I used whatever was on my Fedora 13 box:
[sean@bob ~]$ awk --version GNU Awk 3.1.8 [sean@bob ~]$ sed --version GNU sed version 4.2.1
The difference gets much bigger if you use a more complex regexp.
[sean@bob tmp]$ time awk '/.*output.*start.*/,/.*output.*end.*/' < infile > /dev/null
real 0m0.450s user 0m0.393s sys 0m0.010s [sean@bob tmp]$ time sed -n '/.*output.*start.*/,/.*output.*end.*/p' < infile > /dev/null
real 0m1.726s user 0m1.495s sys 0m0.017s
Awk didn't seem to blink an eye. Strangely enough, since the beginning and ending .*'s are completely superfluous, they seem to throw sed for a loop, even if the middle .* is replaced with a space.
Sean
On Wed, Nov 10, 2010 at 12:37 PM, Gilles Detillieux < grdetil@scrc.umanitoba.ca> wrote:
Interesting! Which version of awk did you test? I have to admit I haven't looked into awk performance in quite some time. My early experience, on older Unix systems (pre-Linux), confirmed what I had read about awk being pretty slow. But I seem to recall that even on older Linux systems, gawk wasn't exactly speedy then either. I imagine the GNU awk developers must have remedied that since, though, if that is indeed what you were testing.
Searching online for discussions on awk performance found one from 2002 suggesting gawk was much faster than nawk, and another from this past August that suggested the opposite. Perhaps the developers of the two have been leap-frogging each other with optimizations to their code?
On 11/10/2010 11:56 AM, Sean Walberg wrote:
Adam and I were having an offline discussion, and some testing shows that AWK outperforms SED by a slight margin:
[sean@bob tmp]$ W=/usr/share/dict/words [sean@bob tmp]$ (tail -1000 $W; echo output start; cat $W; echo output end; head -1000 $W) > infile [sean@bob tmp]$ wc -l infile 481831 infile [sean@bob tmp]$ time awk '/output start/,/output end/' < infile >
/dev/null
real 0m0.411s user 0m0.393s sys 0m0.016s [sean@bob tmp]$ time sed -n '/output start/,/output end/p' < infile > /dev/null
real 0m0.678s user 0m0.631s sys 0m0.029s
I ran it a bunch more times and the results were similar. YMMV, benchmarks are lies, etc.
Sean
On Wed, Nov 10, 2010 at 11:32 AM, Gilles Detillieux <grdetil@scrc.umanitoba.ca mailto:grdetil@scrc.umanitoba.ca> wrote:
I may have misinterpreted the question before. If you want the"output
start" and "output end" marker lines in the output (which I guessyour
grep pipeline would do), then Adam's sed script will do that. Mine, using the "d" commands, will output only the data in between. The shortest awk script to do the same would be: awk '/output start/{s=1};s==1;/output end/{s=0};' or awk '/output end/{s=0};s==1;/output start/{s=1};' The first is a simplification of Adam's, which outputs the outputmarker
lines, while the second, using the same statements in the opposite order, suppresses the markers. Of perl, awk and sed, I suspect sedis
the most lightweight, and probably the quickest, unless perl can outperform sed on larger files. awk has a reputation for beingpretty
slow. I tend to favour sed unless awk or perl makes the job a lot easier. Gilles On 11/10/2010 11:13 AM, Adam Thompson wrote: > The AWK version is functionally identical, and not very much shorter, or > any more elegant: > > awk ‘/output start/ {s=1};{if (s==1) print $0};/output end/ {s=0}’ > > (the perl version can generally be made that small, too.) > > > > I would instead suggest sed(1), since this is precisely what it’s > designed for: > > sed –n ‘/output start/,/output end/p’ < infile > > > > -Adam > > > > > > *From:* roundtable-bounces@muug.mb.ca <mailto:roundtable-bounces@muug.mb.ca> > [mailto:roundtable-bounces@muug.mb.ca <mailto:roundtable-bounces@muug.mb.ca>] *On Behalf Of *Sean Walberg > *Sent:* Wednesday, November 10, 2010 10:56 > *To:* Continuation of Round Table discussion > *Subject:* Re: [RndTbl] Command line challenge: trim garbage from start > and end of a file. > > > > OTTOMH: > > > > perl -n -e 'BEGIN {$state = 0} $state = 1 if ($state == 0 and/output
> start/); $state = 2 if ($state == 1 and /output end/) ; print if > ($state == 1)' < infile > outfile > > I'll bet there's a shorter AWK version though. > > > > Sean > > > > On Wed, Nov 10, 2010 at 10:51 AM, John Lange <john@johnlange.ca <mailto:john@johnlange.ca> > <mailto:john@johnlange.ca <mailto:john@johnlange.ca>>> wrote: > > I have files with the following structure: > > garbage > garbage > garbage > output start > .. good data > .. good data > .. good data > .. good data > output end > garbage > garbage > garbage > > How can I extract the good data from the file trimming the garbage > from the beginning and end? > > The following works just fine but it's dirty because I don't likethe
> fact that I have to pick an arbitrarily large number for the"before"
> and "after" values. > > grep -A 999999 "output start" <infile> | grep -B 999999 "output end" > > newfile > > Can anyone come up with something more elegant? > > -- > John Lange > www.johnlange.ca <http://www.johnlange.ca> <-- Gilles R. Detillieux E-mail: <grdetil@scrc.umanitoba.ca <mailto:grdetil@scrc.umanitoba.ca>> Spinal Cord Research Centre WWW:Dept. Physiology, U. of Manitoba Winnipeg, MB R3E 0J9 (Canada) _______________________________________________ Roundtable mailing list Roundtable@muug.mb.ca <mailto:Roundtable@muug.mb.ca> http://www.muug.mb.ca/mailman/listinfo/roundtable-- Sean Walberg <sean@ertw.com mailto:sean@ertw.com> http://ertw.com/
Roundtable mailing list Roundtable@muug.mb.ca http://www.muug.mb.ca/mailman/listinfo/roundtable
-- Gilles R. Detillieux E-mail: grdetil@scrc.umanitoba.ca Spinal Cord Research Centre WWW: http://www.scrc.umanitoba.ca/ Dept. Physiology, U. of Manitoba Winnipeg, MB R3E 0J9 (Canada) _______________________________________________ Roundtable mailing list Roundtable@muug.mb.ca http://www.muug.mb.ca/mailman/listinfo/roundtable