data->longest = &(data->longest_float);
+ }
+ is_inf = is_inf_internal = 1;
+ if (flags & SCF_DO_STCLASS_OR) /* Allow everything */
+ cl_anything(pRExC_state, data->start_class);
+ flags &= ~SCF_DO_STCLASS;
+ }
+ } else {
+ Newx(newframe,1,scan_frame);
+ paren = stopparen;
+ start = scan+2;
+ end = regnext(scan);
+ }
+ if (newframe) {
+ assert(start);
+ assert(end);
+ SAVEFREEPV(newframe);
+ newframe->next = regnext(scan);
+ newframe->last = last;
+ newframe->stop = stopparen;
+ newframe->prev = frame;
+
+ frame = newframe;
+ scan = start;
+ stopparen = paren;
+ last = end;
+
+ continue;
+ }
}
else if (OP(scan) == EXACT) {
I32 l = STR_LEN(scan);
else
data->start_class->flags |= ANYOF_UNICODE_ALL;
data->start_class->flags &= ~ANYOF_EOS;
- cl_and(data->start_class, &and_with);
+ cl_and(data->start_class, and_withp);
}
flags &= ~SCF_DO_STCLASS;
}
/* 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);
ANYOF_BITMAP_SET(data->start_class, uc);
data->start_class->flags &= ~ANYOF_EOS;
}
- cl_and(data->start_class, &and_with);
+ cl_and(data->start_class, and_withp);
}
flags &= ~SCF_DO_STCLASS;
}
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;
case CURLY:
- mincount = ARG1(scan);
- maxcount = ARG2(scan);
+ if (stopparen>0 && (OP(scan)==CURLYN || OP(scan)==CURLYM)
+ && (scan->flags == stopparen))
+ {
+ mincount = 1;
+ maxcount = 1;
+ } else {
+ mincount = ARG1(scan);
+ maxcount = ARG2(scan);
+ }
next = regnext(scan);
if (OP(scan) == CURLYX) {
I32 lp = (data ? *(data->last_closep) : 0);
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) {
f &= ~SCF_WHILEM_VISITED_POS;
/* This will finish on WHILEM, setting scan, or on NULL: */
- minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext, last, data,
+ minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
+ last, data, stopparen, recursed, NULL,
(mincount == 0
? (f & ~SCF_DO_SUBSTR) : f),depth+1);
else if (flags & SCF_DO_STCLASS_AND) {
/* Switch to OR mode: cache the old value of
* data->start_class */
- StructCopy(data->start_class, &and_with,
+ INIT_AND_WITHP;
+ StructCopy(data->start_class, and_withp,
struct regnode_charclass_class);
flags &= ~SCF_DO_STCLASS_AND;
StructCopy(&this_class, data->start_class,
} else { /* Non-zero len */
if (flags & SCF_DO_STCLASS_OR) {
cl_or(pRExC_state, data->start_class, &this_class);
- cl_and(data->start_class, &and_with);
+ cl_and(data->start_class, and_withp);
}
else if (flags & SCF_DO_STCLASS_AND)
cl_and(data->start_class, &this_class);
nxt = regnext(nxt);
if (OP(nxt) != CLOSE)
goto nogo;
+ if (RExC_open_parens) {
+ RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
+ RExC_close_parens[ARG(nxt1)-1]=nxt+2; /*close->while*/
+ }
/* Now we know that nxt2 is the only contents: */
oscan->flags = (U8)ARG(nxt);
OP(oscan) = CURLYN;
OP(nxt1) = NOTHING; /* was OPEN. */
+
#ifdef DEBUGGING
OP(nxt1 + 1) = OPTIMIZED; /* was count. */
NEXT_OFF(nxt1+ 1) = 0; /* just for consistancy. */
if (OP(nxt) != CLOSE)
FAIL("Panic opt close");
oscan->flags = (U8)ARG(nxt);
+ if (RExC_open_parens) {
+ RExC_open_parens[ARG(nxt1)-1]=oscan; /*open->CURLYM*/
+ RExC_close_parens[ARG(nxt1)-1]=nxt2+1; /*close->NOTHING*/
+ }
OP(nxt1) = OPTIMIZED; /* was OPEN. */
OP(nxt) = OPTIMIZED; /* was CLOSE. */
+
#ifdef DEBUGGING
OP(nxt1 + 1) = OPTIMIZED; /* was count. */
OP(nxt + 1) = OPTIMIZED; /* was count. */
#endif
/* Optimize again: */
study_chunk(pRExC_state, &nxt1, minlenp, &deltanext, nxt,
- NULL, 0,depth+1);
+ NULL, stopparen, recursed, NULL, 0,depth+1);
}
else
oscan->flags = 0;
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_with);
+ cl_and(data->start_class, and_withp);
flags &= ~SCF_DO_STCLASS;
}
}
}
else
data_fake.last_closep = &fake;
+ data_fake.pos_delta = delta;
if ( flags & SCF_DO_STCLASS && !scan->flags
&& OP(scan) == IFMATCH ) { /* Lookahead */
cl_init(pRExC_state, &intrnl);
f |= SCF_WHILEM_VISITED_POS;
next = regnext(scan);
nscan = NEXTOPER(NEXTOPER(scan));
- minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext, last, &data_fake, f,depth+1);
+ minnext = study_chunk(pRExC_state, &nscan, minlenp, &deltanext,
+ last, &data_fake, stopparen, recursed, NULL, f, depth+1);
if (scan->flags) {
if (deltanext) {
- vFAIL("Variable length lookbehind not implemented");
+ FAIL("Variable length lookbehind not implemented");
}
else if (minnext > (I32)U8_MAX) {
- vFAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
+ FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
}
scan->flags = (U8)minnext;
}
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);
}
}
else
data_fake.last_closep = &fake;
data_fake.flags = 0;
+ data_fake.pos_delta = delta;
if (is_inf)
data_fake.flags |= SF_IS_INF;
if ( flags & SCF_DO_STCLASS && !scan->flags
f |= SCF_WHILEM_VISITED_POS;
next = regnext(scan);
nscan = NEXTOPER(NEXTOPER(scan));
-
- *minnextp = study_chunk(pRExC_state, &nscan, minnextp, &deltanext, last, &data_fake, f,depth+1);
+
+ *minnextp = study_chunk(pRExC_state, &nscan, minnextp, &deltanext,
+ last, &data_fake, stopparen, recursed, NULL, f,depth+1);
if (scan->flags) {
if (deltanext) {
- vFAIL("Variable length lookbehind not implemented");
+ FAIL("Variable length lookbehind not implemented");
}
else if (*minnextp > (I32)U8_MAX) {
- vFAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
+ FAIL2("Lookbehind longer than %"UVuf" not implemented", (UV)U8_MAX);
}
scan->flags = (U8)*minnextp;
}
-
+
*minnextp += min;
-
-
+
if (f & SCF_DO_STCLASS_AND) {
const int was = (data->start_class->flags & ANYOF_EOS);
cl_and(data->start_class, &intrnl);
if (was)
data->start_class->flags |= ANYOF_EOS;
- }
+ }
if (data) {
if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
pars++;
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 )
#endif
}
else if (OP(scan) == OPEN) {
- pars++;
+ if (stopparen != (I32)ARG(scan))
+ pars++;
}
else if (OP(scan) == CLOSE) {
+ if (stopparen == (I32)ARG(scan)) {
+ break;
+ }
if ((I32)ARG(scan) == is_par) {
next = regnext(scan);
if (data)
data->flags |= SF_HAS_EVAL;
}
- else if (OP(scan) == LOGICAL && scan->flags == 2) { /* Embedded follows */
+ else if ( PL_regkind[OP(scan)] == ENDLIKE ) {
+ if (flags & SCF_DO_SUBSTR) {
+ SCAN_COMMIT(pRExC_state,data,minlenp);
+ flags &= ~SCF_DO_SUBSTR;
+ }
+ if (data && OP(scan)==ACCEPT) {
+ data->flags |= SCF_SEEN_ACCEPT;
+ if (stopmin > min)
+ stopmin = min;
+ }
+ }
+ 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;
cl_anything(pRExC_state, data->start_class);
flags &= ~SCF_DO_STCLASS;
}
+ else if (OP(scan) == GPOS) {
+ if (!(RExC_rx->extflags & RXf_GPOS_FLOAT) &&
+ !(delta || is_inf || (data && data->pos_delta)))
+ {
+ if (!(RExC_rx->extflags & RXf_ANCH) && (flags & SCF_DO_SUBSTR))
+ RExC_rx->extflags |= RXf_ANCH_GPOS;
+ if (RExC_rx->gofs < (U32)min)
+ RExC_rx->gofs = min;
+ } else {
+ RExC_rx->extflags |= RXf_GPOS_FLOAT;
+ RExC_rx->gofs = 0;
+ }
+ }
#ifdef TRIE_STUDY_OPT
-#ifdef FULL_TRIE_STUDY
+#ifdef FULL_TRIE_STUDY
else if (PL_regkind[OP(scan)] == TRIE) {
/* NOTE - There is similar code to this block above for handling
- BRANCH nodes on the initial study. If you change stuff here
+ BRANCH nodes on the initial study. If you change stuff here
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
data_fake.last_closep = &fake;
-
+ data_fake.pos_delta = delta;
if (flags & SCF_DO_STCLASS) {
cl_init(pRExC_state, &this_class);
data_fake.start_class = &this_class;
if (trie->jump[word]) {
if (!nextbranch)
- nextbranch = tail - trie->jump[0];
- scan= tail - trie->jump[word];
+ nextbranch = trie_node + trie->jump[0];
+ scan= trie_node + trie->jump[word];
/* We go from the jump point to the branch that follows
it. Note this means we need the vestigal unused branches
even though they arent otherwise used.
