RE: Does VW claim is it an ANSI compliant Smalltalk?

> -----Original Message-----
> From: [hidden email] [mailto:[hidden email]] On
> Behalf Of Steve Dahl
> Sent: Wednesday, May 10, 2006 10:06 PM
> To: Travis Griggs
> Cc: VW DEV
> Subject: Re: Does VW claim is it an ANSI compliant Smalltalk?
>
>
> VW claims that it is somewhat ANSI-compliant, and that it wants to
> become more-so over time, but not at the expense of:
> - motherhood and apple pie
> - some other project that looked like it would benefit customers more
> than a little more ANSI compliance
> - breaking customers' code in ways that cost us more good will than we
> earn from adding the ANSI feature.
>
> In the case of binary selectors, it's been the third one that's caused
> us to at least hesitate.
>
> ANSI, as I recall it, specifies arbitrary length binary selectors, no
> restrictions. VisualWorks indeed says that binary selectors can't be
> longer than two characters, but more importantly, it says that if a
> binary selector ends with a -, it can only be one character long.
>
> This is important because - is also a unary prefix for constant numbers.
> If we don't make this restriction, then "3@-4" would be interpreted as
> "3 @- 4" rather than "3 @ -4". I think there are three places where
> customers could get bit by this (but there might be more):
>
> - source code of methods
> - storeStrings of objects, perhaps stored in a database
> - code generators, including but not limited to SomeClass>>storeOn:,
> UIDefiner, and BrowserNavigator's code for creating default methods in
> classes
>
> Addressing all three of these to customers' satisfaction sounds hard to
> me, although I could be wrong. What would be much easier would be if we
> adopted only partial ANSI compliance: we allow arbitrary length of
> binary selectors, but only if the last character is not -. The only
> binary selector that would be allowed to end in - is #-.
>
> So #<--->, #--->, and #&!@?/ would be valid binary selectors, but #<---
> would not, because its last character is -. This is essentially
> completely backward compatible with all existing VW code, but it's not
> 100% ANSI-compatible. Older versions of VW would not be able to file in
> new VW code that used long binary selectors, but that's about the only
> place existing code would break.
>
> The question, for Travis and for anyone else interested, is whether this
> is ANSI enough. Has anyone here seen, in some other Smalltalk, a use of
> a long binary selector ending in -, that was not merely testing the
> limits of binary selectors, but ended in - for a good reason? If we
> disallowed long binary selectors ending in -, would you ever expect code
> written in some other Smalltalk to contain a banned selector?
>
>
> Travis Griggs wrote:
> >
> > I was just reading the spec and noting that binary selectors should  be
> > one or more infix characters, no length limit. But it still seems  to be
> > the case that we're limited to 2.
> >
> > I might want a shortcut for error: aString ya know:
> >
> > self ?!?!?!? aString
> >
> > :)
> >
> > --
> > Travis Griggs
> > Objologist
> > "You A students, you'll be back soon teaching here with me. You B
> > students, you'll actually go on to be real engineers. You C students,
> > you'll go into management and tell the A and B students what to do."  -
> > My Fluid Dynamics Professor whom I have yet to disprove
> >
> >
> >
> > -----------------------------------------
> > DISCLAIMER: This email is bound by the terms and conditions
> > described at
> > http://www.key.net/disclaimer.htm