Let him play in the legacy code. You can just hose him off later before letting him back into the office so he doesn’t track it everywhere.
Unfortunately if you let Junior play in legacy code once, it’ll learn some nasty habits and make more of it from scratch, usually when you’re trying to sleep.
deleted by creator
I don’t want that shit all over me though. That’s why I hired junior!
That’s what the line represents. So they can pull you if you do something nasty.
How do you even fix it, when the design is broken. Sure you can make an alternative function, without all the warts, but how do you convince the “seniors” to not use the old functions, because it is in fact problematic for various reasons.
We have a separate C-library that returns all strings by ringbuffer. Why is that a problem? Call into the ringbuffer enough times, and it will wrap around, overwriting earlier strings. This, and other bad code practices such as broken multi-threaded code (not threadsafe), and more.
If you are creating an alternative implementation and leaving the old one in place, you are not fixing a problem, you are just creating a new one (and a third one because you have duplication of logic).
Either refactor the old function so that it transparently calls the new logic or delete the old function and replace all the existing usage with usage of the new one. It does not need to happen as a single commit. You can check in the new function, tell everyone to use it, and clean up usage of the old one. If anyone tries to use the old implementation, call them out in a code review.
If removing or replacing the old implementation is not possible, at least mark it as deprecated so that anyone using it gets a warning.
It wish it was that easy. It is a deep rooted problem, because of several other factors.
There’s not just one function that is l dependant on the ringbuffer, but multiple functions in multiple libs and dll’s. Some if these dll’s and libs are used by us, some are used by another department and some are used by both. It has to stay C-code, in order for both departments to be able to use the library. The other department is using pInvoke to call it from a C#-application, or from other dlls. We use it as library in a C+±application, but also for us it might happen indirectly by being called from other libs or dlls. A proper fix is therefore an immense task, and requires coordinations between multiple teams.
About deprecation, well, we handle warnings as errors, so adding a deprecated-attribute wouldn’t compile. I could add a warning-assert, but that annoys some of the senior developers, and then I have to remove it again.
About code review, many of the seniors often don’t do them and just commit directly…
To sum up, it’s easier to phase it out instead of turning everything upside down.
I’m sure that ring buffer started out as a “temporary solution”.
God, I would love that so much
deleted by creator
Neither does the senior
Anymore
me when first starting out at a job commenting everything I can
VS
me a couple years in completely lost because I never updated the comments and now none of them make any sense whatsoeverCommenting well is a highly advanced skill. I generally prefer no comments on code since it’s less likely to confuse people and I’ll merrily purge auto-doc comments and anything like
// getId() returns an id
That comment has negative value.
I can’t help it, I always get the mental image of hands clapping sarcastically when I see something like that.
In my experience refactoring lots and lots of crappy code left by devs long gone, a dev who can write useful comments is by and large a dev who can write code clean and simple enough not to need them. If the code doesn’t have informative names and clear separation of concern, chances are a comment won’t help because the dev didn’t really know what they did that worked in the first place.
Generally, yes. However I have been known to document exactly why I’m doing something incredibly stupid - because it’s required but a stupid third party library which, despite being awful, is still better than implementing it myself as a refactor.
a dev who can write useful comments is by and large a dev who can write code clean and simple enough not to need them.
my boss is great in this regard and also always has to keep reminding us to write unit tests 😅
deleted by creator
Comments should only be used to describe stuff that’s otherwise difficult to convey with code.
The best explanation I’ve ever heard is:
Comments should state the ‘Why’ never the ‘What’.
There are some cases though where the code is just complicated for reasons outside of your control, in which case “what” comments are good - but they should never be taken at face value, but only used as a first step in understanding the code. There’s a significant risk of the code not actually doing what the comment says.
Yeah. Most of the time I use comments in my algorithms, as they often use some weird optimized black magic which are difficult to understand without comments.
Like don’t set this value to the obvious default. Bad stuff happens
I write a lot of fairly simple scripts in Bash and PowerShell that should be easily understood by anybody else with moderate experience in the language, but I leave a lot of obvious comments because my coworkers don’t write any code and are extremely skittish about my automations. I add them basically to quell their fears.
Why are coworkers who don’t write any code in the codebase?
These are scripts that manage stuff on a few hundred user endpoints and a few servers. They were doing basically everything manually until I got here, and the only way I could get them on board with my slow introduction of automation is to let them see it. I have to ensure things don’t get too long, complex, or hard to explain, or they start getting nervous.
I’d rather teach people to comment well through my reviews. Much easier to understand two lines of well written function description in English than 20 lines of code.
Yeah well now we have git copilot where your comments include AI suggestions.
Now I have TONS more comments
The leash is good unit testing.
Oh, he’ll just change the unit test if it fails.
Unit tests are there to get an @ignore annotation!
There’s also that long time senior dev who’s overly confident in their abilities and force pushes production breaking code directly to master.
Not for long there isn’t.
Nah, they’re the one who’s contributing most to the project. Mostly because their code is so garbage no one else can work with it. But that’s not a thing the managers take into account.
Yeet that shit into main. Who needs unit tests, it works on my computer?
If it works in my mind it’s ready for production
Top kek
YOLOdev is the best way to dev
Look, it’s me.
Just let me rewrite ONE report from scratch so it doesn’t check a specific unindexed table that it doesn’t actually need to check and causes the report to be killed by MSQL because it takes too long to run.
Please just one rewrite. Please.
Just one little crystal report.
Just one little crystal report.
Ugh 🤮
I mean I agree with the vibe of that image but holy niche.
Make him unit test it all and then they can do changes that don’t break.
Often enough, the old code is so badly intertwined that it’s impossible to actually test. Those are the moments where all you can do is nuke it from orbit.
Well I was going for that… They will surrender before they do any changes.
Why is the junior dev tied to the other one’s penis?
Because everyone sucks at drawing hands
Bro that’s not where the penis is located on the body
Hey, I don’t body shame