*/
- minnext = study_chunk(pRExC_state, &scan, minlenp, &deltanext,
- (regnode *)nextbranch, &data_fake, f,depth+1);
+ minnext = study_chunk(pRExC_state, &scan, minlenp,
+ &deltanext, (regnode *)nextbranch, &data_fake,
+ stopparen, recursed, NULL, f,depth+1);
}
if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
nextbranch= regnext((regnode*)nextbranch);
if (data_fake.flags & (SF_HAS_PAR|SF_IN_PAR))
pars++;
-
+ if (data_fake.flags & SCF_SEEN_ACCEPT) {
+ if ( stopmin > min + min1)
+ stopmin = min + min1;
+ flags &= ~SCF_DO_SUBSTR;
+ if (data)
+ data->flags |= SCF_SEEN_ACCEPT;
+ }
if (data) {
if (data_fake.flags & SF_HAS_EVAL)
data->flags |= SF_HAS_EVAL;
if (flags & SCF_DO_STCLASS_OR) {
cl_or(pRExC_state, data->start_class, &accum);
if (min1) {
- cl_and(data->start_class, &and_with);
+ cl_and(data->start_class, and_withp);
flags &= ~SCF_DO_STCLASS;
}
}
else {
/* Switch to OR mode: cache the old value of
* data->start_class */
- StructCopy(data->start_class, &and_with,
+ INIT_AND_WITHP;
+ StructCopy(data->start_class, and_withp,
struct regnode_charclass_class);
flags &= ~SCF_DO_STCLASS_AND;
StructCopy(&accum, data->start_class,
}
#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);
}
+ if (frame) {
+ last = frame->last;
+ scan = frame->next;
+ stopparen = frame->stop;
+ frame = frame->prev;
+ goto fake_study_recurse;
+ }
finish:
+ assert(!frame);
+ DEBUG_STUDYDATA("pre-fin:",data,depth);
+
*scanp = scan;
*deltap = is_inf_internal ? I32_MAX : delta;
if (flags & SCF_DO_SUBSTR && is_inf)
data->flags &= ~SF_IN_PAR;
}
if (flags & SCF_DO_STCLASS_OR)
- cl_and(data->start_class, &and_with);
+ cl_and(data->start_class, and_withp);
if (flags & SCF_TRIE_RESTUDY)
data->flags |= SCF_TRIE_RESTUDY;
- DEBUG_STUDYDATA(data,depth);
+ DEBUG_STUDYDATA("post-fin:",data,depth);
- return min;
+ 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) {
- Renewc(RExC_rx->data,
- sizeof(*RExC_rx->data) + sizeof(void*) * (RExC_rx->data->count + n - 1),
- char, struct reg_data);
- Renew(RExC_rx->data->what, RExC_rx->data->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)
#else
#define CHECK_RESTUDY_GOTO
#endif
+
/*
- pregcomp - compile a regular expression into internal code
*
* Beware that the optimization-preparation code in here knows about some
* of the structure of the compiled regexp. [I'll say.]
*/
-regexp *
-Perl_pregcomp(pTHX_ char *exp, char *xend, PMOP *pm)
+
+
+
+#ifndef PERL_IN_XSUB_RE
+#define RE_ENGINE_PTR &PL_core_reg_engine
+#else
+extern const struct regexp_engine my_reg_engine;
+#define RE_ENGINE_PTR &my_reg_engine
+#endif
+
+#ifndef PERL_IN_XSUB_RE
+REGEXP *
+Perl_pregcomp(pTHX_ const SV * const pattern, const U32 flags)
+{
+ dVAR;
+ HV * const table = GvHV(PL_hintgv);
+ /* Dispatch a request to compile a regexp to correct
+ regexp engine. */
+ 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, pattern, flags);
+ }
+ }
+ return Perl_re_compile(aTHX_ pattern, flags);
+}
+#endif
+
+REGEXP *
+Perl_re_compile(pTHX_ const SV * const pattern, U32 pm_flags)
{
dVAR;
- register regexp *r;
+ 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
-
GET_RE_DEBUG_FLAGS_DECL;
+ DEBUG_r(if (!PL_colorset) reginitcolors());
- if (exp == NULL)
- FAIL("NULL regexp argument");
-
- RExC_utf8 = pm->op_pmdynflags & PMdf_CMP_UTF8;
+ RExC_utf8 = RExC_orig_utf8 = SvUTF8(pattern);
- RExC_precomp = exp;
- DEBUG_r(if (!PL_colorset) reginitcolors());
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_nestroot = 0;
RExC_size = 0L;
RExC_emit = &PL_regdummy;
RExC_whilem_seen = 0;
+ RExC_charnames = NULL;
+ RExC_open_parens = NULL;
+ 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;
+
#if 0 /* REGC() is (currently) a NOP at the first pass.
* Clever compilers notice this and complain. --jhi */
REGC((U8)REG_MAGIC, (char*)RExC_emit);
RExC_precomp = NULL;
return(NULL);
}
- DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "Required "));
- DEBUG_COMPILE_r(PerlIO_printf(Perl_debug_log, "size %"IVdf" nodes ", (IV)RExC_size));
- DEBUG_PARSE_r(PerlIO_printf(Perl_debug_log, "\nStarting second pass (creation)\n"));
+ 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"
+ "Starting second pass (creation)\n",
+ (IV)RExC_size);
RExC_lastnum=0;
RExC_lastparse=NULL;
});
-
-
/* Small enough for pointer-storage convention?
If extralen==0, this means that we will not need long jumps. */
if (RExC_size >= 0x10000L && RExC_extralen)
if (RExC_whilem_seen > 15)
RExC_whilem_seen = 15;
- /* Allocate space and initialize. */
- Newxc(r, sizeof(regexp) + (unsigned)RExC_size * sizeof(regnode),
- char, regexp);
- if (r == NULL)
+ /* Allocate space and zero-initialize. Note, the two step process
+ of zeroing when in debug mode, thus anything assigned has to
+ happen after that */
+ 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);
-#endif
- r->refcnt = 1;
- r->prelen = xend - exp;
- r->precomp = savepvn(RExC_precomp, r->prelen);
- r->subbeg = NULL;
-#ifdef PERL_OLD_COPY_ON_WRITE
- r->saved_copy = NULL;
+ Zero(ri, sizeof(regexp_internal) + (unsigned)RExC_size * sizeof(regnode), char);
+#else
+ /* bulk initialize base fields with 0. */
+ Zero(ri, sizeof(regexp_internal), char);
#endif
- r->reganch = pm->op_pmflags & PMf_COMPILETIME;
- r->nparens = RExC_npar - 1; /* set early to validate backrefs */
- r->lastparen = 0; /* mg.c reads this. */
- r->substrs = 0; /* Useful during FAIL. */
- r->startp = 0; /* Useful during FAIL. */
- r->endp = 0; /* Useful during FAIL. */
+ /* non-zero initialization begins here */
+ RXi_SET( r, ri );
+ r->engine= RE_ENGINE_PTR;
+ 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;
+ }
- Newxz(r->offsets, 2*RExC_size+1, U32); /* MJD 20001228 */
- if (r->offsets) {
- r->offsets[0] = RExC_size;
+ r->intflags = 0;
+ r->nparens = RExC_npar - 1; /* set early to validate backrefs */
+
+ if (RExC_seen & REG_SEEN_RECURSE) {
+ Newxz(RExC_open_parens, RExC_npar,regnode *);
+ SAVEFREEPV(RExC_open_parens);
+ Newxz(RExC_close_parens,RExC_npar,regnode *);
+ SAVEFREEPV(RExC_close_parens);
}
+
+ /* Useful during FAIL. */
+#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_emit_start = r->program;
- RExC_emit = r->program;
+ 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);
+ if (RExC_recurse_count) {
+ Newxz(RExC_recurse,RExC_recurse_count,regnode *);
+ SAVEFREEPV(RExC_recurse);
+ }
reStudy:
r->minlen = minlen = sawplus = sawopen = 0;
Zero(r->substrs, 1, struct reg_substr_data);
- StructCopy(&zero_scan_data, &data, scan_data_t);
#ifdef TRIE_STUDY_OPT
if ( restudied ) {
+ U32 seen=RExC_seen;
DEBUG_OPTIMISE_r(PerlIO_printf(Perl_debug_log,"Restudying\n"));
- RExC_state=copyRExC_state;
+
+ RExC_state = copyRExC_state;
+ if (seen & REG_TOP_LEVEL_BRANCHES)
+ RExC_seen |= REG_TOP_LEVEL_BRANCHES;
+ else
+ RExC_seen &= ~REG_TOP_LEVEL_BRANCHES;
if (data.last_found) {
SvREFCNT_dec(data.longest_fixed);
SvREFCNT_dec(data.longest_float);
SvREFCNT_dec(data.last_found);
}
+ StructCopy(&zero_scan_data, &data, scan_data_t);
} else {
- copyRExC_state=RExC_state;
+ StructCopy(&zero_scan_data, &data, scan_data_t);
+ copyRExC_state = RExC_state;
}
+#else
+ StructCopy(&zero_scan_data, &data, scan_data_t);
#endif
+
/* Dig out information for optimizations. */
- r->reganch = pm->op_pmflags & 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->reganch |= ROPT_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->reganch |= ROPT_NAUGHTY;
- scan = r->program + 1; /* First BRANCH. */
+ r->intflags |= PREGf_NAUGHTY;
+ 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 */
- if (OP(scan) != BRANCH) { /* Only one top-level choice. */
+ if (!(RExC_seen & REG_TOP_LEVEL_BRANCHES)) { /* Only one top-level choice. */
I32 fake;
STRLEN longest_float_length, longest_fixed_length;
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->reganch |= (OP(first) == MBOL
- ? ROPT_ANCH_MBOL
+ r->extflags |= (OP(first) == MBOL
+ ? RXf_ANCH_MBOL
: (OP(first) == SBOL
- ? ROPT_ANCH_SBOL
- : ROPT_ANCH_BOL));
+ ? RXf_ANCH_SBOL
+ : RXf_ANCH_BOL));
first = NEXTOPER(first);
goto again;
}
else if (OP(first) == GPOS) {
- r->reganch |= ROPT_ANCH_GPOS;
+ r->extflags |= RXf_ANCH_GPOS;
first = NEXTOPER(first);
goto again;
}
- else if (!sawopen && (OP(first) == STAR &&
+ else if ((!sawopen || !RExC_sawback) &&
+ (OP(first) == STAR &&
PL_regkind[OP(NEXTOPER(first))] == REG_ANY) &&
- !(r->reganch & ROPT_ANCH) )
+ !(r->extflags & RXf_ANCH) && !(RExC_seen & REG_SEEN_EVAL))
{
/* turn .* into ^.* with an implied $*=1 */
const int type =
(OP(NEXTOPER(first)) == REG_ANY)
- ? ROPT_ANCH_MBOL
- : ROPT_ANCH_SBOL;
- r->reganch |= type | ROPT_IMPLICIT;
+ ? RXf_ANCH_MBOL
+ : RXf_ANCH_SBOL;
+ r->extflags |= type;
+ r->intflags |= PREGf_IMPLICIT;
first = NEXTOPER(first);
goto again;
}
if (sawplus && (!sawopen || !RExC_sawback)
&& !(RExC_seen & REG_SEEN_EVAL)) /* May examine pos and $& */
/* x+ must match at the 1st pos of run of x's */
- r->reganch |= ROPT_SKIP;
+ r->intflags |= PREGf_SKIP;
/* Scan is after the zeroth branch, first is atomic matcher. */
#ifdef TRIE_STUDY_OPT
- DEBUG_COMPILE_r(
+ DEBUG_PARSE_r(
if (!restudied)
PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
(IV)(first - scan + 1))
);
#else
- DEBUG_COMPILE_r(
+ DEBUG_PARSE_r(
PerlIO_printf(Perl_debug_log, "first at %"IVdf"\n",
(IV)(first - scan + 1))
);
* it happens that c_offset_min has been invalidated, since the
* earlier string may buy us something the later one won't.]
