Re: patch for String.concat

Your Integer.ord(encoding) looks good. I suppose that String.concat 
should now just use Integer.ord when given an integer argument.

	David

Nobuyoshi Nakada wrote:
> 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);
> 
>