Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cornell!uw-beaver!ubc-cs!van-bc!eslvcr!ted
From: ted@eslvcr.UUCP (Ted Powell)
Newsgroups: comp.unix.questions
Subject: Re: Help a novice: Will "sed" do?
Summary: Consider awk for processing multiline records
Keywords: multiline records, awk
Message-ID: <165@eslvcr.UUCP>
Date: 20 Jul 89 03:40:44 GMT
References: <2180@umbc3.UMBC.EDU> <10540@smoke.BRL.MIL>
Reply-To: ted@eslvcr.wimsey.bc.ca (Ted Powell)
Followup-To: comp.unix.questions
Organization: Entropy Limited, Vancouver, BC
Lines: 40

In article <10540@smoke.BRL.MIL> gwyn@brl.arpa (Doug Gwyn) writes:
>In article <2180@umbc3.UMBC.EDU> rostamia@umbc3.UMBC.EDU (Dr. Rouben Rostamian) writes:
>>I need a command or a script that searches a text file for a given
>>word or pattern and prints out all paragraphs that contain that word
>>or pattern.  Paragraphs are blocks of text separated by one or 
>>more blank lines.
>
>It's pretty hard to do this with standard UNIX text-file utilities,
>because most of them work on a line-at-a-time basis.  That means when
>you find the pattern, it's too late to output the preceding lines.

Use awk! See section 3.4 Multiline Records in:
	The AWK Programming Language
	Aho, Alfred V., Kernighan, Brian W., Weinberger, Peter J.
	Addison-Wesley Series in Computer Science
	ISBN 0-201-07981-X

Given input data with paragraphs separated by blank lines, the following 
passes those paragraphs containing "New York" (example taken from page 83):
	... | awk 'BEGIN { RS = ""; ORS = "\n\n" }; /New York/' | ...

Setting RS (input Record Separator) to null makes awk take everything between
successive blank lines as a record. Setting ORS (Output Record Separator) to 
two newlines gives a blank line between output records. The example could also
be done as:
	awk '
	BEGIN { RS = ""; ORS = "\n\n" }
	/New York/
	' input-file >output-file
or the program can be hidden away in a file ( -f progfile ).

Patterns can be _very_ complex, and you can have multiple patterns with 
corresponding actions. In the example, the action is unspecified, and defaults
to outputting the current record. If you don't have access to the book, see the
man page. Note that in SVR3/386 (and possibly other flavours) there is AWK(1)
and NAWK(1) (New AWK). (Old awk is being kept around for a while, presumably
for compatibility reasons.) The book corresponds to NAWK(1). At least in
SVR3/386, awk/nawk come with the basic system. If you haven't ever used awk,
give it a try. If you haven't read the book, check it out -- it has all kinds
of useful examples in a surprisingly wide range of fields.