*/
- minlen = 0;
-
+
data.longest_fixed = newSVpvs("");
data.longest_float = newSVpvs("");
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;
} else /* XXXX Check for BOUND? */
stclass_flag = 0;
data.last_closep = &last_close;
-
+
minlen = study_chunk(pRExC_state, &first, &minlen, &fake, scan + RExC_size, /* Up to end */
- &data, SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag,0);
+ &data, -1, NULL, NULL,
+ SCF_DO_SUBSTR | SCF_WHILEM_VISITED_POS | stclass_flag,0);
CHECK_RESTUDY_GOTO;
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_GPOS) || (r->reganch & ROPT_ANCH_GPOS)))
- r->reganch |= ROPT_CHECK_ALL;
- scan_commit(pRExC_state, &data,&minlen);
+ && !(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,0);
SvREFCNT_dec(data.last_found);
/* Note that code very similar to this but for anchored string
if (longest_float_length
|| (data.flags & SF_FL_BEFORE_EOL
&& (!(data.flags & SF_FL_BEFORE_MEOL)
- || (RExC_flags & PMf_MULTILINE))))
+ || (RExC_flags & RXf_PMf_MULTILINE))))
{
I32 t,ml;
t = (data.flags & SF_FL_BEFORE_EOL /* Can't have SEOL and MULTI */
&& (!(data.flags & SF_FL_BEFORE_MEOL)
- || (RExC_flags & PMf_MULTILINE)));
+ || (RExC_flags & RXf_PMf_MULTILINE)));
fbm_compile(data.longest_float, t ? FBMcf_TAIL : 0);
}
else {
if (longest_fixed_length
|| (data.flags & SF_FIX_BEFORE_EOL /* Cannot have SEOL and MULTI */
&& (!(data.flags & SF_FIX_BEFORE_MEOL)
- || (RExC_flags & PMf_MULTILINE))))
+ || (RExC_flags & RXf_PMf_MULTILINE))))
{
I32 t,ml;
t = (data.flags & SF_FIX_BEFORE_EOL /* Can't have SEOL and MULTI */
&& (!(data.flags & SF_FIX_BEFORE_MEOL)
- || (RExC_flags & PMf_MULTILINE)));
+ || (RExC_flags & RXf_PMf_MULTILINE)));
fbm_compile(data.longest_fixed, t ? FBMcf_TAIL : 0);
}
else {
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];
- r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
+ 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);
PerlIO_printf(Perl_debug_log,
r->check_substr = r->anchored_substr;
r->check_utf8 = r->anchored_utf8;
r->check_offset_min = r->check_offset_max = r->anchored_offset;
- if (r->reganch & ROPT_ANCH_SINGLE)
- r->reganch |= ROPT_NOSCAN;
+ if (r->extflags & RXf_ANCH_SINGLE)
+ r->extflags |= RXf_NOSCAN;
}
else {
r->check_end_shift = r->float_end_shift;
}
/* XXXX Currently intuiting is not compatible with ANCH_GPOS.
This should be changed ASAP! */
- if ((r->check_substr || r->check_utf8) && !(r->reganch & ROPT_ANCH_GPOS)) {
- r->reganch |= RE_USE_INTUIT;
+ if ((r->check_substr || r->check_utf8) && !(r->extflags & RXf_ANCH_GPOS)) {
+ r->extflags |= RXf_USE_INTUIT;
if (SvTAIL(r->check_substr ? r->check_substr : r->check_utf8))
- r->reganch |= RE_INTUIT_TAIL;
+ r->extflags |= RXf_INTUIT_TAIL;
}
/* XXX Unneeded? dmq (shouldn't as this is handled elsewhere)
if ( (STRLEN)minlen < longest_float_length )
struct regnode_charclass_class ch_class;
I32 last_close = 0;
- DEBUG_COMPILE_r(PerlIO_printf(Perl_debug_log, "\n"));
+ 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;
+
minlen = study_chunk(pRExC_state, &scan, &minlen, &fake, scan + RExC_size,
- &data, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS,0);
-
+ &data, -1, NULL, NULL, SCF_DO_STCLASS_AND|SCF_WHILEM_VISITED_POS,0);
+
CHECK_RESTUDY_GOTO;
r->check_substr = r->check_utf8 = r->anchored_substr = r->anchored_utf8
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];
- r->reganch &= ~ROPT_SKIP; /* Used in find_byclass(). */
+ 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);
PerlIO_printf(Perl_debug_log,
/* Guard against an embedded (?=) or (?<=) with a longer minlen than
the "real" pattern. */
+ DEBUG_OPTIMISE_r({
+ PerlIO_printf(Perl_debug_log,"minlen: %"IVdf" r->minlen:%"IVdf"\n",
+ (IV)minlen, (IV)r->minlen);
+ });
+ r->minlenret = minlen;
if (r->minlen < minlen)
r->minlen = minlen;
if (RExC_seen & REG_SEEN_GPOS)
- r->reganch |= ROPT_GPOS_SEEN;
+ r->extflags |= RXf_GPOS_SEEN;
if (RExC_seen & REG_SEEN_LOOKBEHIND)
- r->reganch |= ROPT_LOOKBEHIND_SEEN;
+ r->extflags |= RXf_LOOKBEHIND_SEEN;
if (RExC_seen & REG_SEEN_EVAL)
- r->reganch |= ROPT_EVAL_SEEN;
+ r->extflags |= RXf_EVAL_SEEN;
if (RExC_seen & REG_SEEN_CANY)
- r->reganch |= ROPT_CANY_SEEN;
- Newxz(r->startp, RExC_npar, I32);
- Newxz(r->endp, RExC_npar, I32);
+ r->extflags |= RXf_CANY_SEEN;
+ if (RExC_seen & REG_SEEN_VERBARG)
+ r->intflags |= PREGf_VERBARG_SEEN;
+ if (RExC_seen & REG_SEEN_CUTGROUP)
+ r->intflags |= PREGf_CUTGROUP_SEEN;
+ if (RExC_paren_names)
+ RXp_PAREN_NAMES(r) = (HV*)SvREFCNT_inc(RExC_paren_names);
+ else
+ 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->offs, RExC_npar, regexp_paren_pair);
+ /* assume we don't need to swap parens around before we match */
- DEBUG_r( RX_DEBUG_on(r) );
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);
+#endif
+ return rx;
}
+#undef RE_ENGINE_PTR
-#define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \
- int rem=(int)(RExC_end - RExC_parse); \
- int cut; \
- int num; \
- int iscut=0; \
- if (rem>10) { \
- rem=10; \
- iscut=1; \
- } \
- cut=10-rem; \
- if (RExC_lastparse!=RExC_parse) \
- PerlIO_printf(Perl_debug_log," >%.*s%-*s", \
- rem, RExC_parse, \
- cut + 4, \
- iscut ? "..." : "<" \
- ); \
- else \
- PerlIO_printf(Perl_debug_log,"%16s",""); \
- \
- if (SIZE_ONLY) \
- num=RExC_size; \
+
+SV*
+Perl_reg_named_buff(pTHX_ REGEXP * const rx, SV * const key, SV * const value,
+ const U32 flags)
+{
+ 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);
+ }
+ }
+ 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 {
+ return FALSE;
+ }
+}
+
+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.
+ * If flags is REG_RSN_RETURN_NAME returns an SV* containing the name
+ * If flags is REG_RSN_RETURN_DATA returns the data SV* corresponding
+ * to the parsed name as looked up in the RExC_paren_names hash.
+ * If there is an error throws a vFAIL().. type exception.
