[ruby-talk:7714] "1_000".to_i == 1 bug?

Actually I'm not sure if this is bug, as it could be also a designed
feature, but:

  ruby -e'p "1_000".to_i'
  1

instead of 1000 what I'd expect.

Small patch:

Index: bignum.c
===================================================================
RCS file: /home/cvs/ruby/bignum.c,v
retrieving revision 1.28
diff -u -r1.28 bignum.c
--- bignum.c    2000/11/17 04:41:09     1.28
+++ bignum.c    2000/12/19 20:49:10
@@ -290,6 +290,11 @@
          case 'D': case 'E': case 'F':
            c = c - 'A' + 10;
            break;
+         case '_':
+           if( base != 10 ){
+             rb_raise(rb_eArgError, "non-base 10 string contains '_': %s",
s);
+           }
+           continue;
          default:
            if (badcheck) {
                if (ISSPACE(c)) {

Doesn't solve the whole problem, as there're two cases, long and short
string. So these two examples are handled by strtoul and the above fix and
therefore give different results:

  ruby -e'p "1_0_0_00".to_i'
  1
  ruby -e'p "1_0_0_0_0".to_i'
  10000

I don't dare to hack with strtoul, so I leave it here.

	- Aleksi

BTW. I found the behaviour when I gave 10_000 as a command line argument to
Stephen's formatting code.