Home freenode/#haskell: Logs Calendar

Logs: freenode/#haskell

←Prev  Next→
Page 1 .. 295 296 297 298 299 300 301 302 303 304 305 .. 5022
— 502,152 events total
2020-09-29 15:57:08 → Saukk joins (~Saukk@2001:998:f1:3966:96a6:dee2:2e9:fdf3)
2020-09-29 15:57:08 <Cale> The GHC API needs to be more usable
2020-09-29 15:58:03 <monochrom> As usual, LinearTypes earns paper credits, improving GHC API doesn't.
2020-09-29 15:58:37 → hnOsmium0001 joins (uid453710@gateway/web/irccloud.com/x-hhonmrvgqckxkhob)
2020-09-29 15:58:48 <maerwald> monochrom: dunno, by now there should be enough companies caring about these mundane things... I'm not sure what's going on
2020-09-29 15:58:52 <dolio> It's a little odd that - binds tighter than function application, though.
2020-09-29 15:59:06 <Cale> dolio: Yeah, that's happening with dot now too
2020-09-29 15:59:19 <dolio> Well, I'm not surprised by that.
2020-09-29 15:59:20 <Cale> It's disgusting
2020-09-29 15:59:53 <dolio> I expect "we need to import syntax from JavaScript" to be nonsense.
2020-09-29 15:59:54 <Cale> The *one* rule about precedence that we used to be able to rely on
2020-09-29 15:59:55 → xerox_ joins (~xerox@unaffiliated/xerox)
2020-09-29 16:00:04 <monochrom> Ahhh I see now, two-for-one deal, if you can use spacing to change precedence, you may as well do it to both - and . and nail two popular demands.
2020-09-29 16:00:13 × coot quits (~coot@37.30.59.210.nat.umts.dynamic.t-mobile.pl) (Quit: coot)
2020-09-29 16:00:18 × kaushikc quits (kaushikkau@gateway/shell/matrix.org/x-hwwljnatfkzlwbta) (Quit: Idle for 30+ days)
2020-09-29 16:00:27 → Rudd0 joins (~Rudd0@185.189.115.98)
2020-09-29 16:00:50 × cpressey quits (~cpressey@79-72-202-166.dynamic.dsl.as9105.com) (Quit: WeeChat 1.9.1)
2020-09-29 16:01:04 <dolio> Anyhow, this is why "avoid (success at all costs)" is the wrong reading. :)
2020-09-29 16:01:25 → nbloomf joins (~nbloomf@2600:1700:83e0:1f40:ddd1:5640:16d4:8a5)
2020-09-29 16:02:21 × machinedgod quits (~machinedg@d67-193-126-196.home3.cgocable.net) (Ping timeout: 258 seconds)
2020-09-29 16:05:13 × nbloomf quits (~nbloomf@2600:1700:83e0:1f40:ddd1:5640:16d4:8a5) (Client Quit)
2020-09-29 16:05:19 → howdoi joins (uid224@gateway/web/irccloud.com/x-mlzhxxxzmqvsdejo)
2020-09-29 16:05:23 × justsomeguy quits (~justsomeg@unaffiliated/--/x-3805311) (Ping timeout: 256 seconds)
2020-09-29 16:06:37 → zacts joins (~zacts@dragora/developer/zacts)
2020-09-29 16:07:10 → thir joins (~thir@p200300f27f0fc60094e773283d7bf825.dip0.t-ipconnect.de)
2020-09-29 16:07:49 → luke joins (~luke@bitnomial/staff/luke)
2020-09-29 16:08:03 → Guest18 joins (567e8866@gateway/web/cgi-irc/kiwiirc.com/ip.86.126.136.102)
2020-09-29 16:08:36 <zincy_> Cale: I agree, newbies at my last workplace were all over RecordDotSyntax
2020-09-29 16:09:13 × raehik quits (~raehik@cpc96984-rdng25-2-0-cust109.15-3.cable.virginm.net) (Ping timeout: 264 seconds)
2020-09-29 16:10:20 × turion quits (~turion@ip1f10fe5a.dynamic.kabel-deutschland.de) (Ping timeout: 256 seconds)
2020-09-29 16:10:23 <zincy_> Why is linear types going to be a disaster?
