noob sed question

[stevelorimer]

Hi folks

Please forgive how lame this question must seem to you pros, but I'm a bit stuck here...

On my machine, calling this in a shell:

g++ --version

outputs the following:

g++ (SUSE Linux) 4.4.1 [gcc-4_4-branch revision 150839]
Copyright (C) 2009 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Please can someone help with the sed and/or other shell commands which will be able to parse out the g++ version number. ie, I'm looking for the output from sed to be the following:

4.4.1

At the moment I'm using this command:

g++ --version | sed -e 's/[a-Z+()$/[,;]*//g'

and getting the following:

   4.4.1 -4_4-  150839]
  2009    .
         .
          .

which I guess is kinda half-way there... I'm sure there's an easy to way to strip out the leading whitespace, and then keep only the items up until the first whitespace character found... but at this point my limited knowledge of sed fails miserably!

TIA
Steve

[Irithori]

Different approach, but maybe ok-ish for you?

You could query the rpm DB

rpm -q --queryformat '%{VERSION}\n' gcc

With gcc being the name of the (installed) rpm.
gcc is gcc´s name on CentOS, havent seen a SuSE in a while, so ymmv

[stevelorimer]

Thanks Irithori, that may just work. Gives back 4.4 instead of 4.4.1, but I think that won't be an issue.

I'd still like to know the sed command though, coz no doubt there will be a million more uses for knowing that in the future?!

[Franklin52]

Thanks Irithori, that may just work. Gives back 4.4 instead of 4.4.1, but I think that won't be an issue.

I'd still like to know the sed command though, coz no doubt there will be a million more uses for knowing that in the future?!

Try this:

g++ --version | sed -n '/g++/ s/.*) \(.*\) \[.*/\1/p'

or shorter with awk:

g++ --version | awk '/g++/{print $4}'

[lemons]

Thanks Irithori, that may just work. Gives back 4.4 instead of 4.4.1, but I think that won't be an issue...

For me, the command mentioned by Irithori seems to return:

[adam@computer ~]$ rpm -q --queryformat '%{VERSION}\n' gcc
4.4.4

(on Fedora 12.)

[stevelorimer]

Thanks Franklin52 - I'd say that the sed one works best for me - results from all 3 possible solutions below.

The problem with the awk one is that I get back the 4th word in each line, of which there are 3. Is there a way to get awk to return only the 1st line?

slorimer@slorimer:~> g++ --version | awk '/g++/{print $4}'
4.4.1
Free
software;

lemons, the fact that you're getting back major.minor.revision whereas I'm only getting major.minor means that using the rpm query is not portable

slorimer@slorimer:~> rpm -q --queryformat '%{VERSION}\n' gcc
4.4

So this is the one that works best. Franklin52, please can you talk me through each of the tokens in the sed command? I mean, come on! It looks like some alien language!

slorimer@slorimer:~> g++ --version | sed -n '/g++/ s/.*) \(.*\) \[.*/\1/p'
4.4.1

On another forum I was given this command, which is quite elegant:

slorimer@slorimer:~> g++ --version | grep '[0-9]\.[0-9]\.[0-9]' -o
4.4.1

Thanks for all the help folks, it's much appreciated!
Steve

[srikanth007m]

If awk is ok for you.. Then you can try this

$ gcc --version | grep 4 | awk '{print $3}'
4.2.4

[stevelorimer]

That assumes there will always be a 4 in gcc version - which is most definitely not true.

[jhuizer]

Given this output for `g++ --version`:

g++ (SUSE Linux) 4.4.1 [gcc-4_4-branch revision 150839]
Copyright (C) 2009 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

g++ --version | sed -n '/g++/ s/.*) \(.*\) \[.*/\1/p'

- the -n flag tells sed not to print anything unmatched
- line(s) containing 'g++' are matched

The harder part is s/.*) \(.*\) \[.*/\1/p'
- '.' in a sed expression means any character (exactly one)
- '*' means, repeating the previous pattern, zero or more times
- '.*' means, a series of characters of arbitrary length (not otherwise specified)
- \( ... \) is a subexpression - it is referenced by \1 (you can have up to 9 of these, referenced by \1 ... \9 )
- '[' is a special character (for ranges, such as [1-9]) - to have it treated as the actual character, it has to be "escaped", written as "\["

From left to write:
- it matches anything before a ')' in .*)
- then a space
- then a series of characters (of arbitrary length) as the subexpression (intended to be the actual version string!!)
- [ and the rest of the line is matched

then the subexpression is printed

[jhuizer]

By the way the sed line doesn't work in general; my `g++ --version` says,

g++ (Debian 4.3.2-1.1) 4.3.2
Copyright (C) 2008 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Nothing is matched, as the [ isn't in this output

I'd adapt the sed line to:

g++ --version |   sed -n '/g++/ s/.*) \([0-9]\.[0-9]\.[0-9]\).*/\1/p'

Hoping how that works shouldn't be too hard to decrypt now!