Figure out the core data structures and do this at different level of abstraction. I'm currently reading through bits of the TensorFlow codebase, and starting with TF's GraphDef[0] was a good place to start. This is great because it's high-level enough that I can understand how TF models are structured on-disk and I can follow the deeper threads (e.g. how does device acceleration work) if needed.

https://github.com/tensorflow/tensorflow/blob/master/tensorf...

1. If the original authors are still around, we do a knowledge transfer. 2. Create diagrams about the relationships between all the systems involved. UMLs, ERDs, and all the rest of them 3. Clone the system locally and trace how the data flows from one point to the next. I'd go with logins first since we all know how it should work so it's easy to follow. 4. Lots and lots of trial and error when contributing new features. If it's well made, tests will catch your bugs. If you have a good team, they'll cut you some slack since you're just new and they were also new, too.

Go deep. Find something interesting and pull on that thread and keep going until you figure something out. If you do/don't get that far, start changing some things and see how it (mis)behaves, iterate.

The same way you eat an elephant: one bite at a time.

I don't think so. If you want to eat an elephant, eat it in group. :-)

Check this: https://mitchellh.com/writing/contributing-to-complex-projec...

Talk to the people who understand it well. Unless it is extremely well documented that is the only sane method. If you can get a class diagram it will help.