perlunifaq, uniintro: fix for 80 col display

[p5sagit/p5-mst-13.2.git] / regcomp.h

diff --git a/regcomp.h b/regcomp.h
index d3c75f0..a20d6e1 100644 (file)
--- a/regcomp.h
+++ b/regcomp.h
@@ -18,6 +18,25 @@ typedef OP OP_4tree;                 /* Will be redefined later. */
 /* Be really agressive about optimising patterns with trie sequences? */
 #define PERL_ENABLE_EXTENDED_TRIE_OPTIMISATION 1
 
+/* Use old style unicode mappings for perl and posix character classes
+ *
+ * NOTE: Enabling this essentially breaks character class matching against unicode 
+ * strings, so that POSIX char classes match when they shouldn't, and \d matches 
+ * way more than 10 characters, and sometimes a charclass and its complement either
+ * both match or neither match.
+ * NOTE: Disabling this will cause various backwards compatibility issues to rear 
+ * their head, and tests to fail. However it will make the charclass behaviour 
+ * consistant regardless of internal string type, and make character class inversions
+ * consistant. The tests that fail in the regex engine are basically broken tests.
+ *
+ * Personally I think 5.12 should disable this for sure. Its a bit more debatable for
+ * 5.10, so for now im leaving it enabled.
+ * XXX: It is now enabled for 5.11/5.12
+ *
+ * -demerphq
+ */
+#define PERL_LEGACY_UNICODE_CHARCLASS_MAPPINGS 1
+
 /* Should the optimiser take positive assertions into account? */
 #define PERL_ENABLE_POSITIVE_ASSERTION_STUDY 0
 
@@ -178,7 +197,7 @@ struct regnode_2 {
 
 
 #define ANYOF_BITMAP_SIZE      32      /* 256 b/(8 b/B) */
-#define ANYOF_CLASSBITMAP_SIZE  4      /* up to 40 (8*5) named classes */
+#define ANYOF_CLASSBITMAP_SIZE  4      /* up to 32 (8*4) named classes */
 
 /* also used by trie */
 struct regnode_charclass {
@@ -317,9 +336,9 @@ struct regnode_charclass_class {    /* has [[:blah:]] classes */
 #define ANYOF_NALNUM    1
 #define ANYOF_SPACE     2      /* \s */
 #define ANYOF_NSPACE    3
-#define ANYOF_DIGIT     4
+#define ANYOF_DIGIT     4      /* \d */
 #define ANYOF_NDIGIT    5
-#define ANYOF_ALNUMC    6      /* isalnum(3), utf8::IsAlnum, ALNUMC */
+#define ANYOF_ALNUMC    6      /* [[:alnum:]] isalnum(3), utf8::IsAlnum, ALNUMC */
 #define ANYOF_NALNUMC   7
 #define ANYOF_ALPHA     8
 #define ANYOF_NALPHA    9
@@ -528,10 +547,10 @@ struct reg_data {
 #define check_offset_max substrs->data[2].max_offset
 #define check_end_shift substrs->data[2].end_shift
 
-#define RX_ANCHORED_SUBSTR(rx) ((rx)->anchored_substr)
-#define RX_ANCHORED_UTF8(rx)   ((rx)->anchored_utf8)
-#define RX_FLOAT_SUBSTR(rx)    ((rx)->float_substr)
-#define RX_FLOAT_UTF8(rx)      ((rx)->float_utf8)
+#define RX_ANCHORED_SUBSTR(rx) (((struct regexp *)SvANY(rx))->anchored_substr)
+#define RX_ANCHORED_UTF8(rx)   (((struct regexp *)SvANY(rx))->anchored_utf8)
+#define RX_FLOAT_SUBSTR(rx)    (((struct regexp *)SvANY(rx))->float_substr)
+#define RX_FLOAT_UTF8(rx)      (((struct regexp *)SvANY(rx))->float_utf8)
 
 /* trie related stuff */
 
@@ -567,6 +586,15 @@ struct _reg_trie_state {
   } trans;
 };
 
+/* info per word; indexed by wordnum */
+typedef struct {
+    U16  prev; /* previous word in acceptance chain; eg in
+                * zzz|abc|ab/ after matching the chars abc, the
+                * accepted word is #2, and the previous accepted
+                * word is #3 */
+    U32 len;   /* how many chars long is this word? */
+    U32 accept;        /* accept state for this word */
+} reg_trie_wordinfo;
 
 
 typedef struct _reg_trie_state    reg_trie_state;
@@ -584,15 +612,14 @@ struct _reg_trie_data {
     reg_trie_state  *states;         /* state data */
     reg_trie_trans  *trans;          /* array of transition elements */
     char            *bitmap;         /* stclass bitmap */
-    U32             *wordlen;        /* array of lengths of words */
     U16            *jump;           /* optional 1 indexed array of offsets before tail 
                                         for the node following a given word. */
-    U16                    *nextword;       /* optional 1 indexed array to support linked list
-                                        of duplicate wordnums */
+    reg_trie_wordinfo *wordinfo;     /* array of info per word */
     U16             uniquecharcount; /* unique chars in trie (width of trans table) */
     U32             startstate;      /* initial state - used for common prefix optimisation */
     STRLEN          minlen;          /* minimum length of words in trie - build/opt only? */
     STRLEN          maxlen;          /* maximum length of words in trie - build/opt only? */
+    U32             prefixlen;       /* #chars in common prefix */
     U32             statecount;      /* Build only - number of states in the states array 
                                         (including the unused zero state) */
     U32             wordcount;       /* Build only */
@@ -710,6 +737,7 @@ re.pm, especially to the documentation.
 #define RE_DEBUG_EXTRA_STATE       0x080000
 #define RE_DEBUG_EXTRA_OPTIMISE    0x100000
 #define RE_DEBUG_EXTRA_BUFFERS     0x400000
+#define RE_DEBUG_EXTRA_GPOS        0x800000
 /* combined */
 #define RE_DEBUG_EXTRA_STACK       0x280000
 
@@ -765,6 +793,8 @@ re.pm, especially to the documentation.
 #define DEBUG_TRIE_r(x) DEBUG_r( \
     if (re_debug_flags & (RE_DEBUG_COMPILE_TRIE \
         | RE_DEBUG_EXECUTE_TRIE )) x )
+#define DEBUG_GPOS_r(x) DEBUG_r( \
+    if (re_debug_flags & RE_DEBUG_EXTRA_GPOS) x )
 
 /* initialization */
 /* get_sv() can return NULL during global destruction. */