+ */
+
+#define REG_RSN_RETURN_NULL 0
+#define REG_RSN_RETURN_NAME 1
+#define REG_RSN_RETURN_DATA 2
+
+STATIC SV*
+S_reg_scan_name(pTHX_ RExC_state_t *pRExC_state, U32 flags) {
+ char *name_start = 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
+ = 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) {
+ HE *he_str = NULL;
+ SV *sv_dat = NULL;
+ if ( ! sv_name ) /* should not happen*/
+ Perl_croak(aTHX_ "panic: no svname in reg_scan_name");
+ if (RExC_paren_names)
+ he_str = hv_fetch_ent( RExC_paren_names, sv_name, 0, 0 );
+ if ( he_str )
+ sv_dat = HeVAL(he_str);
+ if ( ! sv_dat )
+ vFAIL("Reference to nonexistent named group");
+ return sv_dat;
+ }
+ else {
+ Perl_croak(aTHX_ "panic: bad flag in reg_scan_name");
+ }
+ /* NOT REACHED */
+ }
+ return NULL;
+}
+
+#define DEBUG_PARSE_MSG(funcname) DEBUG_PARSE_r({ \
+ int rem=(int)(RExC_end - RExC_parse); \
+ int cut; \
+ int num; \
+ int iscut=0; \
+ if (rem>10) { \
+ rem=10; \
+ iscut=1; \
+ } \
+ cut=10-rem; \
+ if (RExC_lastparse!=RExC_parse) \
+ PerlIO_printf(Perl_debug_log," >%.*s%-*s", \
+ rem, RExC_parse, \
+ cut + 4, \
+ iscut ? "..." : "<" \
+ ); \
+ else \
+ PerlIO_printf(Perl_debug_log,"%16s",""); \
+ \
+ if (SIZE_ONLY) \
+ num = RExC_size + 1; \
else \
num=REG_NODE_NUM(RExC_emit); \
if (RExC_lastnum!=num) \
- PerlIO_printf(Perl_debug_log,"|%4d",num); \
+ PerlIO_printf(Perl_debug_log,"|%4d",num); \
else \
- PerlIO_printf(Perl_debug_log,"|%4s",""); \
+ PerlIO_printf(Perl_debug_log,"|%4s",""); \
PerlIO_printf(Perl_debug_log,"|%*s%-4s", \
(int)((depth*2)), "", \
(funcname) \
DEBUG_PARSE_MSG((funcname)); \
PerlIO_printf(Perl_debug_log,"%4s","\n"); \
})
+#define DEBUG_PARSE_FMT(funcname,fmt,args) DEBUG_PARSE_r({ \
+ DEBUG_PARSE_MSG((funcname)); \
+ PerlIO_printf(Perl_debug_log,fmt "\n",args); \
+})
/*
- reg - regular expression, i.e. main body or parenthesized thing
*
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. */
/* Make an OPEN node, if parenthesized. */
if (paren) {
+ if ( *RExC_parse == '*') { /* (*VERB:ARG) */
+ char *start_verb = RExC_parse;
+ STRLEN verb_len = 0;
+ char *start_arg = NULL;
+ unsigned char op = 0;
+ int argok = 1;
+ int internal_argval = 0; /* internal_argval is only useful if !argok */
+ while ( *RExC_parse && *RExC_parse != ')' ) {
+ if ( *RExC_parse == ':' ) {
+ start_arg = RExC_parse + 1;
+ break;
+ }
+ RExC_parse++;
+ }
+ ++start_verb;
+ verb_len = RExC_parse - start_verb;
+ if ( start_arg ) {
+ RExC_parse++;
+ while ( *RExC_parse && *RExC_parse != ')' )
+ RExC_parse++;
+ if ( *RExC_parse != ')' )
+ vFAIL("Unterminated verb pattern argument");
+ if ( RExC_parse == start_arg )
+ start_arg = NULL;
+ } else {
+ if ( *RExC_parse != ')' )
+ vFAIL("Unterminated verb pattern");
+ }
+
+ switch ( *start_verb ) {
+ case 'A': /* (*ACCEPT) */
+ if ( memEQs(start_verb,verb_len,"ACCEPT") ) {
+ op = ACCEPT;
+ internal_argval = RExC_nestroot;
+ }
+ break;
+ case 'C': /* (*COMMIT) */
+ if ( memEQs(start_verb,verb_len,"COMMIT") )
+ op = COMMIT;
+ break;
+ case 'F': /* (*FAIL) */
+ 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 || memEQs(start_verb,verb_len,"MARK") ) {
+ op = MARKPOINT;
+ argok = -1;
+ }
+ break;
+ case 'P': /* (*PRUNE) */
+ if ( memEQs(start_verb,verb_len,"PRUNE") )
+ op = PRUNE;
+ break;
+ case 'S': /* (*SKIP) */
+ if ( memEQs(start_verb,verb_len,"SKIP") )
+ op = SKIP;
+ break;
+ case 'T': /* (*THEN) */
+ /* [19:06] <TimToady> :: is then */
+ if ( memEQs(start_verb,verb_len,"THEN") ) {
+ op = CUTGROUP;
+ RExC_seen |= REG_SEEN_CUTGROUP;
+ }
+ break;
+ }
+ if ( ! op ) {
+ RExC_parse++;
+ vFAIL3("Unknown verb pattern '%.*s'",
+ verb_len, start_verb);
+ }
+ if ( argok ) {
+ if ( start_arg && internal_argval ) {
+ vFAIL3("Verb pattern '%.*s' may not have an argument",
+ verb_len, start_verb);
+ } else if ( argok < 0 && !start_arg ) {
+ vFAIL3("Verb pattern '%.*s' has a mandatory argument",
+ verb_len, start_verb);
+ } else {
+ ret = reganode(pRExC_state, op, internal_argval);
+ if ( ! internal_argval && ! SIZE_ONLY ) {
+ if (start_arg) {
+ SV *sv = newSVpvn( start_arg, RExC_parse - start_arg);
+ ARG(ret) = add_data( pRExC_state, 1, "S" );
+ RExC_rxi->data->data[ARG(ret)]=(void*)sv;
+ ret->flags = 0;
+ } else {
+ ret->flags = 1;
+ }
+ }
+ }
+ if (!internal_argval)
+ RExC_seen |= REG_SEEN_VERBARG;
+ } else if ( start_arg ) {
+ vFAIL3("Verb pattern '%.*s' may not have an argument",
+ verb_len, start_verb);
+ } else {
+ ret = reg_node(pRExC_state, op);
+ }
+ nextchar(pRExC_state);
+ return ret;
+ } else
if (*RExC_parse == '?') { /* (?...) */
- U32 posflags = 0, negflags = 0;
- U32 *flagsp = &posflags;
bool is_logical = 0;
const char * const seqstart = RExC_parse;
paren = *RExC_parse++;
ret = NULL; /* For look-ahead/behind. */
switch (paren) {
- case '<': /* (?<...) */
- RExC_seen |= REG_SEEN_LOOKBEHIND;
+
+ 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 = ',';
- if (*RExC_parse != '=' && *RExC_parse != '!')
- goto unknown;
+ else if (*RExC_parse != '=')
+ named_capture:
+ { /* (?<...>) */
+ char *name_start;
+ SV *svname;
+ paren= '>';
+ case '\'': /* (?'...') */
+ name_start= RExC_parse;
+ svname = reg_scan_name(pRExC_state,
+ SIZE_ONLY ? /* reverse test from the others */
+ REG_RSN_RETURN_NAME :
+ REG_RSN_RETURN_NULL);
+ 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 (SIZE_ONLY) {
+ HE *he_str;
+ SV *sv_dat = NULL;
+ if (!svname) /* shouldnt happen */
+ Perl_croak(aTHX_
+ "panic: reg_scan_name returned NULL");
+ 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 )
+ sv_dat = HeVAL(he_str);
+ if ( ! sv_dat ) {
+ /* croak baby croak */
+ Perl_croak(aTHX_
+ "panic: paren_name hash element allocation failed");
+ } else if ( SvPOK(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);
+ paren = 1;
+ goto capturing_parens;
+ }
+ RExC_seen |= REG_SEEN_LOOKBEHIND;
RExC_parse++;
case '=': /* (?=...) */
case '!': /* (?!...) */
RExC_seen_zerolen++;
+ if (*RExC_parse == ')') {
+ ret=reg_node(pRExC_state, OPFAIL);
+ 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;
nextchar(pRExC_state);
*flagp = TRYAGAIN;
return NULL;
- case 'p': /* (?p...) */
- if (SIZE_ONLY && ckWARN2(WARN_DEPRECATED, WARN_REGEXP))
- vWARNdep(RExC_parse, "(?p{}) is deprecated - use (??{})");
- /* FALL THROUGH*/
+ case '0' : /* (?0) */
+ case 'R' : /* (?R) */
+ 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;
+ 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);
+ num = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
+ }
+ goto gen_recurse_regop;
+ /* NOT REACHED */
+ case '+':
+ if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
+ RExC_parse++;
+ vFAIL("Illegal pattern");
+ }
+ goto parse_recursion;
+ /* NOT REACHED*/
+ case '-': /* (?-1) */
+ if (!(RExC_parse[0] >= '1' && RExC_parse[0] <= '9')) {
+ RExC_parse--; /* rewind to let it be handled later */
+ goto parse_flags;
+ }
+ /*FALLTHROUGH */
+ case '1': case '2': case '3': case '4': /* (?1) */
+ case '5': case '6': case '7': case '8': case '9':
+ RExC_parse--;
+ parse_recursion:
+ num = atoi(RExC_parse);
+ parse_start = RExC_parse - 1; /* MJD */
+ if (*RExC_parse == '-')
+ RExC_parse++;
+ while (isDIGIT(*RExC_parse))
+ RExC_parse++;
+ if (*RExC_parse!=')')
+ vFAIL("Expecting close bracket");
+
+ gen_recurse_regop:
+ if ( paren == '-' ) {
+ /*
+ Diagram of capture buffer numbering.
