Grep Recursive and Count

问题:

Need to search a directories with lots of sub-directories for a string inside files:

I'm using:

grep -c -r "string here" *

How can I total count of finds?

How can I output to file only those files with at least one instance?

回答1:

It works for me (it gets the total number of 'string here' found in each file). However, it does not display the total for ALL files searched. Here is how you can get it:

grep -c -r 'string' file > out && \
    awk -F : '{total += $2} END { print "Total:", total }' out

The list will be in out and the total will be sent to STDOUT.

Here is the output on the Python2.5.4 directory tree:

grep -c -r 'import' Python-2.5.4/ > out && \
    awk -F : '{total += $2} END { print "Total:", total }' out
Total: 11500

$ head out
Python-2.5.4/Python/import.c:155
Python-2.5.4/Python/thread.o:0
Python-2.5.4/Python/pyarena.c:0
Python-2.5.4/Python/getargs.c:0
Python-2.5.4/Python/thread_solaris.h:0
Python-2.5.4/Python/dup2.c:0
Python-2.5.4/Python/getplatform.c:0
Python-2.5.4/Python/frozenmain.c:0
Python-2.5.4/Python/pyfpe.c:0
Python-2.5.4/Python/getmtime.c:0

If you just want to get lines with occurrences of 'string', change to this:

grep -c -r 'import' Python-2.5.4/ | \
    awk -F : '{total += $2; print $1, $2} END { print "Total:", total }'

That will output:

[... snipped]
Python-2.5.4/Lib/dis.py 4
Python-2.5.4/Lib/mhlib.py 10
Python-2.5.4/Lib/decimal.py 8
Python-2.5.4/Lib/new.py 6
Python-2.5.4/Lib/stringold.py 3
Total: 11500

You can change how the files ($1) and the count per file ($2) is printed.

回答2:

Using Bash's process substitution, this gives what I believe is the output you want? (Please clarify the question if it's not.)

grep -r "string here" * | tee >(wc -l)

This runs grep -r normally, with output going both to stdout and to a wc -l process.

回答3:

Some solution with AWK:

grep -r "string here" * | awk 'END { print NR } 1'

Next one is total count, number of files, and number of matches for each, displaying the first match of each one (to display all matches, change the condition to ++f[$1]):

grep -r "string here" * | 
    awk -F: 'END { print "\nmatches: ", NR, "files: ", length(f); 
                   for (i in f) print i, f[i] } !f[$1]++'

Output for the first solution (searching within a directory for "boost::". I manually cut some too long lines so they fit horizontally):

list_inserter.hpp:            return range( boost::begin(r), boost::end(r) );
list_of.hpp:            ::boost::is_array<T>,
list_of.hpp:            ::boost::decay<const T>,
list_of.hpp:            ::boost::decay<T> >::type type;
list_of.hpp:        return ::boost::iterator_range_detail::equal( l, r );
list_of.hpp:        return ::boost::iterator_range_detail::less_than( l, r );
list_of.hpp:        return ::boost::iterator_range_detail::less_than( l, r );
list_of.hpp:        return Os << ::boost::make_iterator_range( r.begin(), r.end() );
list_of.hpp:            return range( boost::begin(r), boost::end(r) );
list_of.hpp:            return range( boost::begin(r), boost::end(r) );
list_of.hpp:            return range( boost::begin(r), boost::end(r) );
ptr_list_of.hpp:                          BOOST_DEDUCED_TYPENAME boost::ptr_...
ptr_list_of.hpp:        typedef boost::ptr_vector<T>       impl_type;
13

Output for the second one

list_inserter.hpp:            return range( boost::begin(r), boost::end(r) );
list_of.hpp:            ::boost::is_array<T>,
ptr_list_of.hpp:                          BOOST_DEDUCED_TYPENAME boost::ptr_...

matches:  13 files:  3
ptr_list_of.hpp 2
list_of.hpp 10
list_inserter.hpp 1

Colors in the result are nice (--color=always for grep), but they break when piped through awk here. So better don't enable them then unless you want to have all your terminal colored afterwards :) Cheers!

回答4:

I would try a combination of find and grep.

find . | xargs grep -c "string here"

Anyway, grep -c -r "string here" * works for me (Mac OS X).

回答5:

grep -rc "my string" ./ | grep :[1-9] >> file_name_by_count.txt

Works like a charm.

回答6:

To output only file names with matches, use:

grep -r -l "your string here" .

It will output one line with the filename for each file which matches the expression searched for.