From: brucedickey <brucedickey@...>
Date: 2001-08-01T03:44:46+09:00
Subject: [ruby-talk:18907] Re: Digits in method names



> -----Original Message-----
> From: Ned Konz [mailto:ned@bike-nomad.com]
> Sent: Tuesday, July 31, 2001 12:22 PM
> To: ruby-talk@ruby-lang.org
> Subject: [ruby-talk:18905] Re: Digits in method names
> 
> 
> On Tuesday 31 July 2001 10:24 am, brucedickey wrote:
> 
> > Most mainstream
> > parsers are limited to one character of look-ahead (if any) 
> (are LALR,
> > LL(1), etc), and that is the reason for the restriction.
> 
> My impression is that typical parsers have one _token_ of 
> lookahead, and that 
> typically a lexer would be responsible for assembling 
> _characters_ into 
> _tokens_. And it's pretty simple to determine where tokens end.

Yes, and the lexer (speaking of a separate lexer and parser) has one
character of lookahead which leads to the restriction on identifiers
starting with digits. I had lumped the lexer and parser together and called
it a "parser" -- I see that wasn't very clear. I'm used to tools that use a
single grammar file for both.

> The ambiguity I'd see would be with the representation of 
> floating point 
> numbers: what is "3e5" ?

A lexer with K characters of lookahead, where K is rather large would gobble
characters until a non-identifier/number character is encountered. A
semantic function could then determine whether the accumulated buffer holds
a string meeting the requirements of a floating point number. 


> -- 
> Ned Konz
> currently: Stanwood, WA
> email:     ned@bike-nomad.com
> homepage:  http://bike-nomad.com
>