The trivial problem here being knowing what kinda of parameter some random function somewhere in your code expects… And your code not randomly breaking in production when someone changes that function after you already used it, unless you wrote unit tests that literally test every single line of code.
Tests are supposed to ensure functionality, not implementation. No useful unit test has ever been made obsolete by a type system. Likewise, if your tests claim your program fulfills its functionality, the type system ads no additional information to the discussion.
My point was that without a type system you need an impossible 100% code coverage, to be sure that you didn’t accidentally mess up some variable assignment or parameter somewhere, since you have nothings to easily and automatically catch those errors.
I think we are actually saying the same thing though…? Tests should really only ensure functionality. It should be the language’s job to help you with obvious implementation errors (such as using a wrong value type).
My sense is that this argument primarily holds for teams without thorough code reviews. For individuals or teams with good reviews, TypeScript adds little except for complex code or massive rewrites. I’m not saying it adds little in absolute terms, but that it adds little once you account for the overhead of using it.
A quick check everytime when you build / package the code is surely more effective than a human code review.
Also the difficulty of coding in a language where there isn’t any static type analysis still remains. How does it even work, do you have to do a manual text search everytime you change some existing function or class?
Nonsense. The compiler can handle type-checking far more quickly and acurately than any code reviewer. When I review code, I want to look at code structure, algorithms, data structures, interface design, contracts, logic, and style.
I don’t want to go through your code line by line cross-referencing every function call to make sure you put the arguments in the right order and checking every member access for typos. That’s a waste of my time, and by extension, the company’s money.
No, this is just some guy who cannot solve trivial problems without Intellij holding his hand, and he feels like that entitles him to an opinion
The trivial problem here being knowing what kinda of parameter some random function somewhere in your code expects… And your code not randomly breaking in production when someone changes that function after you already used it, unless you wrote unit tests that literally test every single line of code.
Tests are supposed to ensure functionality, not implementation. No useful unit test has ever been made obsolete by a type system. Likewise, if your tests claim your program fulfills its functionality, the type system ads no additional information to the discussion.
My point was that without a type system you need an impossible 100% code coverage, to be sure that you didn’t accidentally mess up some variable assignment or parameter somewhere, since you have nothings to easily and automatically catch those errors.
I understand your point, I just choose to reject it for the reasons given.
I think we are actually saying the same thing though…? Tests should really only ensure functionality. It should be the language’s job to help you with obvious implementation errors (such as using a wrong value type).
My sense is that this argument primarily holds for teams without thorough code reviews. For individuals or teams with good reviews, TypeScript adds little except for complex code or massive rewrites. I’m not saying it adds little in absolute terms, but that it adds little once you account for the overhead of using it.
A quick check everytime when you build / package the code is surely more effective than a human code review.
Also the difficulty of coding in a language where there isn’t any static type analysis still remains. How does it even work, do you have to do a manual text search everytime you change some existing function or class?
Nonsense. The compiler can handle type-checking far more quickly and acurately than any code reviewer. When I review code, I want to look at code structure, algorithms, data structures, interface design, contracts, logic, and style.
I don’t want to go through your code line by line cross-referencing every function call to make sure you put the arguments in the right order and checking every member access for typos. That’s a waste of my time, and by extension, the company’s money.