December 2015 - ghc-devs

Re: GHC 8.0 feature freeze imminent
by Ryan Scott 14 Dec '15

14 Dec '15

One thing I definitely want to get in before the 8.0 release is: * A fix for Trac #10716 [1] (adding strictness metadata to GHC generics). This feels like a natural counterpart for Trac #10697 [2] (reworking strictness information to Template Haskell), which currently has a Phab Diff in review [3]. I'd like for these to be released at the same time so that I can use Template Haskell to backport the machinery in generic-deriving [4]. In addition, I realized that my encoding of selector metadata is not very extensible at the moment, and this would give me a good excuse to fix it. Some things that I probably won't get in time for 8.0: * Automatic derivation of Bifunctor/Bifoldable/Bitraversable, and of Eq1/Eq2/Ord1/Ord2/Read1/Read2/Show1/Show2. These would require resolving Trac #10448 [5] and #11135 [6] first, which might need more discussion. Not only that, but I still need to write up a proposal for how each of these classes would be derived, and there would need to be additional discussion on those points... All in all, it doesn't sound like something I could do in two weeks :) * Making Generic1 kind-polymorphic, Trac #10604 [7]. The first step to achieving this was done recently [8], but unfortunately, I wasn't able to make the actual datatypes in GHC.Generics kind-polymorphic for unknown reasons (I'd still like to talk to Richard Eisenberg about this at some point). Not only that, but I'm not sure what the implications of making Generic1 poly-kinded would be yet. I'd like to talk to Andres Löh about this. Ryan S. ----- [1] https://ghc.haskell.org/trac/ghc/ticket/10716 [2] https://ghc.haskell.org/trac/ghc/ticket/10697 [3] https://phabricator.haskell.org/D1605 [4] http://hackage.haskell.org/package/generic-deriving [5] https://ghc.haskell.org/trac/ghc/ticket/10448 [6] https://ghc.haskell.org/trac/ghc/ticket/11135 [7] https://ghc.haskell.org/trac/ghc/ticket/10604 [8] https://ghc.haskell.org/trac/ghc/ticket/10775

3 2

Remaining type=kind items
by Richard Eisenberg 14 Dec '15

14 Dec '15

Yes -- these are very much on the TODO list. Sadly, visible type application didn't make it. It could with another two weeks(?), but I don't think it's worth holding things up further. Unlike D808, visible type application doesn't touch wide swaths of code, so merging it after the ghc-8.0 branch won't cause backporting trouble. I plan on writing a longish section in the user manual about GHC's kind system, updating the PolyKinds discussion already there. Should happen this week. Richard On Dec 14, 2015, at 8:32 AM, Ben Gamari <ben(a)smart-cactus.org> wrote: > Hi Richard! > > This is probably on your to-do list, but here are a few > places where I think the no-kinds work could use some documentation, > > - the introduction of kind equalities could use a brief note in the > release notes > > - visible type application should be mentioned in the release notes > (this made it in, right?) > > - visible type application syntax should be described in the users > guide > > - `-XTypeInType` needs some discussion in the users guide. Be sure > to mention the fact that `*` needs to be imported explicitly and > state precisely how one would do this (since it is a symbol) > > - the notion of levity could use a bit of discussion in the users > guide > > Cheers, > > - Ben

1 0

Re: Performance of pattern checker on OptCoercions
by Richard Eisenberg 12 Dec '15

