I am not a programmer who knows how to program. I know this because global variables are how I fix most the issue I run into, but are constantly told this wrong.
Software dev is full of obscure keywords that describe otherwise pretty simple or basic concepts you stumble upon in practice naturally and that you probably already understand.
singleton: a class/object that is designed to be single use, i.e. only ever instantiated with a single instance. Typically used when you use class/objects more for flow control or to represent the state of the program itself, rather than using it to represent data
immutable: read-only, i.e. unchangeable
dependency injection: basically when you pass a function or object into another function object, thereby extending their effective functionality, typically for modular code and to separate concerns.
Here’s one more of my favourite examples of such a keyword: memoization
But we might need to add more features in the future so it might not just be a list in a few years. Better encapsulate it in a few layers of abstractions just to be safe.
Is it really tempting for people? They’ve given me too many headaches when I’ve had to reformat or add functionality to files.
Unless it’s a simple single use script that fit on the computer screen, I don’t feel like global variables would ever be tempting, unless it’s for constants.
Why is that weirder? The people writing scientific software are, by and large, less good at writing software than people who only specialize in software development. I’d expect there to tons of terrible engineering practices in an old code base like that
Because even trivial things like Fourier transforms (to people like me) are very difficult to understand to those that don’t know them. They took me years to understand. Non scientific software engineers do not understand those. It’s just a different course of education.
You’re also right about old code base as well. Algorithms like these belong in c++ (or C or fortran), and it’s extremely difficult to explain why to people who have no understanding of numerical computing.
in real life, I’m fighting with - I’m not joking - a few dozen “quick patches”. code does not reflect in any point functional requirements, and dude is adamant he’s in the right and supersarcastic in any occasion.
Rarely have I ever actually had consequences for my sins, which tends to be why I don’t go back and fix them…
If tech debt weight is felt in any way, it tends to get fixed. If it’s not felt, it’s just incredibly easy to forget and disregard.
(This is mostly me not learning my lesson well enough from my time being on Tech Debt: The Team. I do try and figure out the correct way to do things, but at the end of the day, I get paid to do what the boss wants as cheaply as possible, not what’s right :/ money dgaf about best practices until someone gets sued for malpractice, but on that logic, maybe the tech debt piper just hasn’t returned for payment from me yet… Only time will tell)
Fair point, I work in a consumer facing, fast turn around, short lived code project industry. Not a typical software project with long life cycles.
These practices would almost certainly bite my company in the ass if we had to maintain anything for longer than year.
Occasionally, we do have to support a client for multiple years, and everytime it’s a hilarious shit show trying to figure out how to keep all the project dependencies up to date. This is likely platform tech debt, and would be the beginning of the problem if we didn’t have the privilege of being able to start over from scratch code-wise for each client’s new order.
I guess I’m just in a lucky spot in the programmer pool where tech debt literally doesn’t hit me as hard as it usually does others, and I just couldn’t identify that before now lol
Instead of saying tech debt isn’t that bad, my tune will change to something else. Like I said, I was on a team at one point that had a worse than usual tech debt problem, and it was unworkably stressful to deal with. Im guessing that experience is more typical of being near tech debt than my other experiences.
Ah yeah, that would be a worry, except I forgot to mention that most of the code I work on usually gets thrown away after like 6 months. Makes tech debt not have nearly as big of an impact on me.
We do have a longer lasting code base that the little widgets I make run off of. That has a much more strict requirements to ensure tech debt is not introduced specifically so we don’t end up in that sort of a position.
That said, and yet we couldn’t even keep it out of our own code base. So yeah, I think my original comment is just wrong because I forgot all the ways tech debt actually has effected me in the past and how my industry’s project cycle is so short term that i rarely have the opportunity to run into tech debt that I caused in a problematic way…
Most software engineers also have to actively maintain and add features to their finished project, and those aspects change a lot about how the problem can be approached.
I failed to take into account why might I have not been effected by tech debt despite occasionally creating it before commenting. Will have to make sure that filter gets a bit stronger lol
If you’re smart you do it the quick and easy way and leave the company before it bites you in the ass. Only suckers stay with the same company for more than a few years
I.e. you did use them, but learned the hard way why you shouldn’t.
