regexp filter a file

[adam-griffiths]

I would like to run a regexp across a large file and print out only the sub matches. I'm happy to use sed, awk, perl but i'm looking for the simpelest way - a one liner hopefully - to find all [HEADLINE](.*)[WORDS] and return only the sub match i.e. the (.*) part.

For example I would like to parse this:

[HEADLINE]
headline 1
[Words]34
[HEADLINE]
headline 2
[Words]66
[HEADLINE]
headline 3
[Words]55

to give this:

headline 1
headline 2
headline 3

I am also interested in general what is the quickest one liner to filter and return for all submatches, that may or may not me multiline, of a regexp in a text file.

Thanks

Adam

[Javasnob]

I don't know which would be quickest, but I would use sed.

[adam-griffiths]

Thanks for your reply - how would I do it in sed? I'm looking for something like:
match the regexp [HEADLINE]\n(.*)[WORDS].

and then print \1, i.e. the characters that were matched in the (.*) bit.

But I don't know how to get sed to do this, can anyone give me a suggested sed script?

[drl]

Hi.

I have used agrep for things like this. You might need to install it, but it's a useful utility to know about. Part of the man page is below.

You might consider csplit, which would create a file for each occurrence, but in some circumstances, that is actually the best part.

However, perl or sed may be your best bet. I've worked on and off on a multi-line matching script in perl, but it is definitely not a one-liner.

Let us know what your solution turns out to be ... cheers, drl

       agrep -d '^From ' 'breakdown;internet' mbox
              outputs  all  mail messages (the pattern '^From ' separates mail
              messages in a mail file) that contain keywords  'breakdown'  and
              'internet'.
-- excerpt from man agrep

[muha]

Substitute foo with bar on lines between RE1 and RE2 but not ON lines containing RE1 and RE2. ~>

sed '/RE1/,/RE2/{;/RE1/b;/RE2/b;s/foo/bar/;}' file

Substitute, anything between [HEADLINE] and [Words] with anything between [HEADLINE] and [Words] , on lines between [HEADLINE] and [Words] but not ON lines containing [HEADLINE] and [Words]. ~>
So basicly, print anything between [HEADLINE] and [Words], last p stands for print.
Don't print the rest, that's the -n option.

$ cat infile.txt
[HEADLINE]
headline 1
[Words]34
[HEADLINE]
headline 2
[Words]66
[HEADLINE]
headline 3
[Words]55
$ sed -n '/\[HEADLINE\]/,/\[Words\]/{;/\[HEADLINE\]/b;/\[Words\]/b;s/.*/&/p;}' infile.txt
headline 1
headline 2
headline 3