Newest at the top
| 2024-11-13 20:47:07 +0100 | <haskellbridge> | <thirdofmay18081814goya> i.e. the general ability to pattern match |
| 2024-11-13 20:47:00 +0100 | <haskellbridge> | <thirdofmay18081814goya> uh well I'm mostly reasoning very loosely about cata/ana, wrt List the significant part in both cases is a pattern match on "cons" into some other datatype (cata very loosely) or a pattern match on some other type onto "List" (ana very loosely) |
| 2024-11-13 20:45:50 +0100 | <tomsmeding> | thirdofmay18081814goya: how should we properly mention you, because it's bridged (and not plumbed) your mxid is not a nick here, so we don't have autocomplete |
| 2024-11-13 20:45:05 +0100 | abrar | (~abrar@pool-72-78-199-167.phlapa.fios.verizon.net) |
| 2024-11-13 20:45:05 +0100 | <EvanR> | had to review the log carefully to see who "goya" is xD |
| 2024-11-13 20:44:55 +0100 | <haskellbridge> | <zwro> tomsmeding: you're absolutely correct. my examples were bad. i'm sure i had some kind of problem with driving Functor in the past but can't remember the exact case. maybe later it'll come to me |
| 2024-11-13 20:44:40 +0100 | <tomsmeding> | goya: if you point to the precise "ana/cata/hylo" methods that you mean (there are various packages defining them in various ways iirc), then we could see how compatible the type signatures would be |
| 2024-11-13 20:44:25 +0100 | <EvanR> | you can still treat it like a representable functor |
| 2024-11-13 20:43:43 +0100 | <tomsmeding> | there are no recursion schemes for the thing because, well, it's not a tree |
| 2024-11-13 20:43:35 +0100 | peterbecich | (~Thunderbi@syn-047-229-123-186.res.spectrum.com) peterbecich |
| 2024-11-13 20:43:28 +0100 | <tomsmeding> | it has a Foldable instance, and there are various methods |
| 2024-11-13 20:43:11 +0100 | <EvanR> | vector package, Data.Vector |
| 2024-11-13 20:43:00 +0100 | <haskellbridge> | <thirdofmay18081814goya> what module for said "Vector"? |
| 2024-11-13 20:42:53 +0100 | abrar | (~abrar@pool-72-78-199-167.phlapa.fios.verizon.net) (Ping timeout: 245 seconds) |
| 2024-11-13 20:42:42 +0100 | <EvanR> | and has a bunch of snazzy rewrite rules for fusion |
| 2024-11-13 20:42:14 +0100 | <EvanR> | but is implemented behind the scenes in a compact way |
| 2024-11-13 20:42:03 +0100 | <tomsmeding> | goya: how are they defined for Vector? |
| 2024-11-13 20:41:55 +0100 | <EvanR> | has a lot of instances like List does |
| 2024-11-13 20:41:46 +0100 | <EvanR> | Vector is pretty good |
| 2024-11-13 20:40:51 +0100 | <haskellbridge> | <thirdofmay18081814goya> can I still think in terms of ana/cata/hylo or will this destroy the performance benefits? |
| 2024-11-13 20:40:38 +0100 | <haskellbridge> | <thirdofmay18081814goya> long time user of "List", now need to use arrays from performance-critical usecase (processing logs, text files of more than 3 million lines) |
| 2024-11-13 20:40:14 +0100 | alexherbo2 | (~alexherbo@2a02-8440-3117-f07c-987b-fc29-77ee-addd.rev.sfr.net) alexherbo2 |
| 2024-11-13 20:37:13 +0100 | ash3en | (~Thunderbi@ip1f10cbd6.dynamic.kabel-deutschland.de) (Client Quit) |
| 2024-11-13 20:34:30 +0100 | <tomsmeding> | (one could of course design the _classes_ differently, but that's a separate topic) |
| 2024-11-13 20:34:08 +0100 | <tomsmeding> | (my point was: I agree with your feelings, but the types imply this: "feels weird" =/=> "there are other choices for GHC's deriving mechanism") |
| 2024-11-13 20:33:35 +0100 | <tomsmeding> | sure |
| 2024-11-13 20:33:23 +0100 | <c_wraith> | Err. Or that foldable shouldn't provide length |
| 2024-11-13 20:33:09 +0100 | <tomsmeding> | but I would argue it's a very sensible choice |
| 2024-11-13 20:33:08 +0100 | <c_wraith> | that's not intuitive, but I'm not comfortable declaring tuples aren't foldable or that foldable provide length |
| 2024-11-13 20:33:02 +0100 | <tomsmeding> | for Ord, Foldable, Traversable you do definitely have a choice of ordering, which constructor does it visit in which order; GHC makes a choice here |
| 2024-11-13 20:32:41 +0100 | <tomsmeding> | zwro: which class? Functor doesn't give one a lot of leeway |
| 2024-11-13 20:32:25 +0100 | ash3en | (~Thunderbi@ip1f10cbd6.dynamic.kabel-deutschland.de) ash3en |
| 2024-11-13 20:32:25 +0100 | <tomsmeding> | I would be the last one to say that `length (1,2,3) == 1` is intuitive |
| 2024-11-13 20:32:20 +0100 | <haskellbridge> | <zwro> but what about trees? |
| 2024-11-13 20:32:14 +0100 | <tomsmeding> | similarly for Foldable on tuples |
| 2024-11-13 20:32:03 +0100 | <haskellbridge> | <zwro> point taken |
| 2024-11-13 20:30:28 +0100 | <tomsmeding> | it may look weird, but by the types it must be this |
| 2024-11-13 20:30:14 +0100 | <tomsmeding> | and Functor, by its kind, maps over the last type-argument |
| 2024-11-13 20:30:06 +0100 | <tomsmeding> | the only sensible Functor instance for (a, b, c) is indeed the one GHC defines for tuples, because it's ((,,) a b c) |
| 2024-11-13 20:29:50 +0100 | <tomsmeding> | the only sensible Functor instance for that Foo is the one it derives |
| 2024-11-13 20:29:40 +0100 | <tomsmeding> | no, the deriving mechanism does the right thing for a data type |
| 2024-11-13 20:29:25 +0100 | <haskellbridge> | <zwro> my point is that there is no obvious way, and the deriving mechanism is opinionated |
| 2024-11-13 20:28:57 +0100 | <tomsmeding> | I was talking about the deriving machinery |
| 2024-11-13 20:28:52 +0100 | <tomsmeding> | that's the Functor instance for tuples, not the rules GHC uses for automatically deriving Functor |
| 2024-11-13 20:28:38 +0100 | <lambdabot> | (0,0,1) |
| 2024-11-13 20:28:36 +0100 | <haskellbridge> | > fmap succ (0,0,0) |
| 2024-11-13 20:28:36 +0100 | <haskellbridge> | <zwro> er |
| 2024-11-13 20:28:24 +0100 | <haskellbridge> | fmap succ (0,0,0) |
| 2024-11-13 20:28:24 +0100 | <haskellbridge> | <zwro> there |
| 2024-11-13 20:28:16 +0100 | <lambdabot> | (0,1) |