I’m often asked about comments in code: when to do it, how to do it, what to put, etc. I’ve recently run into Steve’s post about when to comment your code, left a comment (!) there, and we got to expand our conversation in his podcast/screencast (link at the bottom). I’ve decided to create this post to consolidate the links and info shared during the interview, to make it easy for folks to find the material.
I remembered writing blog posts a couple of times over the years and it’s interesting to see how my opinion on this subject has changed over time.
The first post goes all the way back to 2005, with me asking “can you plesae put some comment on that Regular Expression?”. 15 years later, I still ask my smart friends to get me the RegEx I need, along some comments as to what each piece does!
In 2007, I was big into using Xml Comments, GhostDoct, and Documentor… I’m not anymore, as documented 10 years later with my post “XmlDoc Comments: Auto Generate and Hide the Clutter”. In nutshell, if we’re documenting a public API, yes, by all means let’s put in that documentation, but making it count: commenting the GetCustomers endpoint with “Gets the customers” doesn’t add any value to the effort!
My practice of making comments stick out as a sore thumb posted in 2010 still stands in 2020: I still set up all of my IDEs in that manner and it still produces exactly the outcome as I intended.
When I do want to drop a quick TODO comments in code (watch/listen to the podcast interview when it’s up to know why I might do that), I have templates on my IDEs to automate that: ReSharper in Visual Studio, User Snippet in VS Code, Live Template in RubyMine.
In regards to code that’s commented out, like so:
We should be using a source control system; if we ever want that code back, we have a way to bring it back. So just remove it!
Still using the example above, notice that each if-block is preceeded by a comment. Is it really necessary? How about removing the comment and extracting the expression into a method that tells us the question being asked?
There’s also the “narrator-style” comment:
Narrating every single line of code is very annoying (by the way, I think I wrote the code above many moons ago). If the comments were written initially as a placeholder for the steps that needed to be implemented, let’s make sure to get rid of it when we’re done.
Last but not least, some people say (I’ve said it myself) that comments should document “why” the code was written in such manner. I’d propose a variaton to that: how about documenting the why with some good specs (or tests, if that’s how you prefer)?
Now, I’m not refering to tests that look like this:
Such test doesn’t tell us the “why”; it tells us the “how”. I mean this kind of test (now you’ll see why spec fits better):
Summing it up, this is how I prefer to “comment” code:
- Given-When-Then specs
- Writing code English-first
If I do have a real need to drop an actual comment in code (“why do we have this query in the code that has to potential to perform badly”), I’ll probably drop a quick comment, with a link out to the issue tracker, where I’ll put more context about why the code was left like that, and where a Product Owner can decide when it’s appropriate to address the situation.
Any comments? 🙂
And for audio-only version: