Path: utzoo!attcan!uunet!mcvax!dik
From: dik@cwi.nl (Dik T. Winter)
Newsgroups: comp.unix.wizards
Subject: Re: grep "sorted file" inappropriate (was Re: new grep)
Message-ID: <7573@boring.cwi.nl>
Date: 2 Jul 88 17:37:01 GMT
References: <16237@brl-adm.ARPA> <4219@mtgzz.att.com> <199@elgar.UUCP> <2561@homxc.UUCP> <10132@tekecs.TEK.COM>
Organization: CWI, Amsterdam
Lines: 26

In article <10132@tekecs.TEK.COM> andrew@frip.gwd.tek.com (Andrew Klossner) writes:
 > []
 > 
 > 	"How about an option (only applicable when patterns start with
 > 	"^") to tell grep that the file is lexically sorted (?what
 > 	order?) so it can terminate if the input line is lexically
 > 	greater than the pattern (s).  This is useful for searching
 > 	through sorted lists (files, dictionaries, etc...)"
 > 
 > If your pattern is a string (no meta-characters), what you really want
 > is a binary search, as implemented by the Berkeley look(1) command.
 > This is a different sort of operation than grep's sequential search,
 > and offering it as a separate command makes sense.
 > 
 > If your pattern contains meta-characters, its position in a sorted list
 > isn't necessarily defined.  For example, when do you stop searching
 > when the pattern is "^.rwx"?  You could define rules that specify when
 > to stop searching for each of several different forms of pattern, based
 > on the existing pattern algebra, but this starts to be more complexity
 > than the problem merits.
 > 
And also, ^ at the start of a pattern has already a meaning.
-- 
dik t. winter, cwi, amsterdam, nederland
INTERNET   : dik@cwi.nl
BITNET/EARN: dik@mcvax