Very likely OP is a student, or entry-level programmer, and is avoiding them because they were told to, and just haven’t done enough refactoring & debugging or worked on large enough code bases to ‘get’ it yet.
Unironically: For in-house scripts and toolboxes where I want to set stuff like input directory, output directory etc. for the whole toolbox, and then just run the scripts. There are other easy solutions of course, but this makes it really quick and easy to just run the scripts when I need to.
Depends on what you're doing. Functional programming has its own downsides, especially once you want to write interactive programs, which often depend on global states. Then you either have to rely on atoms, which defeat the purpose of the functional programming, or pass around the program state, which is janly and can be slow.
I personally go multi paradigm. Simpler stuffs are almost functional (did not opt for consting everything due to performance issues), GUI stuff is OOP, etc.
Well, if you’re writing something the user will be looking at and clicking on, you will probably want to have some sort of state management that is global.
Or if you’re writing something that seems really simple and it’s own thing at first but then SURPRISE it is part of the system and a bunch of other programmers have incorporated it into their stuff and the business analyst is inquiring if you could make it configurable and also add a bunch of functionality.
I also had to work with a system where configurations for user space were done as libraries setting global constants. And then we changed it so everything had to be hastily redone so that suddenly every client didn’t have the same config.
I’ve once had a course involving programming and the lecturer rewrote the code, which we were usually using at our institute, making ALL variables global. - Yes, also each and every loop counter and iterator. 🤪
There’s no way you teach a uni course and do this kind of thing unless to demonstrate poor practice/run time difference. Are you sure you were paying attention?
Yes. He really thought it was efficient and would avoid errors if literally all variables were defined in a single Matlab function he called at the beginning of the script. We students all thought: “Man, are you serious?” As we didn’t want to debug such a mess, in our code, we ignored what he was doing and kept using local variables.
Yes, it was a course on finite deformation material models. And no, you do really, really not want to declare each and every variable in your material subroutine globally for the whole finite element program.
That’s why when your job hires new people right out of college they have no idea what they’re doing and now must be trained how to actually do the job. “What, you mean we aren’t writing this enterprise application in python!?”
This is needlessly obtuse. The definition of the word is that it’s non-constant. There isn’t an ISO definition of the word no, but there are many reputable dictionaries out there that will serve as an alternative.
Except that’s exactly what it is allowed to in algebra.
Sure, in most equations you solve in early algebra school there is only one possible value for the variables. But in many equations there can be multiple, or even infinite. It’s an unknown, and the contents can vary (depending on other constraints, ie. The rest of the equation(s)).
“Variable” refers to the label, i.e. a box that can contain anything (like *ptr is a pointer to [something we dont know anything about])
Immutable describes the contents, i.e. the stuff in the box cant change. (like int* ptr describes that the pointer points to an int)
Rust makes it very obvious that there’s a difference between constants and immutable variables, mainly because constants must be compile time constants.
What do you call it when a variable cant change after its definition, but isnt guaranteed to be the same on each function call? (E.g. x is an array that’s passed in, and we’re just checking if element y exists)
It’s not a constant, the contents of that label are “changing”, but the label’s contents cant be modified inside the scope of that function. So it’s a variable, but immutable.
And more generally mutable aliasing references of any sort are evil. Doesn’t mean they’re not useful, just that you need magic protection spells (mutexes, semaphores, fancy lock-free algorithms, atomics, etc) to use them safely. Skip the spell or use she wrong one, and the demon escapes and destroys all you hold dear.
VS Code + vim plugin is sooo slooow. I’m happy it works for you but I can’t wait to finish on boarding my onboarding buddy so I can go back to vim where I belong
I do too… but it’s not perfect. If you use the extension that uses neovim in the background is seems is the best option but still I miss my “never needing a mouse” feeling I had on emacs. I mean maybe is just lack of knowing keybindings… but back on the day when I used spacemacs it was all so intuitive.
You guys recommend VSCodium over VSCode. Is there a working sync solution similar to the one built into VSCode where you can sync all settings and extensions between machines?
Yes! It’s this one open-vsx.org/extension/zokugun/sync-settings I really like it for using a normal repository over a “gist” and so you can also use any git server provider, I think the developer is also a contributor of VSCodium itself
programmer_humor
Hot
This magazine is from a federated server and may be incomplete. Browse more on the original instance.