How would I count the total number of lines present in all the files in a git repository?
git ls-files
gives me a list of files tracked by git.
I'm looking for a command to cat
all those files. Something like
git ls-files | [cat all these files] | wc -l
xargs
will let you cat
all the files together before passing them to wc
, like you asked:
git ls-files | xargs cat | wc -l
But skipping the intermediate cat
gives you more information and is probably better:
git ls-files | xargs wc -l
git diff --stat 4b825dc642cb6eb9a060e54bf8d69288fbee4904
This shows the differences from the empty tree to your current working tree. Which happens to count all lines in your current working tree.
To get the numbers in your current working tree, do this:
git diff --shortstat `git hash-object -t tree /dev/null`
It will give you a string like 1770 files changed, 166776 insertions(+)
.
git hash-object -t tree /dev/null
.
git diff --stat `git hash-object -t tree /dev/null`
git diff --stat `git hash-object -t tree /dev/null` | tail -1
git diff --shortstat `git hash-object -t tree /dev/null`
to get the last line, tail isnt needed.
If you want this count because you want to get an idea of the project’s scope, you may prefer the output of CLOC (“Count Lines of Code”), which gives you a breakdown of significant and insignificant lines of code by language.
cloc $(git ls-files)
(This line is equivalent to git ls-files | xargs cloc
. It uses sh
’s $()
command substitution feature.)
Sample output:
20 text files.
20 unique files.
6 files ignored.
http://cloc.sourceforge.net v 1.62 T=0.22 s (62.5 files/s, 2771.2 lines/s)
-------------------------------------------------------------------------------
Language files blank comment code
-------------------------------------------------------------------------------
Javascript 2 13 111 309
JSON 3 0 0 58
HTML 2 7 12 50
Handlebars 2 0 0 37
CoffeeScript 4 1 4 12
SASS 1 1 1 5
-------------------------------------------------------------------------------
SUM: 14 22 128 471
-------------------------------------------------------------------------------
You will have to install CLOC first. You can probably install cloc
with your package manager – for example, brew install cloc
with Homebrew.
cloc $(git ls-files)
is often an improvement over cloc .
. For example, the above sample output with git ls-files
reports 471 lines of code. For the same project, cloc .
reports a whopping 456,279 lines (and takes six minutes to run), because it searches the dependencies in the Git-ignored node_modules
folder.
cloc --vcs git
these days, which avoids some edge cases with badly named files (or too many of them).
cloc
counts lines of files in a local directory, without ever accessing the network. It doesn’t even know whether the code came from GitHub or not.
I've encountered batching problems with git ls-files | xargs wc -l
when dealing with large numbers of files, where the line counts will get chunked out into multiple total
lines.
Taking a tip from question Why does the wc utility generate multiple lines with "total"?, I've found the following command to bypass the issue:
wc -l $(git ls-files)
Or if you want to only examine some files, e.g. code:
wc -l $(git ls-files | grep '.*\.cs')
wc -l $(git ls-files | find *.m *.h)
wc -l --files0-from=<(git ls-files -z)
. The <(COMMAND)
syntax returns the name of a file whose contents are the result of COMMAND
.
-n
switch with xargs
can be used to increase the maximum number of lines within a chunk
The best solution, to me anyway, is buried in the comments of @ephemient's answer. I am just pulling it up here so that it doesn't go unnoticed. The credit for this should go to @FRoZeN (and @ephemient).
git diff --shortstat `git hash-object -t tree /dev/null`
returns the total of files and lines in the working directory of a repo, without any additional noise. As a bonus, only the source code is counted - binary files are excluded from the tally.
The command above works on Linux and OS X. The cross-platform version of it is
git diff --shortstat 4b825dc642cb6eb9a060e54bf8d69288fbee4904
That works on Windows, too.
For the record, the options for excluding blank lines,
-w/--ignore-all-space,
-b/--ignore-space-change,
--ignore-blank-lines,
--ignore-space-at-eol
don't have any effect when used with --shortstat
. Blank lines are counted.
git mktree </dev/null
or true|git mktree
or git mktree <&-
or :|git mktree
for the keystroke-counters among us :-) - a spare empty tree floating around the repo isn't going to hurt anything.
This works as of cloc 1.68:
cloc --vcs=git
--vcs
didn't work for me, maybe it was removed. cloc .
while at the git repo did work, OTOH.
--vcs=git
worked for me on version v1.90 =) But yes I ran it at the root, it's just an option to tell cloc what it can ignore
I use the following:
git grep ^ | wc -l
This searches all files versioned by git for the regex ^
, which represents the beginning of a line, so this command gives the total number of lines!
I was playing around with cmder (http://gooseberrycreative.com/cmder/) and I wanted to count the lines of html,css,java and javascript. While some of the answers above worked, or
pattern in grep didn't - I found here (https://unix.stackexchange.com/questions/37313/how-do-i-grep-for-multiple-patterns) that I had to escape it
So this is what I use now:
git ls-files | grep "\(.html\|.css\|.js\|.java\)$" | xargs wc -l
I did this:
git ls-files | xargs file | grep "ASCII" | cut -d : -f 1 | xargs wc -l
this works if you count all text files in the repository as the files of interest. If some are considered documentation, etc, an exclusion filter can be added.
Try:
find . -type f -name '*.*' -exec wc -l {} +
on the directory/directories in question
If you want to get the number of lines from a certain author, try the following code:
git ls-files "*.java" | xargs -I{} git blame {} | grep ${your_name} | wc -l
This tool on github https://github.com/flosse/sloc can give the output in more descriptive way. It will Create stats of your source code:
physical lines
lines of code (source)
lines with comments
single-line comments
lines with block comments
lines mixed up with source and comments
empty lines
Depending on whether or not you want to include binary files, there are two solutions.
git grep --cached -al '' | xargs -P 4 cat | wc -l git grep --cached -Il '' | xargs -P 4 cat | wc -l "xargs -P 4" means it can read the files using four parallel processes. This can be really helpful if you are scanning very large repositories. Depending on capacity of the machine you may increase number of processes. -a, process binary files as text (Include Binary) -l '', show only filenames instead of matching lines (Scan only non empty files) -I, don't match patterns in binary files (Exclude Binary) --cached, search in index instead of in the work tree (Include uncommitted files)
The answer by Carl Norum assumes there are no files with spaces, one of the characters of IFS
with the others being tab
and newline
. The solution would be to terminate the line with a NULL byte.
git ls-files -z | xargs -0 cat | wc -l
If you want to find the total number of non-empty lines, you could use AWK:
git ls-files | xargs cat | awk '/\S/{x++} END{print "Total number of non-empty lines:", x}'
This uses regex to count the lines containing a non-whitespace character.
: | git mktree | git diff --shortstat --stdin
Or:
git ls-tree @ | sed '1i\\' | git mktree --batch | xargs | git diff-tree --shortstat --stdin
Success story sharing
grep cpp |
in there before thexargs
, then.git ls-files -z | xargs -0 wc -l
if you have files with spaces in the name.git ls-files | grep -P ".*(hpp|cpp)" | xargs wc -l
where the grep part is any perl regex you want!git ls-files | grep "\.java$" | xargs wc -l