A Python based dashboard for the Rensselaer Center for Open Source Software

Observatory is intended to track a group of open source projects across their source code repositories and blogs. It also includes links to the websites and wikis each project, but doesn't scan those for any updates.

Observatory can directly access repositories (for git, hg, and svn), so the "branches don't work" problem has been eliminated. With this also comes a full redesign. Individual commits are visible in detail, and each project has its own page, which shows authors, smaller contributors, blog posts, commits, and screenshots. Additionally, there is a "feed" view that shows all of the most recent actions.

Observatory is pretty easy to set up. Excluding the Apache and mod_wsgi setup that you would want for a production environment (resources on which are best found elsewhere, most of the dependencies) are packed in the observatory/lib directory.

The following additional packages are required:

• Python Imaging Library (PIL)
• BeautifulSoup

As well as these version control systems:

• git (Subversion is via git svn)
• Mercurial

Everything except git can be acquired from PyPI:

$ pip install mercurial BeautifulSoup pil

Of course, you can also use your distribution's package manager if that's your sort of thing.

Since it deals primarily with external data sources, Observatory uses a set of scripts to pull data down and store it in the local database. There scripts are located in the observatory/dashboard/fetch directory. There are two primary scripts, fetch_blogs.py and fetch_repositories.py, the purpose of each being fairly obvious. For development purposes, there is also a fetch.py script (located up a directory from the others) which runs both fetch scripts.

The primary reason for splitting up the fetching process into two scripts is that previously, the update process was somewhat unpleasant for users: either click the update link (which didn't always work), or wait up to an hour. While blogs update infrequently, commits (should) happen much more frequently. Therefore, it is ideal to run the repository fetching script more often than the blog fetching one. Both scripts tend to run fairly quickly (with the exception of the initial import).

As threads and Python don't get along so well, fetching instead uses multiple processes (via the subprocess module). The amount of processes can be configured in the settings.py file. The usual rules about processes per CPU core don't necessarily apply here, because a lot of the time (especially for blog fetching) is spent waiting for servers to respond.

If you would like to help out with Observatory, that is probably a good thing. You can either use the fork + pull request feature of Github, or just send in patches (use Issues or email).

To set up a development environment is very easy. Just follow the instructions in the "Setup" above, but skip the part about configuring a web server and database, since you can use the Django development server and SQLite.

There are two scripts that are useful:

• observatory/dashboard/demo.py loads up a demo set of projects and users.
• observatory/dashboard/fetch.py runs the repository and blog fetch scripts.

Once both of these have been run, the database will be populated with the most recent commits and blog posts for the demo projects. Then, you can just run the server with ./manage.py runserver in the observatory directory, and access it (typically) at http://localhost:8000/projects.

If you're looking for something to work on, check out the Issues page or ask me, via email or on #rcos on irc.freenode.net.