12 Dec '15

Here's a thought: is there a way you can tell when you're in a bad case? There's surely one easy way: just count how many times your algorithm goes around. But perhaps you can do better, by looking for lots of guards, etc. My idea would be to enable guard checking by default, but to bail when the going gets tough. Bailing could issue a warning that such-and-such a function/case/multi-way-if is too ornate to be checked performantly. The user could then specify `-fno-warn-incomplete-pattern-guards` to turn off pattern-guard checking, `-fno-warn-ornate-guards` simply to disable the warning in the too-tough-to-handle case (but still check other guards in the module), or `-fwarn-incomplete-pattern-guards-at-any-cost` to instruct the check simply to work harder. Thus, pattern-guard checking would be enabled by default, but we would be protected from requiring 10GB of memory to compile a module. I was inspired by the fact that I would be tempted to enable pattern-guard checking by default in my own code, even if the check is non-performant. Then, in years' time, I would have a module that inexplicably takes forever to compile and forget to look here. The design above lets me have my cake and eat it too -- that is, to have my code checked but not worry about surprising compilation times. What do you think? Richard On Dec 12, 2015, at 1:05 PM, George Karachalias <george.karachalias(a)gmail.com> wrote: > > > On Sat, Dec 12, 2015 at 12:01 AM, Simon Peyton Jones <simonpj(a)microsoft.com> wrote: > `f` will generate an empty uncovered set while g will generate: > > uncovered = > > { x |> { x ~ [], x ~ (_:_) } > > , x |> { x ~ (_:_), x ~ [] } } > > > > which is also semantically empty but this cannot be detected until we call the > > term oracle on it to see the inconsistency. > > > > So, why not call the term oracle more aggressively (at every node of the tree), to prune empty sets? It should not be hard to spot common cases; in what you show above it’s very easy. > > > Yes, we've been tempted to do this before but it didn't work. Since the sets can be really > big, we have relied on laziness a lot to save space and time. By calling the oracle more > often we will evaluate more eagerly parts of the sets which may lead to more space leaks. > > Additionally, in my reply for #11195 (https://ghc.haskell.org/trac/ghc/ticket/11195#comment:4) > you can find another example for which solving the constraints eagerly does not necessarily > improve the situation. Sometimes the sets are big and consistent. So checking eagerly does > not actually prune anything. I can see now that my `f` and `g` above were bad examples for this. > > Last but not least, the term oracle is not the most performant part of the checker. I would > certainly avoid calling it more often than we already do. > > Nevertheless, I do not want to drop the work we have done on pattern guards either so I > think that the flag solution is a very good and realistic option for now. If we really manage to > find a nice way to do better we can always update the code which will not be removed from > GHC. I will certainly keep thinking about it :) > > George > > -- > things you own end up owning you

1 0

User manual sub-sections
by Simon Peyton Jones 12 Dec '15

12 Dec '15

Ben, Austin, Could we see the sub-sub-sections of the user manual in the table of contents http://downloads.haskell.org/~ghc/master/users-guide/ The section titles are often very helpful. C.f. the old version https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/index.html Simon

4 5

kind inference (#11203)
by Richard Eisenberg 12 Dec '15

12 Dec '15

Hi Simon and other devs, I struggled a bit today with some issues around kind inference that I wanted to write down. Here are some illuminating test cases, explained below: -- useful definition: data SameKind :: k -> k -> * 1. data T (a :: k1) x = MkT (S a ()) data S (b :: k2) y = MkS (T b ()) 2. data Q (a :: k1) (b :: k2) c = MkQ (SameKind a b) 3. data D1 = MkD1 (forall p (a :: k). p a -> Int) 4. data D2 a b = MkD2 (Proxy (a :: Maybe k)) | MkD3 (Proxy (b :: [k]) (1) should be accepted. But this is rather hard to arrange for, because T and S do not have CUSKs. Their kinds are generalized only after checking the mutually-recursive group. So, to accept this pair, we need to use SigTvs for k1 and k2, to allow them to unify with each other. (This works today.) (2) should be rejected. But it is currently accepted, because k1 and k2 are SigTvs, and so they unify with each other. (3) should be rejected. I have no idea where that k should be bound. Previously, the k was bound in the forall, but after the wildcards-refactor patch, that's no longer possible. (This is taken from test case ghci/scripts/T7873. HEAD accepts this definition, but produces a bogus tycon.) Note that this was accepted correctly before the wildcards-refactor. (4) should be accepted. This demonstrates that k can be introduced in constructors but scopes over the entire definition. The difference with case (3) is that (4) has type variables whose kinds mention k, whereas (3) does not. How should we arrange for all of this behavior? As stated in #11203, I think the use of SigTvs in PolyKinds is highly suspect. I have no good ideas here, though. In case (3), k ends up in the hsq_kvs (soon to be renamed to hsq_implicit) field of the HsQTyVars for D1. After kind inference, I've written a check to make sure k appears free in the kind of a tyvar. That works. The problem is that the check also snags cases (1) and (2): my check uses the name from the LHsQTyVars, and with SigTvs in the mix, (1) and (2) end up with kind variables with names different from the names in the LHsQTyVars. I'm sure there's some way to disentangle all of this, but I've given it a rest for now. This hit me in the fact that performance test case T9872d is like case (1), and the SigTvs threw my check off. But removing it caused problems elsewhere. I'd enjoy discussing next week. Thanks, Richard

