Running “uncommitted”

By default “uncommitted” uses the locate(1) command to scan for repositories, which means that it can operate quickly even over very large filesystems like my home directory:

$ uncommitted ~

But you should be warned: because the locate(1) database is only updated once a day on most systems, this will miss repositories which you have created since its last run. To be absolutely sure to see all current repositories, you should instead ask “uncommitted” to search the filesystem tree itself. To do this on your “devel” directory, for example, you would type this:

$ uncommitted -w ~/devel

Not only will the output of “-w” always be up-to-date, but it is usually faster for small directory trees. The default behavior (which can also be explicitly requested, with “-l”) is faster when the directory tree you are searching is very large.

Should you ever want a list of all repositories, and not just those with uncommitted changes, you can use the “-v” verbose option:

$ uncommitted -v ~

You can always get help by running “uncommitted” without arguments or with the “-h” or “–help” options.

Supported VCs

At the moment, “uncommitted” supports:

• Mercurial (.hg directories)

• Subversion (.svn directories)

It needs to support Git soon, since that DVCS is in widespread use. However, I am not familiar enough with Git’s output to write a function for it myself. When I tried out Git a few minutes ago, its status messages were crazy, with all sorts of hash characters everywhere; it looked like it was suffering from an acute lack of confidence, and was therefore commenting out all of its own output. Anyway, there is probably some simple way to make Git report on uncommitted changes, but I will leave it to a Git fan to figure out how, if they would like to contribute a patch back.

I would also not be opposed to someone contributing a Bazaar plugin. But CVS should probably never be supported by “uncommitted” because that might imply that it is still an acceptible system to be using.

It occurs to me that there might already be some version control abstraction layer that I should be using for this, rather than figuring out how to run each version control system myself; a quick search of PyPI suggests that I take a closer look at the pyvcs project. Maybe that can be a useful direction for the next phase of development!