Ask HN: How do you read code?
Given a new project or an unfamiliar codebase, how do you go about reading and understanding the code?
5 comments
If you have access to a whiteboard...
List all of the directories (or classes, for a smaller project), and draw them as boxes on the board.
Draw arrows between the boxes based on dependencies in the codebase (helps if you have someone already familiar).
... Now start at the bottom of the dependency graph, and browse the code at a medium level.
If you are going to be working in this codebase, go back through and browse the code at deeper level, as if you were doing a code review. You can pretty much assume that, every couple lines, there is always something you would change about the code. Make a note of it.
At this point, you should have a fairly decent idea of how the code works, and how you would change it.
List all of the directories (or classes, for a smaller project), and draw them as boxes on the board.
Draw arrows between the boxes based on dependencies in the codebase (helps if you have someone already familiar).
... Now start at the bottom of the dependency graph, and browse the code at a medium level.
If you are going to be working in this codebase, go back through and browse the code at deeper level, as if you were doing a code review. You can pretty much assume that, every couple lines, there is always something you would change about the code. Make a note of it.
At this point, you should have a fairly decent idea of how the code works, and how you would change it.
I usually end up doing this, but not formally. I dive into the code and eventually find the bottom, and then work up. In the future, I think I'm going to try this and avoid wasting time wandering about trying to keep track mentally of everything as I go along. At the very least, I'll write it out in a notebook.
Thanks for the answer! It sounds like you're taking a bottom-up approach when reading code. Do you think we can use software to build this dependency graph and make it easier to read code?
Yes, technically it is possible. Some languages (and build tools) are a lot easier than others.
For purposes of learning the code, it may not be worth automating the dependency graph. You will probably learn more from tracking it down than you would if it was pre-generated for you.
For purposes of learning the code, it may not be worth automating the dependency graph. You will probably learn more from tracking it down than you would if it was pre-generated for you.
Well you should always ask someone who has worked with it but lately I had to come up quickly with a phone formatter class. I didn't have a clue how it worked, no comments. So I start to rewrite some of it, run it see what happens. Then when I learned what the stuff did I renamed it so it made sense to me. Used real long variable names which were meaningfull not just tmp and temp. Can really recommend it and it was very fast.
Thanks for your answer! Your approach sounds like it works well if you have to write a new component from scratch. What if you had to jump into an existing codebase? How would you go about understanding the code?
I look at what it does, do a quick architecture in my head, then compare it to what I see. This helps me because before I even look at it I have an understanding of how I think it should work rather than an absolutely clean slate.
What do you mean when you say compare it to what you see? How do you take the files and lines of code you read and model it into a working architecture? Where do you start? Do you think we could write a program that can make this process easier? (also, thanks for your answer!)
if you have a good debugger then start at the beginning and run it line by line. if a section looks understandable skip it.
Thanks for your answer sharemywin! What if you were looking at a large project that doesn't have a well-defined "beginning?" How would you pick what code to run through the debugger first?
From left-to-right