Re: [Jscheme-devel] string->number

[Geoffrey K. <ge...@kn...>]

On Jun 7, 2004, at 10:44, Timothy John Hickey wrote:
> This little issue actually goes to the heart of JScheme since what is=20=

> unique about Jscheme is the tight
> integration with Java and that integration is made possibly by our=20
> decision to adopt the Java primitive types
> as the Scheme base types. The Java literal syntax is then essential to=20=

> making JScheme integrate smoothly with
> Java.

This is a worthy goal.  I certainly remember how impressed I was when=20
the first Java language and VM specs came out, so painstakingly=20
developed.  But then even Java has warts, for example:

"How Java=92s Floating-Point Hurts Everyone Everywhere"
http://www.cs.berkeley.edu/~wkahan/JAVAhurt.pdf

Principle of Least Surprise ... What is more of an unpleasant surprise?
1.  (string->number "08") =3D> 8
2.  (string->number "08") =3D> #f
3.  (string->number "010") =3D> 10
4.  (string->number "010") =3D> 8

My feeling is that anyone who wants octal interpretation won't mind=20
saying so:
(string->number "010" 8) =3D> 8

Good morning, Alexander.  How old will you be next year?
    010
I see you entered a leading zero.  You're too clever.  I wasn't=20
expecting that.  Like many 7 year-olds, I bet sometimes you wish you=20
were 10.

Geoffrey
--
Geoffrey S. Knauth | http://knauth.org/gsk