I need to extract the rusage mem= values from gibberish similar to the following:

0]rusage[DS=1:mem=5000]>;MonNov2415:33:54:[49]startedon<somecomputer@somewhere>,Executionm]rusage[mem=10000]>;TueNov1807:03:24:Startedon<somecomputer@somewhere>,ExecutionHomeces<select[linux&&mem>10000&&maxmem32G]rusage[mem=10000]span[hosts=1]>;

In the above example this would be:
5000
10000
10000

Keep in mind that this information will not always be located in the same place in the data stream, if it were I would just use awk.

The normal output of this data is not all scrunched, but I figured it might be easier to parse than something wrapped, broken, and like this:

blah blah blah blah blah rusage[me
m=123456] blah blah blah





Thank you.

Joel

What have you tried so far? What does a regexp such as /mem=\d+/ find?

Please forgive my ignorance, but exactly how would I implement that regular expression. I am sure I could learn sed really well with enough time, but right now it still seems very cryptic.

If test.txt contains
then
gives
(I'm trying to improve my own sed'ing)

This works very well!!! Do you have and Sed or Regular Expression books or links you recommend?

Joel

Try this http://www.ibm.com/developerworks/li...ry/l-sed1.html

Do you always know how many matches to expect ??. Your (latest) reply indicates the "rusage" reference in your initial post is irrelevant. Needing that as preceding data would have made things messy.
I would be inclined to use something like this

That's because it IS cryptic......


When learning SED, start simple, eg:

sed 's/old/new/' filename ##replace all "old" with "new"

sed -n '1p' filename ##print just line 1

sed -n '1,3p' filename ##print lines 1 thru 3

sed -n '/Tom/p' filename ##print all lines containing "Tom"

http://www.grymoire.com/Unix/