+ Top line is the normal capture buffer numbers
+ Botton line is the negative indexing as from
+ the X (the (?-2))
+
+ + 1 2 3 4 5 X 6 7
+ /(a(x)y)(a(b(c(?-2)d)e)f)(g(h))/
+ - 5 4 3 2 1 X x x
+
+ */
+ num = RExC_npar + num;
+ if (num < 1) {
+ RExC_parse++;
+ vFAIL("Reference to nonexistent group");
+ }
+ } else if ( paren == '+' ) {
+ num = RExC_npar + num - 1;
+ }
+
+ ret = reganode(pRExC_state, GOSUB, num);
+ if (!SIZE_ONLY) {
+ if (num > (I32)RExC_rx->nparens) {
+ RExC_parse++;
+ vFAIL("Reference to nonexistent group");
+ }
+ ARG2L_SET( ret, RExC_recurse_count++);
+ RExC_emit++;
+ DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
+ "Recurse #%"UVuf" to %"IVdf"\n", (UV)ARG(ret), (IV)ARG2L(ret)));
+ } else {
+ RExC_size++;
+ }
+ RExC_seen |= REG_SEEN_RECURSE;
+ 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 '?': /* (??...) */
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 */
}
case '(': /* (?(?{...})...) and (?(?=...)...) */
{
+ int is_define= 0;
if (RExC_parse[0] == '?') { /* (?(?...)) */
if (RExC_parse[1] == '=' || RExC_parse[1] == '!'
|| RExC_parse[1] == '<'
goto insert_if;
}
}
+ else if ( RExC_parse[0] == '<' /* (?(<NAME>)...) */
+ || RExC_parse[0] == '\'' ) /* (?('NAME')...) */
+ {
+ char ch = RExC_parse[0] == '<' ? '>' : '\'';
+ 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 != ch)
+ vFAIL2("Sequence (?(%c... not terminated",
+ (ch == '>' ? '<' : ch));
+ RExC_parse++;
+ if (!SIZE_ONLY) {
+ num = add_data( pRExC_state, 1, "S" );
+ 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 if (RExC_parse[0] == 'D' &&
+ RExC_parse[1] == 'E' &&
+ RExC_parse[2] == 'F' &&
+ RExC_parse[3] == 'I' &&
+ RExC_parse[4] == 'N' &&
+ RExC_parse[5] == 'E')
+ {
+ ret = reganode(pRExC_state,DEFINEP,0);
+ RExC_parse +=6 ;
+ is_define = 1;
+ goto insert_if_check_paren;
+ }
+ else if (RExC_parse[0] == 'R') {
+ RExC_parse++;
+ parno = 0;
+ if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
+ parno = atoi(RExC_parse++);
+ while (isDIGIT(*RExC_parse))
+ RExC_parse++;
+ } else if (RExC_parse[0] == '&') {
+ SV *sv_dat;
+ RExC_parse++;
+ sv_dat = reg_scan_name(pRExC_state,
+ SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
+ parno = sv_dat ? *((I32 *)SvPVX(sv_dat)) : 0;
+ }
+ ret = reganode(pRExC_state,INSUBP,parno);
+ goto insert_if_check_paren;
+ }
else if (RExC_parse[0] >= '1' && RExC_parse[0] <= '9' ) {
/* (?(1)...) */
char c;
RExC_parse++;
ret = reganode(pRExC_state, GROUPP, parno);
+ insert_if_check_paren:
if ((c = *nextchar(pRExC_state)) != ')')
vFAIL("Switch condition not recognized");
insert_if:
if (flags&HASWIDTH)
*flagp |= HASWIDTH;
if (c == '|') {
+ if (is_define)
+ vFAIL("(?(DEFINE)....) does not allow branches");
lastbr = reganode(pRExC_state, IFTHEN, 0); /* Fake one for optimizer. */
regbranch(pRExC_state, &flags, 1,depth+1);
REGTAIL(pRExC_state, ret, lastbr);
}
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:
parno = RExC_npar;
RExC_npar++;
+
ret = reganode(pRExC_state, OPEN, parno);
+ if (!SIZE_ONLY ){
+ if (!RExC_nestroot)
+ RExC_nestroot = parno;
+ 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)));
+ RExC_open_parens[parno-1]= ret;
+ }
+ }
Set_Node_Length(ret, 1); /* MJD */
Set_Node_Offset(ret, RExC_parse); /* MJD */
is_open = 1;
}
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);
return(NULL);
if (*RExC_parse == '|') {
if (!SIZE_ONLY && RExC_extralen) {
- reginsert(pRExC_state, BRANCHJ, br);
+ reginsert(pRExC_state, BRANCHJ, br, depth+1);
}
else { /* MJD */
- reginsert(pRExC_state, BRANCH, br);
+ reginsert(pRExC_state, BRANCH, br, depth+1);
Set_Node_Length(br, paren != 0);
Set_Node_Offset_To_R(br-RExC_emit_start, parse_start-RExC_start);
}
}
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 != ':') {
break;
case 1:
ender = reganode(pRExC_state, CLOSE, parno);
+ if (!SIZE_ONLY && RExC_seen & REG_SEEN_RECURSE) {
+ DEBUG_OPTIMISE_MORE_r(PerlIO_printf(Perl_debug_log,
+ "Setting close paren #%"IVdf" to %d\n",
+ (IV)parno, REG_NODE_NUM(ender)));
+ RExC_close_parens[parno-1]= ender;
+ if (RExC_nestroot == parno)
+ RExC_nestroot = 0;
+ }
Set_Node_Offset(ender,RExC_parse+1); /* MJD */
Set_Node_Length(ender,1); /* MJD */
break;
break;
case 0:
ender = reg_node(pRExC_state, END);
+ if (!SIZE_ONLY) {
+ assert(!RExC_opend); /* there can only be one! */
+ RExC_opend = ender;
+ }
break;
}
- REGTAIL_STUDY(pRExC_state, lastbr, ender);
+ REGTAIL(pRExC_state, lastbr, ender);
if (have_branch && !SIZE_ONLY) {
+ if (depth==1)
+ RExC_seen |= REG_TOP_LEVEL_BRANCHES;
+
/* Hook the tails of the branches to the closing node. */
for (br = ret; br; br = regnext(br)) {
const U8 op = PL_regkind[OP(br)];
if (paren == '>')
node = SUSPEND, flag = 0;
- reginsert(pRExC_state, node,ret);
+ reginsert(pRExC_state, node,ret, depth+1);
Set_Node_Cur_Length(ret);
Set_Node_Offset(ret, parse_start + 1);
ret->flags = flag;
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 {
do_curly:
if ((flags&SIMPLE)) {
RExC_naughty += 2 + RExC_naughty / 2;
- reginsert(pRExC_state, CURLY, ret);
+ reginsert(pRExC_state, CURLY, ret, depth+1);
Set_Node_Offset(ret, parse_start+1); /* MJD */
Set_Node_Cur_Length(ret);
}
w->flags = 0;
REGTAIL(pRExC_state, ret, w);
if (!SIZE_ONLY && RExC_extralen) {
- reginsert(pRExC_state, LONGJMP,ret);
- reginsert(pRExC_state, NOTHING,ret);
+ reginsert(pRExC_state, LONGJMP,ret, depth+1);
+ reginsert(pRExC_state, NOTHING,ret, depth+1);
NEXT_OFF(ret) = 3; /* Go over LONGJMP. */
}
- reginsert(pRExC_state, CURLYX,ret);
+ reginsert(pRExC_state, CURLYX,ret, depth+1);
/* MJD hk */
Set_Node_Offset(ret, parse_start+1);
Set_Node_Length(ret,
*flagp = (op != '+') ? (WORST|SPSTART|HASWIDTH) : (WORST|HASWIDTH);
if (op == '*' && (flags&SIMPLE)) {
- reginsert(pRExC_state, STAR, ret);
+ reginsert(pRExC_state, STAR, ret, depth+1);
ret->flags = 0;
RExC_naughty += 4;
}
goto do_curly;
}
else if (op == '+' && (flags&SIMPLE)) {
- reginsert(pRExC_state, PLUS, ret);
+ reginsert(pRExC_state, PLUS, ret, depth+1);
ret->flags = 0;
RExC_naughty += 3;
}
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),
origparse);
}
- if (*RExC_parse == '?') {
+ if (RExC_parse < RExC_end && *RExC_parse == '?') {
nextchar(pRExC_state);
- reginsert(pRExC_state, MINMOD, ret);
+ reginsert(pRExC_state, MINMOD, ret, depth+1);
REGTAIL(pRExC_state, ret, ret + NODE_STEP_REGNODE);
}
- if (ISMULT2(RExC_parse)) {
+#ifndef REG_ALLOW_MINMOD_SUSPEND
+ else
+#endif
+ if (RExC_parse < RExC_end && *RExC_parse == '+') {
+ regnode *ender;
+ nextchar(pRExC_state);
+ ender = reg_node(pRExC_state, SUCCEED);
+ REGTAIL(pRExC_state, ret, ender);
+ reginsert(pRExC_state, SUSPEND, ret, depth+1);
+ ret->flags = 0;
+ ender = reg_node(pRExC_state, TAIL);
+ REGTAIL(pRExC_state, ret, ender);
+ /*ret= ender;*/
+ }
+
+ if (RExC_parse < RExC_end && ISMULT2(RExC_parse)) {
RExC_parse++;
vFAIL("Nested quantifiers");
}
return(ret);
}
+
+/* reg_namedseq(pRExC_state,UVp)
+
+ This is expected to be called by a parser routine that has
+ recognized'\N' and needs to handle the rest. RExC_parse is
+ expected to point at the first char following the N at the time
+ of the call.
+
+ If valuep is non-null then it is assumed that we are parsing inside
+ of a charclass definition and the first codepoint in the resolved
+ string is returned via *valuep and the routine will return NULL.
+ In this mode if a multichar string is returned from the charnames
+ handler a warning will be issued, and only the first char in the
+ sequence will be examined. If the string returned is zero length
+ then the value of *valuep is undefined and NON-NULL will
+ be returned to indicate failure. (This will NOT be a valid pointer
+ to a regnode.)
+
+ If value is null then it is assumed that we are parsing normal text
+ and inserts a new EXACT node into the program containing the resolved
+ string and returns a pointer to the new node. If the string is
+ zerolength a NOTHING node is emitted.
+
+ On success RExC_parse is set to the char following the endbrace.
+ Parsing failures will generate a fatal errorvia vFAIL(...)
+
+ NOTE: We cache all results from the charnames handler locally in
+ the RExC_charnames hash (created on first use) to prevent a charnames
+ handler from playing silly-buggers and returning a short string and
+ then a long string for a given pattern. Since the regexp program
+ size is calculated during an initial parse this would result
+ in a buffer overrun so we cache to prevent the charname result from
+ changing during the course of the parse.
