[ILUG] grepping file contents

[Pádraig Brady]

Brian Foster wrote:
>   | Date: Tue, 10 Oct 2006 13:36:29 +0530
>   | From: "Walter Faleiro" <curtorkar at gmail.com>
>   | 
>   | Currently I am stuck with a mailbox of 950M.  I need to yank
>   | lines from the line number 16186834 to 16204487.  [ ... ]
>   | Can I use a command that will "cut" lines from 16186834 to
>   | 16204487 and redirect the output to another file?
> 
>  sed -n -e 16186834,16204487p
> 
>   should do it;  and to be a shade faster
>   (esp. if the last line, 16204487, is not
>   "close" to the end)
> 
>  sed -n -e 16186834,16204487p -e 16204487q

Even with the quit early, sed is slower than `head | tail`
Even though there is an extra copy involved between head and tail,
sed's line processing and script processing overhead is greater:

$ grep MemTotal /proc/meminfo
MemTotal:      1027836 kB
$ yes 0123456789 | head -20000000 > lines
$ time head -16204487 < lines | tail -17653 > /dev/null
real    0m0.893s
$ time sed -n '16186834,16204487p;16204487q' < lines > /dev/null
real    0m15.140s
$ export LANG=C
$ time head -16204487 < lines | tail -17653 > /dev/null
real    0m0.893s
$ time sed -n '16186834,16204487p;16204487q' < lines > /dev/null
real    0m3.307s

Pádraig.