Re: Regex in sift statements

[Matti Aarnio <mea@nic.funet.fi>]

Felix Lee wrote:
> > > 	The following regex behaves in a way that reveals my ignorance of 
> > > regexes. 
> > > 	(.*)::([A-Z]*)	
> > > 	This will match on CAN:: and CANEDM::Q but fails for CANEDM::QQ
> 
> I think it's a side effect of the token-matching semantics of sift.
> [A-Z] matches a 1-char token, not a char within a token.
> 
> personally, I think the token-matching semantics was a bad idea.  It's
> not quite flexible enough and it's easy to make mistakes.  I'd prefer
> normal regexps augmented with metachars that match whole tokens.

	Ok, I have now coded   string-sift in addition to the present
	token-sift.  The stuff is mostly that of mr. Thomas Knott,
	but not quite, because I preserved the old stuff in there.

	To minimize the impact on the scripts the new "SIFT" is known
	with name  "SSIFT", and the old one is still the "SIFT"..

	I think better would be to call them "SIFT" (string-Sift),
	and "TSIFT" (token-Sift), but it would mean a serious rewrite
	on the config scripts...

	Or perhaps they should be called "SSIFT" and "TSIFT", so that
	nobody would fall on a trap when things change ?

	Comments ?   (... before I make the dump with them ...)
> --

	/Matti Aarnio	<mea@nic.funet.fi>