Re: [PATCH - provisional] H. Merijn Brands idea of buffer numbering.

Message-ID: <9b18b3110702071353l250d8a67x188c4e234e8905c7@mail.gmail.com>

p4raw-id: //depot/perl@30169

diff --git a/pod/perlre.pod b/pod/perlre.pod
index d913c80..5287965 100644 (file)
--- a/pod/perlre.pod
+++ b/pod/perlre.pod
@@ -711,6 +711,29 @@ is equivalent to the more verbose
 
     /(?:(?s-i)more.*than).*million/i
 
+=item C<(?|pattern)>
+X<(?|)> X<Branch reset>
+
+This is the "branch reset" pattern, which has the special property
+that the capture buffers are numbered from the same starting point
+in each branch. 
+
+Normally capture buffers in a pattern are number sequentially, left
+to right in the pattern. Inside of this construct this behaviour is
+overriden so that the captures buffers in each branch share the same
+numbers. The numbering in each branch will be as normal, and any 
+buffers following the use of this pattern will be numbered as though
+the construct contained only one branch, that being the one with the
+most capture buffers in it.
+
+Consider the following pattern. The numbers underneath are which
+buffer number the captured content will be stored in.
+
+
+    # before  ---------------branch-reset----------- after        
+    / ( a )  (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
+    # 1            2         2  3        2     3     4  
+
 =item Look-Around Assertions
 X<look-around assertion> X<lookaround assertion> X<look-around> X<lookaround>
 

diff --git a/regcomp.c b/regcomp.c
index 7a74cfc..c75b5f3 100644 (file)
--- a/regcomp.c
+++ b/regcomp.c
@@ -4962,6 +4962,8 @@ S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
     const I32 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    */
@@ -5212,6 +5214,13 @@ S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
                    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;
@@ -5668,6 +5677,11 @@ S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
        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)
@@ -5769,7 +5783,8 @@ S_reg(pTHX_ RExC_state_t *pRExC_state, I32 paren, I32 *flagp,U32 depth)
            FAIL("Junk on end of regexp");      /* "Can't happen". */
        /* NOTREACHED */
     }
-
+    if (after_freeze)
+        RExC_npar = after_freeze;
     return(ret);
 }
 

diff --git a/t/op/re_tests b/t/op/re_tests
index c047d3a..aa6ec7b 100644 (file)
--- a/t/op/re_tests
+++ b/t/op/re_tests
@@ -1270,3 +1270,13 @@ a*(*F)   aaaab   n       -       -
 (?=xy(?<=(aaxy)))      ..aaxy..        y       $1      aaxy
 
 X(\w+)(?=\s)|X(\w+)    Xab     y       [$1-$2] [-ab]
+
+#check that branch reset works ok.
+(?|a(.)b|d(.(o).)d|i(.)(.)j)(.)        d!o!da  y       $1-$2-$3        !o!-o-a
+(?|a(.)b|d(.(o).)d|i(.)(.)j)(.)        aabc    y       $1-$2-$3        a--c
+(?|a(.)b|d(.(o).)d|i(.)(.)j)(.)        ixyjp   y       $1-$2-$3        x-y-p
+(?|(?|(a)|(b))|(?|(c)|(d)))    a       y       $1      a
+(?|(?|(a)|(b))|(?|(c)|(d)))    b       y       $1      b
+(?|(?|(a)|(b))|(?|(c)|(d)))    c       y       $1      c
+(?|(?|(a)|(b))|(?|(c)|(d)))    d       y       $1      d
+(.)(?|(.)(.)x|(.)d)(.) abcde   y       $1-$2-$3-$4-$5- b-c--e--