From: Christoph <chr_news@...>
Date: 2003-10-15T18:10:14+09:00
Subject: stringy range bug

This is a multi-part message in MIME format.

------=_NextPart_000_0080_01C3930C.DC54EE30
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit

Hi,

currently there is a bug in open ranges with 
identical boundaries  strings.

---
p ("a"..."a").to_a  # ["a"]
---


I also have a back log of old Range questions. 
Shouldn't a failure like

---
class Class
  alias succ superclass
end

class A
end

(A..Object).each {} # raises ... `<=>' for nil:NilClass (NoMethodError)
---

be fixed? (the patch should be straight forward) 
- and why isn't it possible to form ranges with incomparable 
boundaries  - this would make ranges more usefull
when iterating in trees - example

--
class Class
  alias succ superclass
end

class A
end
class B
end

(A..B)   # raises an argument error
---

rather than raising an error we could iterate up to the  
``least common successor'' - in our example this would
be Object?


/Christoph

 







------=_NextPart_000_0080_01C3930C.DC54EE30
Content-Type: application/octet-stream;
	name="string.diff"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment;
	filename="string.diff"

--- string.c.1.176	2003-10-15 10:09:59.382163200 +0200=0A=
+++ string.c	2003-10-15 10:09:25.347153600 +0200=0A=
@@ -1174,9 +1174,12 @@=0A=
 {=0A=
     VALUE current, after_end;=0A=
     ID succ =3D rb_intern("succ");=0A=
+    int limits_cmp;=0A=
 =0A=
     StringValue(end);=0A=
-    if (rb_str_cmp(beg, end) > 0) return beg;=0A=
+    limits_cmp =3D rb_str_cmp(beg, end);=0A=
+=0A=
+    if ((limits_cmp > 0) || (excl && limits_cmp =3D=3D 0)) return beg;=0A=
     after_end =3D rb_funcall(end, succ, 0, 0);=0A=
     current =3D beg;=0A=
     while (!rb_str_equal(current, after_end)) {=0A=

------=_NextPart_000_0080_01C3930C.DC54EE30--