| |
| ▲ | 9rx 4 days ago | parent [-] | | "Parse, don't validate", while familiar to all HN users (it feels like it gets posted here every week), is orthogonal and does not address what we are talking about. To extrapolate, let's stay with the email address type example. In Typescript, you can define an EmailAddress type as: type EmailAddress = string & { __brand: "EmailAddress" }
If you are feeling saucy, you can even define it as: type EmailAddress = `${string}@${string}` & { __brand: "EmailAddress" }
But nether of these prove to me, the user of your API, that an EmailAddress value is actually a RFC-compliant email address. Your "parse" function, if you want to think in those terms, is quite free to slip in noncompliant characters (even if only by accident) that I, the consumer, am not expecting. The only way for me to have confidence in your promise that EmailAddress is RFC-compliant is to lean on your tests written around EmailAddress production.That isn't true for languages with better type systems. In those you can define EmailAddress such that it is impossible for you to produce anything that isn't RFC-compliant. But Typescript does not fit into the category of those languages. It has to rely on testing to define the contract. | | |
| ▲ | IceDane 4 days ago | parent [-] | | Let's imagine we are working with some hypothetical language in which what you describe is possible. At some point, you will have to write a function where you validate/parse some arbitrary string, and it then returns some sort of `Email` type as a result. That function will probably return something like `Option<Email>` because you could feed it an invalid email. The implementation for that function can also be wrong, in exactly the same way the implementation for the typescript equivalent could be wrong. You would have to test it just the same. The guarantees provided by the typescript function are exactly equivalent, except for the fact that you do technically have an escape hatch where you can "force" the creation of a branded `Email` without using the provided safe constructor, where the other language might completely prevent this - but I've already addressed this. In practice, it doesn't matter. You only make the safe constructor available to the user, so they would have to explicitly go out of their way to construct an invalid branded `Email`, and if they do, well, that's not really your problem. | | |
| ▲ | 9rx 3 days ago | parent [-] | | > Let's imagine we are working with some hypothetical language in which what you describe is possible. Or even a real one. The hypothetical part is that you'd be able to understand it. The languages with complex type systems are complete bears to work with, as I am sure you can image. It isn't just some historical curiosity as to why we prefer to use testing over types to define the contract. It is way more practical. But, no matter how you feel about that, it remains that we do define the contract in tests in the languages people actually use. > The implementation for that function can also be wrong, in exactly the same way the implementation for the typescript equivalent could be wrong. The compiler would give you an error if you got the syntax wrong, and in isolation it's fair that you could, say, get the domain name wrong as long as it is syntactically valid. I suppose what I failed to convey, making some assumptions about your understanding of type systems, is that the types would not just specify RFC-compliance. You would also spec out other dependencies such that you also couldn't even provide the wrong domain name without a compiler error. So, no, the implementation of the function couldn't be wrong. You could mis-spec the contract, of course. Then the function could be "wrong". Maybe this is the intent behind what you are trying to say here. But the idea has always been here that I would also read and accept the contract, so it wouldn't actually be "wrong". It would be exactly what we both expect. My side will be written to what the contact specifies. If you try to deviate from the contract later, you'll face the wrath of the compiler. As you point out, that's not the case in Typescript. If you try to change the contract (even if by accident) the compiler/typechecker will never say a word. I won't know until my code starts breaking. Your tests are the only thing you have to keep you in line with what we agreed upon. Tests define the contract in its case. | | |
| ▲ | IceDane 2 days ago | parent [-] | | > Or even a real one. The hypothetical part is that you'd be able to understand it. The languages with complex type systems are complete bears to work with, as I am sure you can image. I have written Haskell professionally for several years and worked with Idris and other similar languages. I'm perfectly aware of what is possible. > The compiler would give you an error if you got the syntax wrong, and in isolation it's fair that you could, say, get the domain name wrong as long as it is syntactically valid. I suppose what I failed to convey, making some assumptions about your understanding of type systems, is that the types would not just specify RFC-compliance. You would also spec out other dependencies such that you also couldn't even provide the wrong domain name without a compiler error. So, no, the implementation of the function couldn't be wrong. Wow, wouldn't that be something. This is hilariously wrong. You can build tiny functions and then compose those into larger functions and then the larger functions are all correct but only as long as their components are correct. It would be lovely if there was some magic construct that prevented you from writing incorrect code. Maybe it could be extended to hacker news comments, and then we could have just skipped this entire discussion. > You could mis-spec the contract, of course. Then the function could be "wrong". Maybe this is the intent behind what you are trying to say here. But the idea has always been here that I would also read and accept the contract, so it wouldn't actually be "wrong". It would be exactly what we both expect. My side will be written to what the contact specifies. If you try to deviate from the contract later, you'll face the wrath of the compiler. A truly baffling take. Bravo! Even on Hacker News, that was truly something. I don't even know where to begin. I guess I don't really have to, since when it comes to explaining the ease with which you transfer words straight from your ass to hacker news comments, it does a better job than I could possibly hope to do. > As you point out, that's not the case in Typescript. If you try to change the contract (even if by accident) the compiler/typechecker will never say a word. I won't know until my code starts breaking. Your tests are the only thing you have to keep you in line with what we agreed upon. Tests define the contract in its case. At this point, it's fairly clear that you are just thoroughly confused about how any of this works and what is actually possible. If you disagree, I'd love to hear how you would implement some sort of construct that allows you to write a function of the type `String -> Email` that doesn't throw if the email is invalid, can process arbitrary user input (not just string literals in your code), and somehow makes it a compiler error if your code is ever incorrect at all. PS: Explicitly returning and carrying around proof of validity doesn't count. | | |
| ▲ | 9rx 2 days ago | parent [-] | | > that doesn't throw if the email is invalid Why can't it 'throw' if the email is invalid? You can — and would — encode that into the contract, detailing exactly what that means, so no problem for the user. They will write their code expecting that condition and everything will be fine. There is nothing special about failure cases. It is only a state, just like every other state. > At this point, it's fairly clear that you are just thoroughly confused about how any of this works and what is actually possible. Ad hominem is a logical fallacy. It is surprising that a comment that claims to understand programming language types in great detail would also commit an obvious logical error, given how important logic is in that context. But, stranger things have happened, I suppose. Unfortunately, there is nothing we can take from a logical error. Assuming it was trying to be written in good faith and free of errors, no doubt we'll get a meaningful explanation of what was actually trying to be expressed with code examples to follow. Looking forward to it. | | |
| ▲ | IceDane 2 days ago | parent [-] | | Thank you for (once again) making it clear to any reader that you are not to be taken seriously. If only we could edit older comments so that the reader didn't have to suffer through all your nonsense to find out. | | |
| ▲ | 9rx 2 days ago | parent [-] | | The reader has no idea who the author is. Nor do they need to as the words on the page already provide sufficient information to determine if the words are worthy of being taken seriously or not. Trying to assign meaning from the author onto the message is not logically sound. Words stand on their own. Turning to logical fallacies in an attempt to steer the reader away from understanding that Typescript relies on testing to downplay the original misunderstanding that you mention realizing cannot be edited now isn't going to work, I'm afraid. The reader is able to see through such a thin veneer. They could be convinced by a technical exchange directed at the topic at hand, but it is telling that we never saw anything of the sort. |
|
|
|
|
|
|
|
