| 5 Jun 2021 |
 Las | * Aeson frequently runs out of memory IIRC | 17:03:38 |
 sterni (he/him) | aeson should already be compiled I'm pretty sure | 17:03:57 |
| sterni (he/him) | since I guess distribution-nixpkgs depends on it | 17:04:08 |
| Las | Really? I must have done something for it not to be fetched from cache then | 17:04:15 |
| Las | For GHC 9? | 17:04:20 |
| sterni (he/him) | yes | 17:04:25 |
| Las | Huh | 17:04:28 |
 toonn | maralorn: Why would you use builtins.attrValues rather than `with lib.maintainers; [ a b ];`? | 17:04:56 |
| sterni (he/him) | when did you try? non standard packages were not built on hydra until a few weeks ago I suspect for non x86_64-linux | 17:04:56 |
| sterni (he/him) | well it wouldn't have the discussed scoping issue | 17:05:26 |
 maralorn | toonn: Because it guarantees that a b come actually from lib.maintainers. | 17:05:46 |
| sterni (he/him) | but I'm still sticking with [ lib.maintainers.a lib.maintainers.b] tbh | 17:05:48 |
| toonn | Wait, with doesn't shadow? | 17:06:23 |
| maralorn | toonn: Nope, it doesn‘t. | 17:06:31 |
| sterni (he/him) | it does not | 17:06:30 |
| Las | If it did that would be horrible | 17:06:36 |
| sterni (he/him) | well both solutions to the issue are kinda bad tbh | 17:06:53 |
| Las | let x = y; in with z; [x] would have a different meaning depending on what z contains | 17:07:07 |
| sterni (he/him) | the root problem is that as soon as with is involved you can never statically reason about bindings unfortunately | 17:07:17 |
| maralorn | And even if it did. I also want to have assurance that I didn‘t accidentally pick an attribute that does not exist in `lib.maintainers | 17:07:25 |
| Las | the root problem is that nix is untyped | 17:07:28 |
| maralorn | * And even if it did. I also want to have assurance that I didn‘t accidentally pick an attribute that does not exist in lib.maintainers. | 17:07:30 |
| toonn | The types aren't even the biggest problem here. Nix needs better scoping primitives. | 17:08:20 |
| maralorn | maralorn sets the topice to: „Welcome to the Nix & Haskell channel. Days since someone demanded static typing in Nix: 0" | 17:09:27 |
| maralorn | * maralorn sets the topic to: „Welcome to the Nix & Haskell channel. Days since someone demanded static typing in Nix: 0" | 17:09:37 |
 fgaz | In reply to @sternenseemann:systemli.org
it does not Whatt‽ isn't that the opposite of basically every ml-style language? | 17:25:35 |
| fgaz | I always assumed it shadowed | 17:25:46 |
| maralorn | fgaz: There is an issue about that and Eelcos reasoning why it does that is actually quite convincing. | 17:26:26 |
| sterni (he/him) | the correct answer is ofc that with was a bad idea in the first place ;) | 17:26:53 |
| maralorn | I think the main advantage of with not shadowing is, that you can never break code by introducing a new value into the withed attrset. | 17:27:25 |
