From mboxrd@z Thu Jan  1 00:00:00 1970
To: 9fans@cse.psu.edu
From: Ralph Corderoy <ralph@inputplus.co.uk>
Message-ID: <3828.3f264674.96335@blake.inputplus.co.uk>
References: <1064.3f2242e3.bedc3@blake.inputplus.co.uk>, <4e9a248eadf51083ac5d918159014eda@collyer.net>
Subject: Re: [9fans] text utility
Date: Wed, 30 Jul 2003 08:25:46 +0000
Topicbox-Message-UUID: 0a5a96bc-eacc-11e9-9e20-41e7f4b1d025

Hi Geoff,

> One could use "pr -tn" instead of the little awk script, but the
> difference in speed really isn't that great.  Reading a cached ~2.5MB
> file of ~250K lines,
>
> cpu% time awk '{print NR, $0}' test >/dev/null
> 2.85u 0.13s 3.53r 	 awk {print NR, $0} test
> cpu% time pr -tn test >/dev/null
> 1.75u 0.13s 2.18r 	 pr -tn test

Fair enough.  Under AIX and Linux cat's a win.  Using yes(1) so awk's
little work to do:

    $ yes y | head -n 250000 | time awk '{print NR, $0}' >/dev/null
    real 1.13
    user 1.06
    sys 0.01
    $ yes y | head -n 250000 | time cat -n >/dev/null
    real 0.08
    user 0.05
    sys 0.00

I used it heavily when doing ad-hoc queries on files with millions of
lines.  But, like you say, if it's not present, it's no use.

Cheers,

--
Ralph Corderoy.      http://inputplus.co.uk/ralph/     http://troff.org/