Recursively count lines of code in source dir heirarchy

michaelsanford · 03-31-2006, 08:03 PM

I've done a bit of reading on this, but my shell scripting leaves to be desired. I want to be able to output the total number of lines in my entire project, which contains sub-folders. I got this far:

Code:

cat `ls -R1 | grep -v ./ | grep -v .gif | grep -v .jpg`

Passing this to `wc -l', however, doesn't work. In fact, this example is broken at `cat' because I can't figure out how to prepend the full path to the file names (or use any other hack) and I get results like this:

Code:

cat: fck_contextmenu.css: No such file or directory
cat: fck_dialog.css: No such file or directory
cat: fck_editor.css: No such file or directory

simply because those files reside in sub-directories of the cwd but are `ls'ed as ./<file> for all `cat' knows.

Any ideas around this? I'm sure it's quite simple but I can't seem to catch it.

aluser · 03-31-2006, 08:53 PM

I usually do something like

Code:

find . \! -iname \*.gif \! -iname \*.jpg -type f -print | xargs wc -l

There are of course several other ways to skin the same cat.

aluser · 03-31-2006, 08:56 PM

oh yeah, if your project is very large, you want something like

Code:

find . \! ... -print | xargs cat | wc -l

Reason being that, if xargs sees what would be "too long" a command line for your system, it invokes the command you give it several times

95se · 03-31-2006, 09:02 PM

Wrote a short script for you. It uses file instead of grepping, to check if it's a text or ascii file (source files will show up as like, ASCII C++ file text)

Code:

#!/bin/bash

wc_total=0

function cnt {
    for file in `ls $1`; do
	if [ -d $file ]; then
	    cnt $file
	else
	    if [ "`file -b $file | grep -i '\(ascii\|text\)'`" != "" ]; then
		let "wc_total += `cat $file | wc -l`"
	    fi
	fi
    done
	
}

cnt .
echo $wc_total
exit 0

michaelsanford · 03-31-2006, 09:08 PM

Thank you both! This is a little tidbit I've wanted to add to my reports (just as an FYI) for quite some time. Very much appreciated.

95se · 04-01-2006, 10:35 AM

I just looked at and realized it may be missing one thing, use this one instead...

Code:

#!/bin/bash

wc_total=0

function cnt {
    cd $1
    for file in `ls`; do
	if [ -d $file ]; then
	    cnt $file
	else
	    if [ "`file $file | grep -i '\(ascii\|text\)'`" != "" ]; then
		let "wc_total += `cat $file | wc -l`"
	    fi
	fi
    done
    cd ..
}

cnt .
echo $wc_total
exit 0

ta0kira · 04-01-2006, 01:56 PM

I don't think it needs to be that complex; maybe this:

Code:

#!/bin/bash

#'count-lines' script

#put each pattern on it's own line
patterns="`echo "$*" | sed "s/\[ \]+/\n/g"`"

#add the '-name' option to each line
patterns="`echo "$patterns" | sed "s/^/-name /g"`"

#add the '-o' option to all but the last line
patterns="`echo "$patterns" | sed "s/$/ -o/g"`"

cat `find -type f $patterns` | grep -c ".*"

Call the script providing patterns on the command line:

Code:

> count-lines "*.cpp *.hpp *.h"

Sorry, didn't get a chance to try it, but I've used the last line of the script providing patterns "manually". The 'sed' stuff should turn the above patterns into (excluding newlines):

Code:

-name *.cpp -o -name *.hpp -o -name *.h

ta0kira

addy86 · 04-01-2006, 04:51 PM

Have you tried sloccount? (http://www.dwheeler.com/sloccount/)

michaelsanford · 04-01-2006, 04:59 PM

Thank you for the updates and additional scripts, and I will check out sloccount.

I love this forum.