Matching RegExps using grep

[Tsuki]

Hi, I'm probably being really dumb here, but can anyone see why this doesn't work?

I've got a bunch of text (curled from a forum that doesn't support notifications / syndication). Somewhere in the middle is text like so:

09:32, Wed 25 Jul by Tsuki
That's surrounded by a bunch of HTML, and I'm trying to extract it by piping it into grep. Here's my code:

egrep --only-matching --regexp=[0-9]+:[0-9]+,\\s\\w+\\s\\w+\\s\\w+\\sby\\s\\w+

Any ideas?

[stylishpants]

You're using \w and \s a lot. These are only available when you're using perl-compatible regular expressions. egrep uses POSIX regular expressions.

You could accomplish this with perl:


perl -ne 'print if /[0-9]+:[0-9]+,\s\w+\s\w+\s\w+\sby\s\w+/'

Or rewrite your previous regexp to use '[A-Za-z0-9_]' in place of '\w' and a bunch of whitespace chars inside a character class instead of '\s'.

Another option is to install and use pcregrep (it's in the repos.)


fhanson@fh:~$ echo "09:32, Wed 25 Jul by Tsuki" | pcregrep -o '\d+:\d+,\s\w+\s\w+\s\w+\sby\s\w+'
09:32, Wed 25 Jul by Tsuki

Also, unless you hate yourself, surround your non-trivial regular expressions with single quotes so that they won't get escaped by the shell. Regular expressions are tricky enough to parse visually without also having to consider shell escaping.



# Harder to read:
egrep --only-matching --regexp=[0-9]+:[0-9]+,\\s\\w+\\s\\w+\\s\\w+\\sby\\s\\w+

# Easier to read: (still won't work though due to the problem mentioned above)
egrep -o '[0-9]+:[0-9]+,\s\w+\s\w+\s\w+\sby\s\w+'

[Tsuki]

Thank you! I didn't realise \s and \w weren't POSIX-compatible. My script's now working!
(Thanks for the hint about the quotes, too. Not sure why I didn't think of that...)