From mboxrd@z Thu Jan  1 00:00:00 1970
To: 9fans@cse.psu.edu
Subject: Re: [9fans] regexp to match paragraphs in troff documents
From: "Russ Cox" <rsc@swtch.com>
Date: Thu,  7 Jun 2007 13:46:15 -0400
In-Reply-To: <082ffdbcfda12b51637105e95ea479c8@mteege.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Message-Id: <20070607174226.379901E8C1C@holo.morphisms.net>
Topicbox-Message-UUID: 7aa1138c-ead2-11e9-9d60-3106f5b1d025

matthias:
> I've tried to pipe paragraphs of a troff document to fmt but I
> have problems with the correct regular expression. My first attempt
> "/^\.[A-Z]++.*\n(^[^]*)*\n^\.[A-Z]++.*\n/" matches only any second
> paragraph because the expression is overlapping. Does anyone have a nice
> idea to match troff parapgraphs?

It looks like you'd be happy with

	,y/^\..*\n/ |fmt

or, avoiding fmt of zero-length ranges (just a little faster)

	,y/(^\..*\n)+/ |fmt

rog:
> for other cases, i suppose it might be nice to have non-greedy matching,
> in which case you could do something like:
> ,x/^\.[A-Z][A-Z].*\n(.*\n)*?\.[A-Z][A-Z]\n/
> russ: how easy do you think it would be to put non-greedy matching into
> the acme/sam regexp engine?

it's trivial but it doesn't make sense.

in plan 9 regular expressions (as in awk), the semantics
are that the leftmost longest overall match is chosen,
even if that means not repeating a * as much as possible.

for example, consider /a*(ab)?/ against "aab".
perl will match "aa" because the a* greedily grabs "aa"
leaving (ab)? no choice but to match the empty string.
plan 9 will match "aab" because that is a longer match:
the a* selflessly matches less so that the overall
expression can match more.

since plan 9 doesn't have the greedy-like-perl * operator,
it doesn't make sense to think about adding a
non-greedy-like-perl * operator.

the y iterator handles about 90% of the reasons people use
non-greedy operators, so i'm happy to leave things as is.

russ