From: Massimiliano Mirra <list@...>
Date: 2002-02-05T03:14:58+09:00
Subject: Re: bash-like command splitting (for regexp wizards)

On Mon, Feb 04, 2002 at 11:57:23AM +0900, Jeff 'japhy' Pinyan wrote:

> >dividers = ['\|\|', '&&', ';']
> >divs = dividers.collect {|div| "(#{div})"}.join('|')
                                   ^      ^
> That regex does not do what you think it does...
> 
>   /([^jeff]+)|jeff/
> 
> does not mean "one or more non-jeff's in a row, OR jeff", but rather "one
> or more non-j's, non-e's, and non-f's in a row, OR jeff".

But that regex is (note parentheses above):

/([^(\|\|)|(&&)|(;)]+|(\|\|)|(&&)|(;)/

> >'ls || cd dir; grep "word; sentence" && beep'
> This is a job for a more explicit parser -- one that recognizes quotes and
> escapes and all.

Now that I've seen what the solution looks like, I agree.

Before heading for regexps to solve this, I tried to get the hang of
yacc/bison and still keep my sanity, but I have little background in C
and none at all in parsers.  I am trying out rockit, which feels much
better to me than yacc/bison-like, however it is still in alpha stage.

Can someone point out some easy, task-oriented yacc documentation (if
any such thing exists)?  All I've found is very feature-oriented, I'd
prefer (shall I say need?) something in the ``to get this, do this''
style.

> I come from a Perl background, so I apologize if my Ruby isn't quite as
> idiomatic as many might be used to.
>   SPLIT_RE = %r{
>     \|\|
>       |  
>     ;
[...]

Wow.  Do you all come to Perl with a degree in Egyptology or do you
get one honoris causa afterwards?  :-)

Thanks a lot!

Massimiliano