From: "Ara.T.Howard" <ahoward@...>
Date: 2004-07-22T07:32:03+09:00
Subject: Re: substring by range parameter (bug?)

On Thu, 22 Jul 2004, D T wrote:

>>
>> for ranges, start must be <= end and -1 is not <= -2
>>
>> eg. for the string "foobar":
>>
>>     ---- --- ----
>>     char idx nidx
>>     ---- --- ----
>>     f    0    -6
>>     o    1    -5
>>     o    2    -4
>>     b    3    -3
>>     a    4    -2
>>     r    5    -1
>>
>>   ~ > irb
>>   irb(main):001:0> "foobar"[-2..-1]
>>   => "ar"
>>   irb(main):002:0> "foobar"[-6..-4]
>>   => "foo"
>>   irb(main):003:0> "foobar"[-3..-1]
>>   => "bar"
>>
>> cheers.
>>
>
> Thanks. I know that.

sorry - did not read your post closely enough.

> That is why I expect "foobar"[-1..-2] returns nil as "foobar"[-1..-3] did.
> But it returns an empty string instead of nil, it seems to me that this is a
> bug.

hmmmm.  does seem strange.

> In fact, if start is -n and end is < -(n+1), it will return nil.  ( like
> -2..-4, -2..-5 )
> The problem is only when start is -n and end is -(n+1), it always return
> empty string.  (like -2..-3, -3..-4)

strange.

> How does it possible that a substring is empty string ?

well, there is always

   idx = 0
   len = 0
   'foobar'[idx, len]

> And the tricky part is, it do existe and you can re-assign to it too. like:
> s = "foobar"; s[-3..-4] = "xxx";
> So, my question is, is this a bug, or special feature for substring when
> range is -n..-(n+1) ?

i can only imagine it happen somehow because

   len = end - start_inclusive + 1

so

   0 = (-4 - -3) + 1

and 0 len give back empty string... kind of looks like it from the source:

   static VALUE
   rb_str_aref(str, indx)
       VALUE str;
       VALUE indx;
   {
       long idx;

       switch (TYPE(indx)) {

<snip>

         default:
           /* check if indx is Range */
           {
               long beg, len;
               switch (rb_range_beg_len(indx, &beg, &len, RSTRING(str)->len, 0)) {
                 case Qfalse:
                   break;
                 case Qnil:
                   return Qnil;
                 default:
                   return rb_str_substr(str, beg, len);
               }
           }
           idx = NUM2LONG(indx);
           goto num_index;
       }
       return Qnil;                /* not reached */
   }


i've looked at rb_range_beg_len and this looks to be the problem - but how to
fix it?  should these types of ranges throw an error in the first place?  like

-1...-1

??

-a
--
===============================================================================
| EMAIL   :: Ara [dot] T [dot] Howard [at] noaa [dot] gov
| PHONE   :: 303.497.6469
| A flower falls, even though we love it;
| and a weed grows, even though we do not love it. 
|   --Dogen
===============================================================================