2024/11/19

Newest at the top

2024-11-19 12:21:35 +0100ubert(~Thunderbi@178.115.41.15.wireless.dyn.drei.com) ubert
2024-11-19 12:11:12 +0100 <kuribas> For example, do can be overloaded by defining your own (>>=) function.
2024-11-19 12:10:49 +0100sawilagar(~sawilagar@user/sawilagar) sawilagar
2024-11-19 12:10:30 +0100sawilagar(~sawilagar@user/sawilagar) (Remote host closed the connection)
2024-11-19 12:10:02 +0100ubert(~Thunderbi@178.115.41.15.wireless.dyn.drei.com) (Ping timeout: 252 seconds)
2024-11-19 12:09:56 +0100notzmv(~daniel@user/notzmv) (Read error: Connection reset by peer)
2024-11-19 12:07:20 +0100 <kuribas> It's a nice to have in idris, because it allows syntactic overloading, but I am not sure it is necessary to have.
2024-11-19 12:06:18 +0100 <kuribas> And type checking very slow.
2024-11-19 12:06:05 +0100 <kuribas> Which makes getting accurate error message difficult.
2024-11-19 12:05:52 +0100 <kuribas> One blocking issue for me with idris and error messages is the ad-hoc polymorphism.
2024-11-19 12:05:35 +0100 <kuribas> yeah ...
2024-11-19 12:05:33 +0100ubert(~Thunderbi@178.115.41.15.wireless.dyn.drei.com) ubert
2024-11-19 12:05:23 +0100 <Hecate> maybe improve idris too while you're at it? :D
2024-11-19 12:05:16 +0100ubert(~Thunderbi@178.115.41.15.wireless.dyn.drei.com) (Read error: Connection reset by peer)
2024-11-19 12:05:14 +0100 <Hecate> there are indeed
2024-11-19 12:05:07 +0100 <kuribas> When dabbling with idris I feel there is a lot of low hanging fruit there.
2024-11-19 12:04:53 +0100 <Hecate> best case, an LLM will feed off of your work later
2024-11-19 12:04:52 +0100 <kuribas> Hecate: agreed :)
2024-11-19 12:04:40 +0100 <Hecate> kuribas: ah, DT languages with good ergonomics are something that need you to advance the state of the art
2024-11-19 12:04:35 +0100 <kuribas> Or maybe elaborate a bit then?
2024-11-19 12:04:11 +0100 <Hecate> kuribas: To rephrase your suggestion: "Would it be possible to train an LLM on an existing dataset to generate new things, which an LLM simply cannot?"
2024-11-19 12:04:06 +0100 <int-e> and LLMs are terrible at that
2024-11-19 12:04:02 +0100 <kuribas> I am happy with GHC error messages, but I am interested in creating a general purpose dependently typed language.
2024-11-19 12:04:00 +0100 <int-e> but error messages need to be accurate
2024-11-19 12:03:41 +0100 <kuribas> More beginner friendly.
2024-11-19 12:03:30 +0100 <int-e> kuribas: define "better"
2024-11-19 12:03:28 +0100 <Hecate> kuribas: no it's not outrageous, it's ridiculous :P
2024-11-19 12:03:20 +0100 <lambdabot> the eta-reduction property does not hold
2024-11-19 12:03:20 +0100 <int-e> @ghc
2024-11-19 12:03:04 +0100 <kuribas> Is that so outrageous?
2024-11-19 12:03:01 +0100 <lambdabot> Its main purpose is to encapsulate the Horrible State Hack
2024-11-19 12:03:01 +0100 <int-e> @ghc
2024-11-19 12:02:33 +0100 <Hecate> :P
2024-11-19 12:02:32 +0100 <Hecate> you silly
2024-11-19 12:02:31 +0100 <Hecate> kuribas: hahahahahahaha
2024-11-19 12:01:56 +0100 <kuribas> Would it be possible to train a llm to generate better error messages?
2024-11-19 11:53:16 +0100__monty__(~toonn@user/toonn) toonn
2024-11-19 11:23:37 +0100mange(~user@user/mange) (Quit: Zzz...)
2024-11-19 11:21:54 +0100tzh(~tzh@c-76-115-131-146.hsd1.or.comcast.net) (Quit: zzz)
2024-11-19 11:19:23 +0100 <Leary> (where `type f ! k = forall t. f (k @ t)`)
2024-11-19 11:18:31 +0100 <Leary> jackdk: For e.g. `type data Wibble = Wobble Foo | Nibble Wibble Wibble` I'm writing TH that ought to generate `deriving instance Show (Wibble @ t)` given that we already have `Show! Foo` in scope, but really it will generate `deriving instance (Show! Foo, Show! Wibble, Show! Wibble) => Show (Wibble @ t)`.
2024-11-19 11:13:12 +0100 <jackdk> Something like that. I don't know enough about your problem to be sure
2024-11-19 11:12:58 +0100lxsameer(~lxsameer@Serene/lxsameer) lxsameer
2024-11-19 11:07:21 +0100Guest4(~Guest4@102.91.71.26) (Client Quit)
2024-11-19 11:07:15 +0100Guest4(~Guest4@102.91.71.26)
2024-11-19 11:06:41 +0100CrunchyFlakes(~CrunchyFl@ip1f13e94e.dynamic.kabel-deutschland.de)
2024-11-19 11:06:35 +0100mari-estel(~mari-este@user/mari-estel) mari-estel
2024-11-19 11:06:05 +0100Guest4(~Guest4@102.91.71.26) (Client Quit)
2024-11-19 11:05:43 +0100Guest4(~Guest4@102.91.71.26)
2024-11-19 11:02:09 +0100 <Leary> jackdk: Hmmm. So the idea is to hide the messy instances by tying them to a private newtype wrapper, then produce a standalone `deriving instance ... via ...` declaration, for which GHC would infer a clean context? Sounds feasible...