Re: patch for String.concat

Hi,

At Sun, 21 Oct 2007 15:50:39 +0900,
David Flanagan wrote in [ruby-core:12797]:
> It occurs to me that my patch for concat may be overkill.  Appending 
> codpoints to strings is presumably for backward compatibility with previous 
> versions when ?x returned an integer instead of a string. 
> 
> So making << work with codepoints > 255 may not be necessary.  Instead, 
> perhaps it would make more sense for it to continue to work only for 
> codepoints < 256, as it did in 1.8. 

Maybe true, but the way to append or convert a codepoint to a
string will be still necessary.  My thought at this moment is
to add an optional encoding argument to Integer#chr.

> The current behavior is unsatisfactory, however.  If I attempt to concat a 
> Unicode codepoint (> 255) to an ASCII string, it doesn't give an error, but 
> just appends the low-order bits and silently discards the rest.  A bug, 
> perhaps in enc/ascii.c? 

It's done in regenc.c, onigenc_single_byte_code_to_mbc() and
onigenc_single_byte_code_to_mbclen(), actually.


Index: common.mk
===================================================================
--- common.mk	(revision 13752)
+++ common.mk	(working copy)
@@ -455,5 +455,5 @@ math.$(OBJEXT): {$(VPATH)}math.c {$(VPAT
   {$(VPATH)}defines.h {$(VPATH)}intern.h {$(VPATH)}missing.h
 numeric.$(OBJEXT): {$(VPATH)}numeric.c {$(VPATH)}ruby.h {$(VPATH)}config.h \
-  {$(VPATH)}defines.h {$(VPATH)}intern.h \
+  {$(VPATH)}defines.h {$(VPATH)}intern.h {$(VPATH)}encoding.h \
   {$(VPATH)}missing.h
 object.$(OBJEXT): {$(VPATH)}object.c {$(VPATH)}ruby.h {$(VPATH)}config.h \
Index: numeric.c
===================================================================
--- numeric.c	(revision 13752)
+++ numeric.c	(working copy)
@@ -12,4 +12,5 @@
 
 #include "ruby/ruby.h"
+#include "ruby/encoding.h"
 #include <ctype.h>
 #include <math.h>
@@ -1803,23 +1804,38 @@ int_pred(VALUE num)
 /*
  *  call-seq:
- *     int.chr    => string
+ *     int.chr([encoding])    => string
  *
- *  Returns a string containing the ASCII character represented by the
- *  receiver's value.
+ *  Returns a string containing the character represented by the
+ *  receiver's value according to +encoding+.
  *
  *     65.chr    #=> "A"
  *     230.chr   #=> "\346"
+ *     255.chr(Encoding::UTF_8)   #=> "\303\277"
  */
 
 static VALUE
-int_chr(VALUE num)
+int_chr(int argc, VALUE *argv, VALUE num)
 {
-    char c;
+    int n;
     long i = NUM2LONG(num);
+    rb_encoding *enc;
+    VALUE str;
 
-    if (i < 0 || 0xff < i)
+    switch (argc) {
+      case 0:
+	enc = rb_enc_primary();
+	break;
+      case 1:
+	enc = rb_to_encoding(argv[0]);
+	break;
+      default:
+	rb_raise(rb_eArgError, "wrong number of arguments (%d for 0 or 1)", argc);
+	break;
+    }
+    if (i < 0 || (n = rb_enc_codelen(i, enc)) <= 0)
 	rb_raise(rb_eRangeError, "%ld out of char range", i);
-    c = i;
-    return rb_str_new(&c, 1);
+    str = rb_enc_str_new(0, n, enc);
+    rb_enc_mbcput(i, RSTRING_PTR(str), enc);
+    return str;
 }
 
@@ -3075,5 +3091,5 @@ Init_Numeric(void)
     rb_define_method(rb_cInteger, "next", int_succ, 0);
     rb_define_method(rb_cInteger, "pred", int_pred, 0);
-    rb_define_method(rb_cInteger, "chr", int_chr, 0);
+    rb_define_method(rb_cInteger, "chr", int_chr, -1);
     rb_define_method(rb_cInteger, "to_i", int_to_i, 0);
     rb_define_method(rb_cInteger, "to_int", int_to_i, 0);


-- 
Nobu Nakada