understand -w option with grep

geeyathink · 02-20-2008, 09:02 AM

I had wanted to include some abbreviations in my searches. cu, al and maybe a few more. I have not been able to get them to match only whole words ie "cu" and "al" They seem to match "cut" "all" "alabama' "cucumber" etc..

Is it correct that the -w option with grep would cause the pattern "cu" to not return a match on "cut".
To use the -w option with grep do I just include it along with any other options at the beginning?
I did not see in man where '-woic -f patternfile file' should conflict with one another.

How far off base am I this time?

I can include code or more context if needed. My script is getting big. I would only post what I hope is the appropriate section if that would do.

Thanks very much,

geeyathink · 02-20-2008, 11:03 AM

I had a few minutes to mess with this at a command line

If I put the word I expect to return "garden" then follow it by the other words that should not be in the file "al" "cu"

Code:

[teabear@junker ~]$ grep -iow -e "garden" -e "cu" -e "al"  /home/teabear/desktop/ok/tster
Garden
Garden
Garden
Garden
Garden
Garden
Garden
Garden
Garden
al

It returns more like I thought it would

But if I reverse the order of the patterns

Code:

[teabear@junker ~]$ grep -iow -e "cu" -e "al" -e "garden" /home/teabear/desktop/ok/tster
al
al
al
Garden
Garden
al
Garden
cu
Garden
Garden
Garden
al
al

It returned a load of the "al" and "cu"

If I seach without the word I was expexting to return

Code:

[teabear@junker ~]$ grep -iow -e "cu" -e "al"  /home/teabear/desktop/ok/tster
[teabear@junker ~]$

It returns nothing at all

tster is a html file about gardening, I just changed the long name it had to tster.

I get very similar results with a plain text file.

I fear it may be time for me to understand regular expressions. Hopefully not yet, just figuring how to use
these plain commands would be great for now if possible.

pixellany · 02-20-2008, 01:04 PM

I don't think there should be any issue combining those options.

Note:
Instead of:
grep -e cu -e al filename (the quotes are not needed)

why not:
egrep 'cu|al' filename

?

geeyathink · 02-20-2008, 04:03 PM

I think I created a small html file with kompoze, to have better control over testing this out.

It only has seven words in it:

linux computer network
cucumber dual almost cups

I got the same behaviour as in the above post when I changed to egrep 'pattern|pattern|pattern' etc...

Anything else anyone thinks for me to look into is great

Thank you,

sundialsvcs · 02-20-2008, 09:13 PM

The essential purpose of "-w" is to say that, in order to "match," the specified pattern must be both preceded and followed by a non-word character (or the beginning/end of line).

I don't use "-w" in any situation more complicated than that.

geeyathink · 02-21-2008, 04:17 AM

Thanks sundialsvcs,

Yes that is I believe all I want out of it

here is output of that seven word file with -ow options

Code:

[teabear@junker ~]$ egrep -ow  'networ|computer|linux|cu|al'  /home/teabear/desktop/tsthtml
linux
computer
networ
al
al
cu
cu
cu

here it is with only the -w option

Code:

[teabear@junker ~]$ egrep -w  'networ|computer|linux|cu|al'  /home/teabear/desktop/tsthtml
<html><head><meta content="text/html; charset=ISO-8859-1" http-equiv="content-type"><title>mmmam</title></head><body>linux computer network<br><br>dual almost cucumber cups</body></html>
[teabear@junker ~]$

The color is missing but in the second output the searched terms are highlighted. Maybe the html stuff is is interfering?

I a going to go double check with plain text file

Yes same behaviour with a text file,

Code:

[teabear@junker ~]$ egrep -w  'computer|linux|cu|al'  /home/teabear/desktop/testtext
linux computer aluminum cucumber
aluminum cucumber linux computer
[teabear@junker ~]$

Code:

[teabear@junker ~]$ egrep -wo  'computer|linux|cu|al'  /home/teabear/desktop/testtext
linux
computer
al
cu
cu
al
cu
cu
linux
computer
[teabear@junker ~]$

so it isn't just the html format thats messing me up

What could I have screwed up to be causing this? Some kind of envirioment
setting?

Thanks again to all,

I reinstalled only my / and still had the problem, tried a while longer to figure out the prob, gave up and then reinstalled /home and now the problem is gone.

I suspect that during my experimenting with grep,sed,awk etc... I accidently screwed up a config file that bash looks at. Just a guess. Don't put much stock in it. I have yet to be correct about anything System or Programming related.