namazu-dev(ring)

[Junio Hamano <junio@xxxxxxxxxxx>]

以下は、.namazurc の REPLACE の左辺に regexp 、右辺に pattern を
書けるようにするパッチです。

    REPLACE    /repository/(.*)%2Cv	http://h/cvsweb/\1
    REPLACE    /repository/(.*)/Attic/(.*)%2Cv	http://h/cvsweb/\1/\2

と .namazurc に書くと、namazu で CVS レポジトリの昔の版にある文字
列まで全体に索引をつけておいて、検索した結果を、cvsweb で表示する、
なんてことができるようになります。

本当なら、struct replace を全部書換えて、

struct replace_elem {
    struct replace_elem *next;
    uchar *src;
    uchar *dst;
    /* The following fields are NULL if this is a traditional
     * string substitution
     */
    struct re_pattern_buffer *src_exp;
    struct subst_elem {
	enum { literal, regex_regno } subst_type;
	union {
	    uchar *literal_string;
	    int regex_regno;
        } u;
    };
};

なんてやって、conf.c で REPLACE を読み込むところで regexp のコン
パイルと、置き換え文字列の準備まで先に全部やってしまうのがよいの
ですが、かなり大規模な書き換えになるので、TODO-namazu-2.0 にある
「Replace と Alias をもっときれいに書く」までのつなぎとして、変更
個所が最小限になるような形でパッチにしてあります。

まずは使ってみてください。

RCS file: /circus/cvsroot/namazu/src/re_match.c,v
retrieving revision 1.4
diff -u -r1.4 re_match.c
--- src/re_match.c	1999/06/12 14:29:31	1.4
+++ src/re_match.c	1999/07/09 05:41:03
@@ -50,6 +50,82 @@
     strcpy(tmp, s);
 
     while (list.src) {
+        if (strpbrk (list.src->str, ".*")) {
+	    struct re_registers regs;
+	    Regexp *re;
+	    int mlen;
+	    int is_a_regexp_match = 0;
+
+	    regs.allocated = 0;
+	    re = ALLOC(Regexp);
+	    MEMZERO((char *)re, Regexp, 1);
+	    re->buffer = 0;
+	    re->allocated = 0;
+	    if (re_compile_pattern (list.src->str, strlen (list.src->str), re))
+	      /* re_comp fails; maybe it was not a regexp substitution
+	       * after all.  Fall back to string substitution for backward
+	       * compatibility.
+	       */
+	      is_a_regexp_match = 0;
+	    else if (0 < (mlen = re_match (re, tmp, strlen (tmp), 0, &regs))) {
+	      /* We got a match.  Try to replace the string. */
+	      uchar repl[BUFSIZE];
+	      uchar *subst = list.dst->str;
+	      /* Assume we are doing regexp match for now; if any of the
+	       * substitution fails, we will switch back to the straight
+	       * string substitution.
+	       */
+	      is_a_regexp_match = 1;
+	      for (i = j = 0; subst[i]; i++) {
+		/* i scans through RHS of sed-style substitution.
+		 * j points at the string being built.
+		 */
+		if ((subst[i] == '\\') &&
+		    ('0' <= subst[++i]) &&
+		    (subst[i] <= '9')) {
+		  /* A backslash followed by a digit---regexp substitution.
+		   * Note that a backslash followed by anything else is
+		   * silently dropped (including a \\ sequence) and is
+		   * passed on to the else clause.
+		   */
+		  int regno = subst[i] - '0';
+		  int ct;
+		  if (re->re_nsub <= regno) {
+		    /* Oops; this is a bad substitution.  Just give up
+		     * and use straight string substitution for backward
+		     * compatibility.
+		     */
+		    is_a_regexp_match = 0;
+		    break;
+		  }
+		  for (ct = regs.beg[regno]; ct < regs.end[regno]; ct++)
+		    repl[j++] = tmp[ct];
+		}
+		else {
+		  /* Either ordinary character, or an unrecognized \ sequence.
+		   * Just copy it.
+		   */
+		  repl[j++] = subst[i];
+		}
+	      }
+	      if (is_a_regexp_match) {
+		/* Good.  Regexp substitution worked and we now have a good
+		 * string in repl.
+		 * The part that matched and being replaced is 0 to mlen-1
+		 * in tmp; tmp[mlen] through the end of it should be
+		 * concatenated to the end of the resulting string.
+		 */
+		repl[j] = 0;
+		strcpy (s, repl);
+		strcpy (s + j, tmp + mlen);
+	      }
+	      re_free_registers (&regs);
+	    }
+	    re_free_pattern (re);
+	    if (is_a_regexp_match)
+	      return 0;
+	    /* Otherwise, we fall back to string substitution */
+	}
 	n_src = strlen(list.src->str);
 	n_dst = strlen(list.dst->str);