+
+ */
+STATIC regnode *
+S_reg_namedseq(pTHX_ RExC_state_t *pRExC_state, UV *valuep)
+{
+ char * name; /* start of the content of the name */
+ char * endbrace; /* endbrace following the name */
+ SV *sv_str = NULL;
+ SV *sv_name = NULL;
+ STRLEN len; /* this has various purposes throughout the code */
+ bool cached = 0; /* if this is true then we shouldn't refcount dev sv_str */
+ regnode *ret = NULL;
+
+ if (*RExC_parse != '{') {
+ vFAIL("Missing braces on \\N{}");
+ }
+ name = RExC_parse+1;
+ endbrace = strchr(RExC_parse, '}');
+ if ( ! endbrace ) {
+ RExC_parse++;
+ vFAIL("Missing right brace on \\N{}");
+ }
+ RExC_parse = endbrace + 1;
+
+
+ /* 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} */
+ 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) ) {
+ cp = 0xFFFD;
+ }
+ if (cp > 0xff)
+ RExC_utf8 = 1;
+ if ( valuep ) {
+ *valuep = cp;
+ return NULL;
+ }
+ string = (char)cp;
+ sv_str= newSVpvn(&string, 1);
+ } else {
+ /* fetch the charnames handler for this scope */
+ HV * const table = GvHV(PL_hintgv);
+ SV **cvp= table ?
+ hv_fetchs(table, "charnames", FALSE) :
+ NULL;
+ SV *cv= cvp ? *cvp : NULL;
+ HE *he_str;
+ int count;
+ /* create an SV with the name as argument */
+ sv_name = newSVpvn(name, endbrace - name);
+
+ if (!table || !(PL_hints & HINT_LOCALIZE_HH)) {
+ vFAIL2("Constant(\\N{%s}) unknown: "
+ "(possibly a missing \"use charnames ...\")",
+ SvPVX(sv_name));
+ }
+ if (!cvp || !SvOK(*cvp)) { /* when $^H{charnames} = undef; */
+ vFAIL2("Constant(\\N{%s}): "
+ "$^H{charnames} is not defined",SvPVX(sv_name));
+ }
+
+
+
+ if (!RExC_charnames) {
+ /* make sure our cache is allocated */
+ RExC_charnames = newHV();
+ sv_2mortal((SV*)RExC_charnames);
+ }
+ /* see if we have looked this one up before */
+ he_str = hv_fetch_ent( RExC_charnames, sv_name, 0, 0 );
+ if ( he_str ) {
+ sv_str = HeVAL(he_str);
+ cached = 1;
+ } else {
+ dSP ;
+
+ ENTER ;
+ SAVETMPS ;
+ PUSHMARK(SP) ;
+
+ XPUSHs(sv_name);
+
+ PUTBACK ;
+
+ count= call_sv(cv, G_SCALAR);
+
+ if (count == 1) { /* XXXX is this right? dmq */
+ sv_str = POPs;
+ SvREFCNT_inc_simple_void(sv_str);
+ }
+
+ SPAGAIN ;
+ PUTBACK ;
+ FREETMPS ;
+ LEAVE ;
+
+ if ( !sv_str || !SvOK(sv_str) ) {
+ vFAIL2("Constant(\\N{%s}): Call to &{$^H{charnames}} "
+ "did not return a defined value",SvPVX(sv_name));
+ }
+ if (hv_store_ent( RExC_charnames, sv_name, sv_str, 0))
+ cached = 1;
+ }
+ }
+ if (valuep) {
+ char *p = SvPV(sv_str, len);
+ if (len) {
+ STRLEN numlen = 1;
+ if ( SvUTF8(sv_str) ) {
+ *valuep = utf8_to_uvchr((U8*)p, &numlen);
+ if (*valuep > 0x7F)
+ RExC_utf8 = 1;
+ /* XXXX
+ We have to turn on utf8 for high bit chars otherwise
+ we get failures with
+
+ "ss" =~ /[\N{LATIN SMALL LETTER SHARP S}]/i
+ "SS" =~ /[\N{LATIN SMALL LETTER SHARP S}]/i
+
+ This is different from what \x{} would do with the same
+ codepoint, where the condition is > 0xFF.
+ - dmq
+ */
+
+
+ } else {
+ *valuep = (UV)*p;
+ /* warn if we havent used the whole string? */
+ }
+ if (numlen<len && SIZE_ONLY && ckWARN(WARN_REGEXP)) {
+ vWARN2(RExC_parse,
+ "Ignoring excess chars from \\N{%s} in character class",
+ SvPVX(sv_name)
+ );
+ }
+ } else if (SIZE_ONLY && ckWARN(WARN_REGEXP)) {
+ vWARN2(RExC_parse,
+ "Ignoring zero length \\N{%s} in character class",
+ SvPVX(sv_name)
+ );
+ }
+ if (sv_name)
+ SvREFCNT_dec(sv_name);
+ if (!cached)
+ SvREFCNT_dec(sv_str);
+ return len ? NULL : (regnode *)&len;
+ } else if(SvCUR(sv_str)) {
+
+ 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,
+ (U8)(FOLD ? (LOC ? EXACTFL : EXACTF) : EXACT));
+ s= STRING(ret);
+
+ if ( RExC_utf8 && !SvUTF8(sv_str) ) {
+ sv_utf8_upgrade(sv_str);
+ } else if ( !RExC_utf8 && SvUTF8(sv_str) ) {
+ RExC_utf8= 1;
+ }
+
+ p = SvPV(sv_str, len);
+ pend = p + len;
+ /* len is the length written, charlen is the size the char read */
+ for ( len = 0; p < pend; p += charlen ) {
+ if (UTF) {
+ UV uvc = utf8_to_uvchr((U8*)p, &charlen);
+ if (FOLD) {
+ STRLEN foldlen,numlen;
+ U8 tmpbuf[UTF8_MAXBYTES_CASE+1], *foldbuf;
+ uvc = toFOLD_uni(uvc, tmpbuf, &foldlen);
+ /* Emit all the Unicode characters. */
+
+ for (foldbuf = tmpbuf;
+ foldlen;
+ foldlen -= numlen)
+ {
+ uvc = utf8_to_uvchr(foldbuf, &numlen);
+ if (numlen > 0) {
+ const STRLEN unilen = reguni(pRExC_state, uvc, s);
+ s += unilen;
+ len += unilen;
+ /* In EBCDIC the numlen
+ * and unilen can differ. */
+ foldbuf += numlen;
+ if (numlen >= foldlen)
+ break;
+ }
+ else
+ break; /* "Can't happen." */
+ }
+ } else {
+ const STRLEN unilen = reguni(pRExC_state, uvc, s);
+ if (unilen > 0) {
+ s += unilen;
+ len += unilen;
+ }
+ }
+ } else {
+ len++;
+ REGC(*p, s++);
+ }
+ }
+ if (SIZE_ONLY) {
+ RExC_size += STR_SZ(len);
+ } else {
+ STR_LEN(ret) = len;
+ RExC_emit += STR_SZ(len);
+ }
+ Set_Node_Cur_Length(ret); /* MJD */
+ RExC_parse--;
+ nextchar(pRExC_state);
+ } else {
+ ret = reg_node(pRExC_state,NOTHING);
+ }
+ if (!cached) {
+ SvREFCNT_dec(sv_str);
+ }
+ if (sv_name) {
+ SvREFCNT_dec(sv_name);
+ }
+ return ret;
+
+}
+
+
/*
- - regatom - the lowest level
+ * reg_recode
*
- * 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.
+ * It returns the code point in utf8 for the value in *encp.
+ * value: a code value in the source encoding
+ * encp: a pointer to an Encode object
*
- * [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]
+ * If the result from Encode is not a single character,
+ * it returns U+FFFD (Replacement character) and sets *encp to NULL.
*/
+STATIC UV
+S_reg_recode(pTHX_ const char value, SV **encp)
+{
+ STRLEN numlen = 1;
+ 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)
+ uv = SvUTF8(sv)
+ ? utf8n_to_uvchr((U8*)s, newlen, &numlen, UTF8_ALLOW_DEFAULT)
+ : *(U8*)s;
+
+ if (!newlen || numlen != newlen) {
+ uv = UNICODE_REPLACEMENT;
+ *encp = NULL;
+ }
+ return uv;
+}
+
+
+/*
+ - regatom - the lowest level
+
+ 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);
- if (RExC_flags & PMf_MULTILINE)
+ if (RExC_flags & RXf_PMf_MULTILINE)
ret = reg_node(pRExC_state, MBOL);
- else if (RExC_flags & PMf_SINGLELINE)
+ else if (RExC_flags & RXf_PMf_SINGLELINE)
ret = reg_node(pRExC_state, SBOL);
else
ret = reg_node(pRExC_state, BOL);
nextchar(pRExC_state);
if (*RExC_parse)
RExC_seen_zerolen++;
- if (RExC_flags & PMf_MULTILINE)
+ if (RExC_flags & RXf_PMf_MULTILINE)
ret = reg_node(pRExC_state, MEOL);
- else if (RExC_flags & PMf_SINGLELINE)
+ else if (RExC_flags & RXf_PMf_SINGLELINE)
ret = reg_node(pRExC_state, SEOL);
else
ret = reg_node(pRExC_state, EOL);
break;
case '.':
nextchar(pRExC_state);
- if (RExC_flags & PMf_SINGLELINE)
+ if (RExC_flags & RXf_PMf_SINGLELINE)
ret = reg_node(pRExC_state, SANY);
else
ret = reg_node(pRExC_state, REG_ANY);
}
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 */
*flagp |= HASWIDTH|SIMPLE;
}
break;
- case 'n':
- case 'r':
- case 't':
- case 'f':
- case 'e':
- case 'a':
- case 'x':
- case 'c':
- case '0':
- goto defchar;
+ case 'N':
+ /* Handle \N{NAME} here and not below because it can be
+ multicharacter. join_exact() will join them up later on.