2020-09-29 16:11:55 <Cale> Because it's not very useful, and yet in order to attempt to use it in various circumstances, people are going to want linear-generalised versions of various higher order things from every package in the known universe.
2020-09-29 16:12:02 <quicksilver> I was always afraid this would happen with record syntax
2020-09-29 16:12:15 <Cale> This will clutter the world, and yet serve very little purpose
2020-09-29 16:12:27 × thir quits (~thir@p200300f27f0fc60094e773283d7bf825.dip0.t-ipconnect.de) (Ping timeout: 240 seconds)
2020-09-29 16:13:32 → raehik joins (~raehik@cpc96984-rdng25-2-0-cust109.15-3.cable.virginm.net)
2020-09-29 16:14:26 <zincy_> Cale: Yeah it seems like a lot of work for a solution to a very niche set of problems
2020-09-29 16:14:36 <Cale> A lot of the downsides of attempting to use linear types are not particularly obvious at first sight. For example, the paper which introduced them includes as one of its examples, a socket library which could only be used if you never want to do any I/O in your server that could potentially throw an exception you might want to catch without taking down the entire server.
2020-09-29 16:14:59 × luke quits (~luke@bitnomial/staff/luke) (Quit: sleep)
2020-09-29 16:15:00 → nbloomf joins (~nbloomf@2600:1700:83e0:1f40:94ea:ab3b:3884:ed6e)
2020-09-29 16:15:16 <zincy_> haha
2020-09-29 16:15:17 × puffnfresh_ quits (~puffnfres@45.76.124.5) (Quit: ZNC 1.8.0 - https://znc.in)
2020-09-29 16:15:31 <Cale> This is the kind of thing that beginners will hang themselves on, and blame Haskell's type system for being frustrating and incomprehensible, and the annoying thing is that in this case, they will be right.
2020-09-29 16:15:36 → puffnfresh_ joins (~puffnfres@45.76.124.5)
2020-09-29 16:16:16 <zincy_> Cale: Do you feel that there are too many obscure extensions on the market?
2020-09-29 16:16:21 <Cale> yes
2020-09-29 16:16:59 <zincy_> They feel a bit like pokemon.
2020-09-29 16:17:05 <Cale> I think it's long past time to stop making new extensions at least until people have better ideas
2020-09-29 16:17:18 <zincy_> In the old days you had ones you loved and new ones keep coming out and you dont care anymore
2020-09-29 16:18:09 <quicksilver> I don't mind a plethora of experimental extensions as long as they don't become a new default
2020-09-29 16:18:19 <quicksilver> I suppose that's hard to control/predict
2020-09-29 16:18:44 <Cale> The thing is, the language people have to learn is basically the union of all the extensions that exist
2020-09-29 16:18:55 <int-e> quicksilver: the problem with those extensions is somebody will use them :-/
2020-09-29 16:18:57 <Cale> because once it's in GHC, someone's going to use it
2020-09-29 16:18:59 <zincy_> I thought Haskell would be great at "taming" bad programmers. But it is amazing the ingenuity Haskell can give you to create bad code.
2020-09-29 16:19:22 <Cale> Yeah, I've seen some remarkably terrible things done with singletons
2020-09-29 16:19:24 → cole-h joins (~cole-h@c-73-48-197-220.hsd1.ca.comcast.net)
2020-09-29 16:19:52 <zincy_> Shudder
2020-09-29 16:19:53 <Cale> People using the type system to guarantee properties that nobody ever actually cared about, and which turn programming into a bureaucratic nightmate
2020-09-29 16:19:57 <Cale> nightmare*
2020-09-29 16:20:09 <zincy_> Haha this is so true
2020-09-29 16:20:32 <Cale> As much as I love dependent types, I also fear what bad programmers will do with them
2020-09-29 16:20:32 → wonko7 joins (~wonko7@2a01:e35:2ffb:7040:8c5f:831a:4acc:45dd)
2020-09-29 16:20:41 <int-e> I want dependable types.
2020-09-29 16:20:58 <monochrom> @quote int-e I want dependable types.
2020-09-29 16:20:58 <lambdabot> No quotes match. Take a stress pill and think things over.