1 1

A comment on Introspective-Haskell
by Levent Erkok 12 Dec '15

12 Dec '15

I just came across https://ghc.haskell.org/trac/ghc/ticket/11081, and the corresponding wiki-page: https://ghc.haskell.org/trac/ghc/wiki/TemplateHaskell/Introspective I think this is a terrific idea. In the past, I've tried both TH and haskell-src-exts to do relatively simple things, but ended-up abandoning them due to the inherent complexity of source level haskell that had very little to do with what I really cared about. Being able to get your hands on Core at the regular Haskell level would truly simplify life, and I suspect would open the flood-gates for a lot of people to develop extremely cool/useful artifacts, making the GHC/Haskell experience even better. I hope this idea is taken further and sees the light-of-day. Richard: Did you have any further thoughts about possible plans? -Levent.

4 8

RE: Performance of pattern checker on OptCoercions
by Simon Peyton Jones 11 Dec '15

11 Dec '15

`f` will generate an empty uncovered set while g will generate: uncovered = { x |> { x ~ [], x ~ (_:_) } , x |> { x ~ (_:_), x ~ [] } } which is also semantically empty but this cannot be detected until we call the term oracle on it to see the inconsistency. So, why not call the term oracle more aggressively (at every node of the tree), to prune empty sets? It should not be hard to spot common cases; in what you show above it’s very easy. Simon From: George Karachalias [mailto:george.karachalias@gmail.com] Sent: 11 December 2015 11:27 To: Ben Gamari <ben(a)well-typed.com> Cc: Simon Peyton Jones <simonpj(a)microsoft.com>; Richard Eisenberg <eir(a)cis.upenn.edu>; GHC developers <ghc-devs(a)haskell.org> Subject: Re: Performance of pattern checker on OptCoercions Hello Ben, On Fri, Dec 11, 2015 at 11:58 AM, Ben Gamari <ben(a)well-typed.com<mailto:ben@well-typed.com>> wrote: Hi George, Richard has encountered a bit of a performance cliff when merging his no-kinds work. In particular OptCoercions now results in multiple gigabytes of memory consumption during compilation due to the pattern checker. The problem seems to be the opt_trans_rule binding, which has numerous equations, each of which has patterns of various complexities and guards. Might this be another case where disabling the pattern checker is unavoidable? I am afraid so. I have just responded to the ticket about it. The essence is the difference between `f` and `g` below: f x = case x of [] -> ... (_:_) -> ... g y | [] <- y = ... | (_:_) <- y = ... `f` will generate an empty uncovered set while g will generate: uncovered = { x |> { x ~ [], x ~ (_:_) } , x |> { x ~ (_:_), x ~ [] } } which is also semantically empty but this cannot be detected until we call the term oracle on it to see the inconsistency. Since pattern guards can pattern match against any variable whilst case expressions match a single expression (`x` above) I can not make the check treat them the same. From what I see, until now the pattern guards in opt_trans_co involved mostly pattern matching with Maybe which has only two constructors. I can easily assume that this is the reason we did not have such a problem until now. So many guards are already a challenge for the pattern match checker but maybe changing this (by this I mean to not use pattern guards on types with many constructors because this is the most expensive thing for the whole check) is enough to make GHC bootstrap. I hope this helps, I am really confident that this is the cause of the problem. Nevertheless, I will look into it more to see if I can find another source. George -- things you own end up owning you

1 0

Re: Performance of pattern checker on OptCoercions
by Richard Eisenberg 11 Dec '15

