Using find and grep

[scm]

Someone recently posted:

Here is what works for me, to use grep on only those files found by find:

find /usr/src/ -name '*.h' | while read FILE
do
    if grep 'integer_' $FILE
    then
          echo $FILE
   fi  
done

The simplest way to do this is:

find /usr/src -name '*.h' | xargs grep 'integer_'

[i92guboj]

Someone recently posted:

The simplest way to do this is:

find /usr/src -name '*.h' | xargs grep 'integer_'

Not really, this is simpler because you don't need a pipe:

find /usr/src -name '*.h' -exec grep 'integer_' '{}' \;

But to tell the truth, you don't even need find:

grep -r 'integer_' /usr/src/

[scm]

Not really, this is simpler because you don't need a pipe:

find /usr/src -name '*.h' -exec grep 'integer_' '{}' \;

The disadvantages with this approach are that it has to invoke grep for each filename, so performance may be an issue, and also since you're only grepping a single file at a time, matching lines will be shown without a filename, so not very useful. Of course, you could overcome the latter with the /dev/null trick, but you'll still require grep to be exec-ed for each file.

But to tell the truth, you don't even need find:

grep -r 'integer_' /usr/src/

True, if you're using GNU grep, but if you're trying to write portable scripts you'd do well to consider the versions of grep you're likely to need to accommodate on some quite elderly UNIX systems.

[ghostdog74]

Someone recently posted:

The simplest way to do this is:

find /usr/src -name '*.h' | xargs grep 'integer_'

i am a bit lazy to try this out, say if there is too many *.h files, will it produce some "argument too long" blah blah errors. Just a thought.

[i92guboj]

The disadvantages with this approach are that it has to invoke grep for each filename, so performance may be an issue, and also since you're only grepping a single file at a time, matching lines will be shown without a filename, so not very useful. Of course, you could overcome the latter with the /dev/null trick, but you'll still require grep to be exec-ed for each file.

Well, that's true.

True, if you're using GNU grep, but if you're trying to write portable scripts you'd do well to consider the versions of grep you're likely to need to accommodate on some quite elderly UNIX systems.

Yes, but since this is mostly a linux forum, we can assume that all people around here use gnu stuff. But you indeed have the point: if portability is an issue, we should not use obscure features.

i am a bit lazy to try this out, say if there is too many *.h files, will it produce some "argument too long" blah blah errors. Just a thought.

Yes. and that's why the other thread suggests using the while construct:

find .... | while read file

This will work always, as long as everything is correctly written and quoted. Which is why this way is preferable sometimes.

[scm]

i am a bit lazy to try this out, say if there is too many *.h files, will it produce some "argument too long" blah blah errors. Just a thought.

Not a very good thought, however - xargs was primarily invented precisely to overcome the early limitations of the shell environment so unless GNU xargs is broken it will always make the argument list fit within the environment space. If I could guarantee that the find list would fit the environment I'd have done

grep 'integer_' $(find /usr/src -name '*.h')