+ Also this makes sure that things like /\N{BLAH}+/ and
+ \N{BLAH} being multi char Just Happen. dmq*/
+ ++RExC_parse;
+ 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 != '\'' && 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);
+ U32 num = 0;
+ SV *sv_dat = reg_scan_name(pRExC_state,
+ SIZE_ONLY ? REG_RSN_RETURN_NULL : REG_RSN_RETURN_DATA);
+ ch= (ch == '<') ? '>' : (ch == '{') ? '}' : '\'';
+ if (RExC_parse == name_start || *RExC_parse != 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;
+
+ /* 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 'g':
case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9':
{
- const I32 num = atoi(RExC_parse);
-
- if (num > 9 && num >= RExC_npar)
+ I32 num;
+ 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_npar - num;
+ if (num < 1)
+ vFAIL("Reference to nonexistent or unclosed group");
+ }
+ 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 (!SIZE_ONLY && num > (I32)RExC_rx->nparens)
- vFAIL("Reference to nonexistent group");
+ 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");
+ }
RExC_sawback = 1;
ret = reganode(pRExC_state,
(U8)(FOLD ? (LOC ? REFFL : REFF) : REF),
break;
case '#':
- if (RExC_flags & PMf_EXTENDED) {
- while (RExC_parse < RExC_end && *RExC_parse != '\n')
- RExC_parse++;
- if (RExC_parse < RExC_end)
+ if (RExC_flags & RXf_PMf_EXTENDED) {
+ 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 & PMf_EXTENDED)
- p = regwhite(p, RExC_end);
- switch (*p) {
+ if (RExC_flags & RXf_PMf_EXTENDED)
+ 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':
+ /* 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 = grok_hex(p, &numlen, &flags, NULL);
p += numlen;
}
+ if (PL_encoding && ender < 0x100)
+ goto recode_encoding;
break;
case 'c':
p++;
--p;
goto loopdone;
}
+ if (PL_encoding && ender < 0x100)
+ goto recode_encoding;
+ break;
+ recode_encoding:
+ {
+ SV* enc = PL_encoding;
+ ender = reg_recode((const char)(U8)ender, &enc);
+ if (!enc && SIZE_ONLY && ckWARN(WARN_REGEXP))
+ vWARN(p, "Invalid escape in the specified encoding");
+ RExC_utf8 = 1;
+ }
break;
case '\0':
if (p >= RExC_end)
ender = *p++;
break;
}
- if (RExC_flags & 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) {
break;
}
- /* If the encoding pragma is in effect recode the text of
- * any EXACT-kind nodes. */
- if (PL_encoding && PL_regkind[OP(ret)] == EXACT) {
- const STRLEN oldlen = STR_LEN(ret);
- SV * const sv = sv_2mortal(newSVpvn(STRING(ret), oldlen));
-
- if (RExC_utf8)
- SvUTF8_on(sv);
- if (sv_utf8_downgrade(sv, TRUE)) {
- const char * const s = sv_recode_to_utf8(sv, PL_encoding);
- const STRLEN newlen = SvCUR(sv);
-
- if (SvUTF8(sv))
- RExC_utf8 = 1;
- if (!SIZE_ONLY) {
- GET_RE_DEBUG_FLAGS_DECL;
- DEBUG_COMPILE_r(PerlIO_printf(Perl_debug_log, "recode %*s to %*s\n",
- (int)oldlen, STRING(ret),
- (int)newlen, s));
- Copy(s, STRING(ret), newlen, char);
- STR_LEN(ret) += newlen - oldlen;
- RExC_emit += STR_SZ(newlen) - STR_SZ(oldlen);
- } else
- RExC_size += STR_SZ(newlen) - STR_SZ(oldlen);
- }
- }
-
return(ret);
}
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;
if (UCHARAT(RExC_parse) == ']')
goto charclassloop;
+parseit:
while (RExC_parse < RExC_end && UCHARAT(RExC_parse) != ']') {
charclassloop:
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
+ multichar strings being returned. I kinda wonder
+ if this makes sense as it does change the behaviour
+ from earlier versions, OTOH that behaviour was broken
+ as well. */
+ UV v; /* value is register so we cant & it /grrr */
+ if (reg_namedseq(pRExC_state, &v)) {
+ goto parseit;
+ }
+ value= v;
+ }
+ break;
case 'p':
case 'P':
{
value = grok_hex(RExC_parse, &numlen, &flags, NULL);
RExC_parse += numlen;
}
+ if (PL_encoding && value < 0x100)
+ goto recode_encoding;
break;
case 'c':
value = UCHARAT(RExC_parse++);
break;
case '0': case '1': case '2': case '3': case '4':
case '5': case '6': case '7': case '8': case '9':
- {
- I32 flags = 0;
- numlen = 3;
- value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
- RExC_parse += numlen;
- break;
- }
+ {
+ I32 flags = 0;
+ numlen = 3;
+ value = grok_oct(--RExC_parse, &numlen, &flags, NULL);
+ RExC_parse += numlen;
+ if (PL_encoding && value < 0x100)
+ goto recode_encoding;
+ break;
+ }
+ recode_encoding:
+ {
+ SV* enc = PL_encoding;
+ value = reg_recode((const char)(U8)value, &enc);
+ if (!enc && SIZE_ONLY && ckWARN(WARN_REGEXP))
+ vWARN(RExC_parse,
+ "Invalid escape in the specified encoding");
+ break;
+ }
default:
if (!SIZE_ONLY && isALPHA(value) && ckWARN(WARN_REGEXP))
vWARN2(RExC_parse,
else {
ANYOF_FLAGS(ret) |= ANYOF_UNICODE;
Perl_sv_catpvf(aTHX_ listsv,
- "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-');
- }
- }
-
- range = 0; /* this was not a true range */
- }
-
- if (!SIZE_ONLY) {
- const char *what = NULL;
- char yesno = 0;
-
- if (namedclass > OOB_NAMEDCLASS)
- optimize_invert = FALSE;
- /* Possible truncation here but in some 64-bit environments
- * the compiler gets heartburn about switch on 64-bit values.
- * 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 ANYOF_ASCII:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_ASCII);
- else {
-#ifndef EBCDIC
- for (value = 0; value < 128; value++)
- ANYOF_BITMAP_SET(ret, value);
-#else /* EBCDIC */
- for (value = 0; value < 256; value++) {
- if (isASCII(value))
- ANYOF_BITMAP_SET(ret, value);
- }
-#endif /* EBCDIC */
- }
- yesno = '+';
- what = "ASCII";
- break;
- case ANYOF_NASCII:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NASCII);
- else {
-#ifndef EBCDIC
- for (value = 128; value < 256; value++)
- ANYOF_BITMAP_SET(ret, value);
-#else /* EBCDIC */
- for (value = 0; value < 256; value++) {
- if (!isASCII(value))
- ANYOF_BITMAP_SET(ret, value);
- }
-#endif /* EBCDIC */
- }
- 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;
- case ANYOF_DIGIT:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
- else {
- /* consecutive digits assumed */
- for (value = '0'; value <= '9'; value++)
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '+';
- what = "Digit";
- break;
- case ANYOF_NDIGIT:
- if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
- else {
- /* consecutive digits assumed */
- for (value = 0; value < '0'; value++)
- ANYOF_BITMAP_SET(ret, value);
- for (value = '9' + 1; value < 256; value++)
- ANYOF_BITMAP_SET(ret, value);
- }
- yesno = '!';
- what = "Digit";
- 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);
+ "%04"UVxf"\n%04"UVxf"\n", (UV)prevvalue, (UV) '-');
}
- yesno = '!';
- what = "SpacePerl";
- break;
- case ANYOF_UPPER:
+ }
+
+ range = 0; /* this was not a true range */
+ }
+
+
+
+ if (!SIZE_ONLY) {
+ const char *what = NULL;
+ char yesno = 0;
+
+ if (namedclass > OOB_NAMEDCLASS)
+ optimize_invert = FALSE;
+ /* Possible truncation here but in some 64-bit environments
+ * the compiler gets heartburn about switch on 64-bit values.
+ * A similar issue a little earlier when switching on value.
+ * --jhi */
+ switch ((I32)namedclass) {
+ 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_UPPER);
+ ANYOF_CLASS_SET(ret, ANYOF_ASCII);
else {
- for (value = 0; value < 256; value++)
- if (isUPPER(value))
- ANYOF_BITMAP_SET(ret, value);
+#ifndef EBCDIC
+ for (value = 0; value < 128; value++)
+ ANYOF_BITMAP_SET(ret, value);
+#else /* EBCDIC */
+ for (value = 0; value < 256; value++) {
+ if (isASCII(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+#endif /* EBCDIC */
}
yesno = '+';
- what = "Upper";
+ what = "ASCII";
break;
- case ANYOF_NUPPER:
+ case ANYOF_NASCII:
if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NUPPER);
+ ANYOF_CLASS_SET(ret, ANYOF_NASCII);
else {
- for (value = 0; value < 256; value++)
- if (!isUPPER(value))
- ANYOF_BITMAP_SET(ret, value);
+#ifndef EBCDIC
+ for (value = 128; value < 256; value++)
+ ANYOF_BITMAP_SET(ret, value);
+#else /* EBCDIC */
+ for (value = 0; value < 256; value++) {
+ if (!isASCII(value))
+ ANYOF_BITMAP_SET(ret, value);
+ }
+#endif /* EBCDIC */
}
yesno = '!';
- what = "Upper";
- break;
- case ANYOF_XDIGIT:
+ what = "ASCII";
+ break;
+ case ANYOF_DIGIT:
if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_XDIGIT);
+ ANYOF_CLASS_SET(ret, ANYOF_DIGIT);
else {
- for (value = 0; value < 256; value++)
- if (isXDIGIT(value))
- ANYOF_BITMAP_SET(ret, value);
+ /* consecutive digits assumed */
+ for (value = '0'; value <= '9'; value++)
+ ANYOF_BITMAP_SET(ret, value);
}
yesno = '+';
- what = "XDigit";
+ what = "Digit";
break;
- case ANYOF_NXDIGIT:
+ case ANYOF_NDIGIT:
if (LOC)
- ANYOF_CLASS_SET(ret, ANYOF_NXDIGIT);
+ ANYOF_CLASS_SET(ret, ANYOF_NDIGIT);
else {
- for (value = 0; value < 256; value++)
- if (!isXDIGIT(value))
- ANYOF_BITMAP_SET(ret, value);
+ /* consecutive digits assumed */
+ for (value = 0; value < '0'; value++)
+ ANYOF_BITMAP_SET(ret, value);
+ for (value = '9' + 1; value < 256; value++)
+ ANYOF_BITMAP_SET(ret, value);
}
yesno = '!';
- what = "XDigit";
- break;
+ what = "Digit";
+ break;
case ANYOF_MAX:
/* this is to handle \p and \P */
break;
{
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)
RExC_parse++;
continue;
}
- if (RExC_flags & PMf_EXTENDED) {
+ if (RExC_flags & RXf_PMf_EXTENDED) {
if (isSPACE(*RExC_parse)) {
RExC_parse++;
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);
}
+ 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);
}
if (SIZE_ONLY) {
SIZE_ALIGN(RExC_size);
RExC_size += 2;
+ /*
+ We can't do this:
+
+ assert(2==regarglen[op]+1);
+
+ Anything larger than this has to allocate the extra amount.
