nobu	2011-01-23 08:00:12 +0900 (Sun, 23 Jan 2011)

  New Revision: 30635

  http://svn.ruby-lang.org/cgi-bin/viewvc.cgi?view=rev&revision=30635

  Log:
    * string.c (str_nth_len, str_utf8_nth): return the rest length together.
    * string.c (rb_str_substr): get rid of measure the length always
      to improve performance for huge string.

  Modified files:
    trunk/ChangeLog
    trunk/string.c

Index: ChangeLog
===================================================================
--- ChangeLog	(revision 30634)
+++ ChangeLog	(revision 30635)
@@ -1,3 +1,10 @@
+Sun Jan 23 08:00:09 2011  Nobuyoshi Nakada  <nobu@r...>
+
+	* string.c (str_nth_len, str_utf8_nth): return the rest length together.
+
+	* string.c (rb_str_substr): get rid of measure the length always
+	  to improve performance for huge string.
+
 Sun Jan 23 00:40:10 2011  KOSAKI Motohiro  <kosaki.motohiro@g...>
 
 	* test/test_syslog.rb: Fix to make a lot of test failure if
@@ -9,7 +16,7 @@
 	  when multiple merge keys are specified.
 
 	* test/psych/test_merge_keys.rb: tests for multi-merge key support
-	
+
 Sat Jan 22 11:33:04 2011  Aaron Patterson <aaron@t...>
 
 	* ext/psych/lib/psych/visitors/to_ruby.rb: merge keys are actually
@@ -29,7 +36,7 @@
 
 	* ext/psych/parser.c (parse): fix assertion error when reusing a
 	  parser after an exception has been raised
-	
+
 	* test/psych/test_parser.rb: test for assertion error
 
 Sat Jan 22 04:09:22 2011  Aaron Patterson <aaron@t...>
Index: string.c
===================================================================
--- string.c	(revision 30634)
+++ string.c	(revision 30635)
@@ -1420,9 +1420,10 @@
     return rb_check_string_type(str);
 }
 
-char*
-rb_enc_nth(const char *p, const char *e, long nth, rb_encoding *enc)
+static char*
+str_nth_len(const char *p, const char *e, long *nthp, rb_encoding *enc)
 {
+    long nth = *nthp;
     if (rb_enc_mbmaxlen(enc) == 1) {
         p += nth;
     }
@@ -1435,12 +1436,16 @@
 
         while (p < e && 0 < nth) {
             e2 = p + nth;
-            if (e < e2)
+            if (e < e2) {
+                *nthp = nth;
                 return (char *)e;
+            }
             if (ISASCII(*p)) {
                 p2 = search_nonascii(p, e2);
-                if (!p2)
+                if (!p2) {
+		    *nthp = nth;
                     return (char *)e2;
+                }
                 nth -= p2 - p;
                 p = p2;
             }
@@ -1448,26 +1453,35 @@
             p += n;
             nth--;
         }
-        if (nth != 0)
+        *nthp = nth;
+        if (nth != 0) {
             return (char *)e;
+        }
         return (char *)p;
     }
     else {
-        while (p<e && nth--) {
+        while (p < e && nth--) {
             p += rb_enc_mbclen(p, e, enc);
         }
     }
     if (p > e) p = e;
+    *nthp = nth;
     return (char*)p;
 }
 
+char*
+rb_enc_nth(const char *p, const char *e, long nth, rb_encoding *enc)
+{
+    return str_nth_len(p, e, &nth, enc);
+}
+
 static char*
 str_nth(const char *p, const char *e, long nth, rb_encoding *enc, int singlebyte)
 {
     if (singlebyte)
 	p += nth;
     else {
-	p = rb_enc_nth(p, e, nth, enc);
+	p = str_nth_len(p, e, &nth, enc);
     }
     if (!p) return 0;
     if (p > e) p = e;
@@ -1492,8 +1506,9 @@
 
 #ifdef NONASCII_MASK
 static char *
-str_utf8_nth(const char *p, const char *e, long nth)
+str_utf8_nth(const char *p, const char *e, long *nthp)
 {
+    long nth = *nthp;
     if ((int)SIZEOF_VALUE < e - p && (int)SIZEOF_VALUE * 2 < nth) {
 	const VALUE *s, *t;
 	const VALUE lowbits = sizeof(VALUE) - 1;
@@ -1516,13 +1531,14 @@
 	}
 	p++;
     }
+    *nthp = nth;
     return (char *)p;
 }
 
 static long
 str_utf8_offset(const char *p, const char *e, long nth)
 {
-    const char *pp = str_utf8_nth(p, e, nth);
+    const char *pp = str_utf8_nth(p, e, &nth);
     return pp - p;
 }
 #endif
@@ -1603,16 +1619,18 @@
 	    if (beg < 0) return Qnil;
 	}
     }
-    else if (beg > 0 && beg > str_strlen(str, enc)) {
+    else if (beg > 0 && beg > RSTRING_LEN(str)) {
 	return Qnil;
     }
     if (len == 0) {
+	if (beg > str_strlen(str, enc)) return Qnil;
 	p = 0;
     }
 #ifdef NONASCII_MASK
     else if (ENC_CODERANGE(str) == ENC_CODERANGE_VALID &&
         enc == rb_utf8_encoding()) {
-        p = str_utf8_nth(s, e, beg);
+        p = str_utf8_nth(s, e, &beg);
+        if (beg > 0) return Qnil;
         len = str_utf8_offset(p, e, len);
     }
 #endif
@@ -1621,15 +1639,15 @@
 
 	p = s + beg * char_sz;
 	if (p > e) {
-	    p = e;
-	    len = 0;
+	    return Qnil;
 	}
         else if (len * char_sz > e - p)
             len = e - p;
         else
 	    len *= char_sz;
     }
-    else if ((p = str_nth(s, e, beg, enc, 0)) == e) {
+    else if ((p = str_nth_len(s, e, &beg, enc)) == e) {
+	if (beg > 0) return Qnil;
 	len = 0;
     }
     else {

--
ML: ruby-changes@q...
Info: http://www.atdot.net/~ko1/quickml/