Contributing to Allura¶
For developers interested in hacking on Allura or its components, this guide will (hopefully) be a roadmap to help you get started and guide you on your way.
Along the way, you will no doubt have questions that aren’t addressed here. For help, you can get in touch with other Allura developers on the developer mailing list (firstname.lastname@example.org) or in the #allura channel on the Freenode IRC network.
Before hacking on Allura, you’ll need to get an Allura instance up and running so you can see and test the changes you make. You can install Allura from scratch, or by using our Docker container images. Instructions for these approaches can be found here:
Allura is comprised of a handful of separate services, all of which must be running in order for the system to be fully functional. These services (and how to start them) are covered in the install documentation, but are mentioned again here simply to reiterate the components of a complete Allura system.
- MongoDB - database
- Solr - searching/indexing
- Web server - the Allura web application
- Taskd - background task daemon
- Inbound email handler - processes email sent to the Allura instance (e.g., a reply to a ticket notification email)
The logs for Allura services can be found in /var/log/allura/. The most important of these is allura.log, as it will contain log messages for all Allura application code.
Solr - Allura artifact data is indexed into Solr so that it can be searched. In general, you won’t need to know much about Solr in order to work on Allura.
Turbogears - Allura is built on the TurboGears web framework. Understanding TG controller basics and object dispatch, TurboGears’ mechanism for routing an HTTP request to the code that should handle it, are critical for understanding how a request is handled by Allura.
Ming - Allura interfaces with MongoDB through Ming, a library which provides an Object Document Mapper for MongoDB. Fortunately, the query syntax is mostly identical to that of native MongoDB, so the learning curve is pretty flat.
EasyWidgets - An HTML template and form validation library used by Allura. The learning curve on EasyWidgets is, ironically, not easy. Be prepared to dig into the source if you want to do something complicated with EW. Fortunately, there are lots of exmaples in the Allura source already.
Jinja - HTML template library used by Allura.
~$ npm install -g broccoli-timepiece ~$ cd ~/src/allura ~$ npm run watch
Finding Something to Work On¶
Tickets that are relatively simple and good for new contributors have a “bitesize” label, and can be found here: https://forge-allura.apache.org/p/allura/tickets/search/?q=labels%3Abitesize+AND+status%3Aopen
Find one that looks good, and leave a comment on the ticket or mailing list to let us know you’re working on it. If you get stuck, remember that we’re available to help on the mailing list or IRC.
The core Allura platform code is in the Allura/ directory in the top-level of the repo. The Forge*/ directories contain Allura “tools” - plugins that extend the core platform. For an overview of the platform and services it provides, read the Platform Tour documentation. If you’re interested in developing a new Allura plugin, you may find this blog series helpful.
Tracing a Request¶
Whether you’re fixing a bug or adding a new feature, one of your first questions will be, “Where is the code that is handling this request (or serving this page)?” For a new contributor, answering this question can be surprisingly challenging. Here are some tips to help you out:
1. The root controller for the entire application is in Allura/allura/controllers/root.py - dispatch for every request begins here. It is possible (albeit difficult) to trace the path your request will take through the code from this starting point if you have a thorough knowledge of Turbogears’ request dispatch mechanics. But, nobody wants to do this if they can avoid it.
2. Is the page being served part of a tool (e.g. Ticket Tracker, Wiki, etc)? Most of the time, the answer is yes. If you know which tool is handling the request, you can skip right to the root controller for that tool. To find the root controller, first find the main entry point for the tool, which is defined in the [allura] section of the tool’s setup.py file. So, for example, if you know the request is being handled by a Ticket Tracker, look in ForgeTracker/setup.py and you’ll see that that its entry point is forgetracker.tracker_main:ForgeTrackerApp. Each Allura tool instance defines a root attribute which is its root controller. So once you’ve found the main tool class, you can find its root controller and begin tracing your request from there.
3. Search for things! grep and equivalents are your friends. If you’re looking at an html page and want to find the controller code for it, try searching the code base for some (static) text on the page. If your search successfully turns up an html page, search again on the name of the html file. There’s a good change you’ll find the controller method that renders that page.
If you’ve never used ipdb before, you’ll find it’s a great tool for interactive debugging of Python code. In order to use ipdb to debug Allura, you’ll first need to make sure that the process you’re debugging is running in the foreground. In most cases you’ll be debugging either the web app process or the taskd (background worker) process.
First, make sure sure ipdb is installed in your virtual environment:
pip install ipdb
Then, find the line of code where you want to start the interactive debugger, and insert this line above it:
import ipdb; ipdb.set_trace()
Now, kill any running web or taskd procs and restart them in the foreground:
cd Allura # web pkill -f gunicorn; gunicorn --reload --paste development.ini # taskd pkill "^taskd"; paster taskd development.ini --nocapture
Make a request to the web app, and when your line of code is hit, a debug session will start on the console where the process is running.
For more information about using pdb, see the official documentation.
To run all the tests, execute ./run_tests in the repo root. To run tests for a single package, for example forgetracker:
cd ForgeTracker && nosetests
To learn more about the nose test runner, consult the documentation.
When writing code for Allura, don’t forget that you’ll need to also create tests that cover behaviour that you’ve added or changed. You may find this short guide helpful.
Submitting a Merge Request¶
Before submitting a merge request, make sure your changes conform to our contribution guidelines. Once your changes are finished and tested, submit them to be merged back into the main repo:
- Fork the main Allura repo from here: https://forge-allura.apache.org/p/allura/git/
- Commit and push your changes to your fork
- Submit a Merge Request from your fork