+ If we changed this to be:
+
+ RExC_size += (1 + regarglen[op]);
+
+ then it wouldn't matter. Its not clear what side effect
+ might come from that so its not done so far.
+ -- dmq
+ */
return(ret);
}
+ 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);
}
* Means relocating the operand.
*/
STATIC void
-S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd)
+S_reginsert(pTHX_ RExC_state_t *pRExC_state, U8 op, regnode *opnd, U32 depth)
{
dVAR;
register regnode *src;
register regnode *dst;
register regnode *place;
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",PL_reg_name[op]);
if (SIZE_ONLY) {
- RExC_size += NODE_STEP_REGNODE + offset;
+ RExC_size += size;
return;
}
src = RExC_emit;
- RExC_emit += NODE_STEP_REGNODE + offset;
+ RExC_emit += size;
dst = RExC_emit;
+ if (RExC_open_parens) {
+ int paren;
+ /*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);*/
+ RExC_open_parens[paren] += size;
+ } else {
+ /*DEBUG_PARSE_FMT("open"," - %s","ok");*/
+ }
+ if ( RExC_close_parens[paren] >= opnd ) {
+ /*DEBUG_PARSE_FMT("close"," - %d",size);*/
+ RExC_close_parens[paren] += size;
+ } else {
+ /*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);
SV * const mysv=sv_newmortal();
DEBUG_PARSE_MSG((scan==p ? "tail" : ""));
regprop(RExC_rx, mysv, scan);
- PerlIO_printf(Perl_debug_log, "~ %s (%d)\n",
- SvPV_nolen_const(mysv), REG_NODE_NUM(scan));
+ PerlIO_printf(Perl_debug_log, "~ %s (%d) %s %s\n",
+ SvPV_nolen_const(mysv), REG_NODE_NUM(scan),
+ (temp == NULL ? "->" : ""),
+ (temp == NULL ? PL_reg_name[OP(val)] : "")
+ );
});
if (temp == NULL)
break;
SV * const mysv=sv_newmortal();
DEBUG_PARSE_MSG((scan==p ? "tsdy" : ""));
regprop(RExC_rx, mysv, scan);
- PerlIO_printf(Perl_debug_log, "~ %s (%s) (%d)\n",
+ PerlIO_printf(Perl_debug_log, "~ %s (%d) -> %s\n",
SvPV_nolen_const(mysv),
- reg_name[exact],
- REG_NODE_NUM(scan));
+ REG_NODE_NUM(scan),
+ 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) {
(r->check_substr == r->float_substr
&& r->check_utf8 == r->float_utf8
? "(checking floating" : "(checking anchored"));
- if (r->reganch & ROPT_NOSCAN)
+ if (r->extflags & RXf_NOSCAN)
PerlIO_printf(Perl_debug_log, " noscan");
- if (r->reganch & ROPT_CHECK_ALL)
+ if (r->extflags & RXf_CHECK_ALL)
PerlIO_printf(Perl_debug_log, " isall");
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->reganch & ROPT_ANCH) {
+ if (r->extflags & RXf_ANCH) {
PerlIO_printf(Perl_debug_log, "anchored");
- if (r->reganch & ROPT_ANCH_BOL)
+ if (r->extflags & RXf_ANCH_BOL)
PerlIO_printf(Perl_debug_log, "(BOL)");
- if (r->reganch & ROPT_ANCH_MBOL)
+ if (r->extflags & RXf_ANCH_MBOL)
PerlIO_printf(Perl_debug_log, "(MBOL)");
- if (r->reganch & ROPT_ANCH_SBOL)
+ if (r->extflags & RXf_ANCH_SBOL)
PerlIO_printf(Perl_debug_log, "(SBOL)");
- if (r->reganch & ROPT_ANCH_GPOS)
+ if (r->extflags & RXf_ANCH_GPOS)
PerlIO_printf(Perl_debug_log, "(GPOS)");
PerlIO_putc(Perl_debug_log, ' ');
}
- if (r->reganch & ROPT_GPOS_SEEN)
- PerlIO_printf(Perl_debug_log, "GPOS ");
- if (r->reganch & ROPT_SKIP)
+ if (r->extflags & RXf_GPOS_SEEN)
+ PerlIO_printf(Perl_debug_log, "GPOS:%"UVuf" ", (UV)r->gofs);
+ if (r->intflags & PREGf_SKIP)
PerlIO_printf(Perl_debug_log, "plus ");
- if (r->reganch & ROPT_IMPLICIT)
+ if (r->intflags & PREGf_IMPLICIT)
PerlIO_printf(Perl_debug_log, "implicit ");
- PerlIO_printf(Perl_debug_log, "minlen %ld ", (long) r->minlen);
- if (r->reganch & ROPT_EVAL_SEEN)
+ 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);
- if (OP(o) >= reg_num) /* regnode.type is unsigned */
+
+ if (OP(o) > REGNODE_MAX) /* regnode.type is unsigned */
/* 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");
- sv_catpv(sv, reg_name[OP(o)]); /* Take off const! */
+ Perl_croak(aTHX_ "Corrupted regexp opcode %d > %d", (int)OP(o), (int)REGNODE_MAX);
+ 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">",
(UV)trie->startstate,
- (IV)trie->laststate-1,
+ (IV)trie->statecount-1, /* -1 because of the unused 0 element */
(UV)trie->wordcount,
(UV)trie->minlen,
(UV)trie->maxlen,
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 )
+ 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 == LOGICAL)
+ 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,
+ 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;
return prog->check_substr ? prog->check_substr : prog->check_utf8;
}
+/*
+ pregfree()
+
+ 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_ struct regexp *r)
+Perl_pregfree(pTHX_ REGEXP *r)
{
- dVAR;
+ SvREFCNT_dec(r);
+}
+void
+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();
- if (RX_DEBUG(r)){
- SV *dsv= sv_newmortal();
- RE_PV_QUOTED_DECL(s, (r->reganch & ROPT_UTF8),
- dsv, r->precomp, r->prelen, 60);
+ {
+ SV *dsv= sv_newmortal();
+ 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->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':
- SvREFCNT_dec((SV*)r->data->data[n]);
+ case 'S':
+ 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 (RX_DEBUG(r)) {
- 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(ri->data->what);
+ Safefree(ri->data);
+ }
+
+ Safefree(ri);
+}
+
+#define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
+#define av_dup_inc(s,t) (AV*)SvREFCNT_inc(sv_dup((SV*)s,t))
+#define hv_dup_inc(s,t) (HV*)SvREFCNT_inc(sv_dup((SV*)s,t))
+#define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
+
+/*
+ re_dup - duplicate a regexp.
+
+ 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)
+#ifndef PERL_IN_XSUB_RE
+void
+Perl_re_dup_guts(pTHX_ const REGEXP *sstr, REGEXP *dstr, CLONE_PARAMS *param)
+{
+ dVAR;
+ 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 (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;
+ }
+ }
+ }
+
+ RXp_PAREN_NAMES(ret) = hv_dup_inc(RXp_PAREN_NAMES(ret), param);
+
+ if (ret->pprivate)
+ RXi_SET(ret,CALLREGDUPE_PVT(dstr,param));
+
+ 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
+
+ ret->mother_re = NULL;
+ ret->gofs = 0;
+}
+#endif /* PERL_IN_XSUB_RE */
+
+/*
+ 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;
+
+ if (ri->data) {
+ struct reg_data *d;
+ const int count = ri->data->count;
+ int i;
+
+ Newxc(d, sizeof(struct reg_data) + count*sizeof(void *),
+ char, struct reg_data);
+ Newx(d->what, count, U8);
+
+ d->count = count;
+ for (i = 0; i < count; i++) {
+ d->what[i] = ri->data->what[i];
+ switch (d->what[i]) {
+ /* legal options are one of: sSfpontTu
+ see also regcomp.h and pregfree() */
+ case 's':
+ case 'S':
+ 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(ri->data->data[i], d->data[i],
+ struct regnode_charclass_class);
+ 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*)ri->data->data[i]);
+ OP_REFCNT_UNLOCK;
+ break;
+ case 'T':
+ /* 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.
+ */
+ 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'", ri->data->what[i]);
}
}
- Safefree(r->data->what);
- Safefree(r->data);
+
+ reti->data = d;
}
- Safefree(r->startp);
- Safefree(r->endp);
- Safefree(r);
+ else
+ reti->data = NULL;
+
+ reti->name_list_idx = ri->name_list_idx;
+
+#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
+
+ 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)
+ 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);
DUMPUNTIL(NEXTOPER(node), 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_LTGT
+ 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",(next - dist) - start);
+ PerlIO_printf(Perl_debug_log, "(%"UVuf")\n",
+ (UV)((dist ? this_trie + dist : next) - start));
if (dist) {
if (!nextbranch)
- nextbranch= next - trie->jump[0];
- DUMPUNTIL(next - dist, nextbranch);
- }
+ nextbranch= this_trie + trie->jump[0];
+ DUMPUNTIL(this_trie + dist, nextbranch);
+ }
if (nextbranch && PL_regkind[OP(nextbranch)]==BRANCH)
nextbranch= regnext((regnode *)nextbranch);
} else {
}
if (op == CURLYX || op == OPEN)
indent++;
- else if (op == WHILEM)
- indent--;
}
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;
}