11 Dec '15

If I understand correctly, you're saying that the problem is due to my use of guards, not anything with fancy types. If that's correct, then this sort of thing might be a trap anyone could stumble into. (The code in OptCoercion is fairly routine for Haskell.) It seems like disabling checking guards by default may be best, but I'd love a flag (not with -Wall) to enable it optionally. Richard On Dec 11, 2015, at 6:27 AM, George Karachalias <george.karachalias(a)gmail.com> wrote: > Hello Ben, > > On Fri, Dec 11, 2015 at 11:58 AM, Ben Gamari <ben(a)well-typed.com> wrote: > > Hi George, > > Richard has encountered a bit of a performance cliff when merging his > no-kinds work. In particular OptCoercions now results in multiple > gigabytes of memory consumption during compilation due to the pattern > checker. The problem seems to be the opt_trans_rule binding, which has > numerous equations, each of which has patterns of various complexities > and guards. Might this be another case where disabling the pattern > checker is unavoidable? > > I am afraid so. I have just responded to the ticket about it. The essence is the > difference between `f` and `g` below: > > f x = case x of > [] -> ... > (_:_) -> ... > > g y | [] <- y = ... > | (_:_) <- y = ... > > `f` will generate an empty uncovered set while g will generate: > > uncovered = > { x |> { x ~ [], x ~ (_:_) } > , x |> { x ~ (_:_), x ~ [] } } > > which is also semantically empty but this cannot be detected until we call the > term oracle on it to see the inconsistency. Since pattern guards can pattern match > against any variable whilst case expressions match a single expression (`x` above) > I can not make the check treat them the same. > > From what I see, until now the pattern guards in opt_trans_co involved mostly > pattern matching with Maybe which has only two constructors. I can easily assume > that this is the reason we did not have such a problem until now. So many guards > are already a challenge for the pattern match checker but maybe changing this (by > this I mean to not use pattern guards on types with many constructors because this > is the most expensive thing for the whole check) is enough to make GHC bootstrap. > > I hope this helps, I am really confident that this is the cause of the problem. > Nevertheless, I will look into it more to see if I can find another source. > > George > > -- > things you own end up owning you

2 1

type operators and colon in GHC
by John Leo 11 Dec '15

11 Dec '15

I sent this to haskell-cafe a little while ago but didn't get a response, so I thought I'd try here. I'd guess this is a case of the GHC user guide needing an update, but I'd like an expert opinion. --- According to sections 7.4.3 and 7.4.4 of the latest GHC documentation https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/data-type-e… you can define (7.4.3) an infix type constructor as long as it begins with a colon, for example data a :*: b = Foo a b and furthermore (7.4.4) you can define an infix operator without having to use a colon if you enable the TypeOperators extension: data a * b = Foo a b However if I try the former without using TypeOperators I get this compiler error in 7.10.2: Illegal declaration of a type or class operator ‘:*:’ Use TypeOperators to declare operators in type and declarations Using TypeOperators fixes this, but then * without colon also works so I don't see the point of using colon anymore. My guess is this was some some kind of historical distinction which is no longer valid and the documentation needs to be updated. Is this true, or am I missing something? John

3 3

[ANNOUNCE] Glasgow Haskell Compiler version 7.10.3
by Ben Gamari 11 Dec '15

11 Dec '15

Hello everyone, We are pleased to announce the release of GHC 7.10.3: https://downloads.haskell.org/~ghc/7.10.3/ There can be found source tarballs and binary distributions for 64-bit and 32-bit modern Linux (GMP 5.0 or later), CentOS (GMP 4.0), Windows, and 64-bit Mac OS X platforms. These binaries and tarballs have an accompanying SHA256SUMS file signed by my GPG key id (0x97DB64AD). Significant fixes in release include changes to the simplifier's treatment of rules, the handling of Mac OS X frameworks, and support for response files to work around the restrictive command line length limit on Windows. As always, a full accounting of the changes present in this release can be found in the release notes [1]. The previous release, 7.10.2, was well-behaved save a couple notable bugs; while we have merged a good number of bug fixes in 7.10.3 they were were largely low risk and so we expect that this release should be similiarly stable. A notable exception is the upgrade of the Windows compiler toolchain to GCC 5.2. Typically we would refrain from making such large changes in a point release but Windows users have been long suffering at the hand of the old toolchain (e.g. lack of response file support, #8596, and lack of SEH support). We expect that this change should fix far more than breaks. Unfortunately, due to some oversights in the release process there are two source tarballs for this release. The first, ghc-7.10.3-src.tar.{bz2,xz}, does not include the release notes in the users guide. This is fixed in the second patchlevel release, ghc-7.10.3a-src.tar.{bz2,xz}. It is recommended that users wanting a source release use the ghc-7.10.3a-src tarballs. GHC 7.10.3 will very likely be the last release in the GHC 7 series. In the coming weeks we will be beginning the release process for GHC 8.0, which will include a number of exciting features including the merger of kinds with types, injective type families, imporved DWARF debugging information, applicative-do syntax, a Strict language extension synonyms mechanism, and more. See the GHC 8.0.1 status page for details [2]. Happy compiling! Cheers, - Ben [1] http://downloads.haskell.org/~ghc/7.10.3/docs/html/users_guide//release-7-1… [2] https://ghc.haskell.org/trac/ghc/wiki/Status/GHC-8.0.1

3 3