2020-09-29 16:21:02 <monochrom> err heh
2020-09-29 16:21:08 <monochrom> @remember int-e I want dependable types.
2020-09-29 16:21:09 <lambdabot> Nice!
2020-09-29 16:21:09 <zincy_> Maybe there should be a dependent types certification like a driving licence
2020-09-29 16:21:13 <zincy_> or an oath to do no harm
2020-09-29 16:21:16 <Cale> Even good programmers have a hard time finding the right way to define things in a dependently typed setting to guarantee the properties that are desired and still maintaining usability
2020-09-29 16:22:01 <dolio> Right. That is publishable research.
2020-09-29 16:22:07 <zincy_> What people do with custom typeclasses can be painful
2020-09-29 16:23:05 <monochrom> I disbelieve that dependent typing and/or exploiting curry-howard is a good way to express general invariants. I disbelieve the argument "I don't need to learn a different language for specifications".
2020-09-29 16:23:24 <Cale> There are places to use all the fancy new stuff in GHC, and I've been genuinely quite happy with some of the practical things I've been able to do with GADTs and DMap for example
2020-09-29 16:23:45 → bitmagie joins (~Thunderbi@200116b80673370090ed3453f880395e.dip.versatel-1u1.de)
2020-09-29 16:23:54 <Cale> But using the fancy things has to be a last resort
2020-09-29 16:24:05 <monochrom> Re-using the type language for general invariants is worse, not better, because it is not re-using, it is shoe-horning.
2020-09-29 16:24:12 <Cale> Plain old Haskell 98 is pretty damn good 99% of the time
2020-09-29 16:24:20 × elcaro quits (~anonymous@45.32.191.75) (Remote host closed the connection)
2020-09-29 16:24:22 → thir joins (~thir@p200300f27f0fc60094e773283d7bf825.dip0.t-ipconnect.de)
2020-09-29 16:24:24 <zincy_> Yeah
2020-09-29 16:24:41 × nbloomf quits (~nbloomf@2600:1700:83e0:1f40:94ea:ab3b:3884:ed6e) (Quit: My MacBook has gone to sleep. ZZZzzz…)
2020-09-29 16:25:25 × urdh quits (~urdh@unaffiliated/urdh) (Ping timeout: 246 seconds)
2020-09-29 16:26:18 → puffnfre- joins (~puffnfres@45.76.124.5)
2020-09-29 16:26:49 × puffnfresh_ quits (~puffnfres@45.76.124.5) (Read error: Connection reset by peer)
2020-09-29 16:26:59 → urdh joins (~urdh@unaffiliated/urdh)
2020-09-29 16:27:31 <Cale> Well, when I say "all the new stuff", I'm still actually excluding a fair amount... there are a fair number of extensions that we just don't even begin to use, like all the weird record overloading shenanigans
2020-09-29 16:28:10 <merijn> Cale: Also, GADTs and fancy things are like spices
2020-09-29 16:28:12 → nbloomf joins (~nbloomf@2600:1700:83e0:1f40:b183:d428:bf80:8a35)
2020-09-29 16:28:14 <merijn> A little goes a long way
2020-09-29 16:28:15 <Cale> I guess I'm referring to the stuff that's more in the direction of dependent types, like GADTs and type families and QuantifiedConstraints and such
2020-09-29 16:28:22 <Cale> yeah
2020-09-29 16:28:40 <Cale> There's a lot that I could do if I could write just *one* type level lambda :)
2020-09-29 16:29:03 <dolio> Is there an extension where record syntax doesn't generate the accessor?
2020-09-29 16:29:08 <zincy_> monochrom: What would you recommend as a first dip of the toe into proving properties about programs?
2020-09-29 16:29:32 × thir quits (~thir@p200300f27f0fc60094e773283d7bf825.dip0.t-ipconnect.de) (Ping timeout: 246 seconds)
2020-09-29 16:29:40 <zincy_> monochrom: Operational/Denotational semantics, was thinking software foundations but Isabelle looks fun too
2020-09-29 16:29:40 <merijn> dolio: Yes
←Prev  Next→
Page 1 .. 295 296 297 298 299 300 301 302 303 304 305 .. 5022

All times are in UTC.