Will anything break if Hashable got Eq as a superclass? #318

phadej · 2021-08-30T08:57:47Z

Hashable and Eq have to be compatible for unordered-containers to work properly, so having Eq as superclass of Hashable seems natural to me.

My question: is there any non-contrived use case for Hashable which will be ruled out by having Eq superclass? Bonus points if that use case is in realm of hashing-based data structures, because "This class (Hashable) exists for the benefit of hashing-based data structures."

EDIT: I'm asking here first, for smaller audience. Will later ask on libraries list & other fora.

The text was updated successfully, but these errors were encountered:

treeowl · 2021-08-30T09:04:15Z

It's hard to imagine one applicable to hashing-based data structures. For one that isn't, maybe something like computable reals?

phadej · 2021-08-30T09:05:45Z

computable reals

You mean that they cannot be Eqd? But can they be hashed either?

treeowl · 2021-08-30T09:10:35Z

Depends on your requirements. Obviously if your numbers are all the same out to N places then you'll get a lot of collisions.

…

On Mon, Aug 30, 2021, 5:05 AM Oleg Grenrus ***@***.***> wrote: computable reals You mean that they cannot be Eqd? But can they be hashed either? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#318 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOOF7IAIEEADFMVZPSRCELT7NCXJANCNFSM5DBNILHA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

treeowl · 2021-08-30T09:12:31Z

I think you're asking a very good question, and your suggestion may well be a good idea.

…

On Mon, Aug 30, 2021, 4:57 AM Oleg Grenrus ***@***.***> wrote: Hashable and Eq have to be compatible for unordered-containers to work properly, so having Eq as superclass of Hashable seems natural to me. My question: is there any non-contrived use case for Hashable which will be ruled out by having Eq superclass? Bonus points if that use case is in realm of hashing-based data structures, because "This class (Hashable) exists for the benefit of hashing-based data structures." — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#318>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOOF7M7H5EUWB377OGYHI3T7NBZNANCNFSM5DBNILHA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

phadej · 2021-08-30T09:13:01Z

But what that hash would be useful for? Specifically what algorithm (polymorphic in Hashable type) would be applicable to this (hashing N first digits of computable reals) (because if it's not polymorhic, you don't need to use Hashable).

treeowl · 2021-08-30T09:21:38Z

It was pure speculation. I know of none.

…

On Mon, Aug 30, 2021, 5:13 AM Oleg Grenrus ***@***.***> wrote: But what that hash would be useful for? Specifically what algorithm (polymorphic in Hashable type) would be applicable to this (because if it's not polymorhic, you don't need to use Hashable). — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#318 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOOF7OTNMRSMKWZS3TXVALT7NDSPANCNFSM5DBNILHA> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

phadej · 2021-10-31T16:04:49Z

https://hackage.haskell.org/package/hashable-1.4.0.0 is on Hackage.

If the bounds are simply relaxed to allow hashable-1.4.0.0 everything compiles fine and all tests pass.

However another alternative or additional option is to make a new release with hashable >=1.4.0.0 lower bound and cleaned up type-signatures.

phadej · 2021-10-31T17:21:59Z

I was imprecise, wigh GHC-7:

% cabal build -w ghc-7.10.3 --constraint='hashable >=1.4'
Resolving dependencies...
Build profile: -w ghc-7.10.3 -O1
In order, the following will be built (use -v for more details):
 - unordered-containers-0.2.14.0 (lib) (first run)
Configuring library for unordered-containers-0.2.14.0..
Preprocessing library for unordered-containers-0.2.14.0..
Building library for unordered-containers-0.2.14.0..
[1 of 9] Compiling Data.HashMap.Internal.List ( Data/HashMap/Internal/List.hs, /code/other-haskell/unordered-containers/dist-newstyle/build/x86_64-linux/ghc-7.10.3/unordered-containers-0.2.14.0/build/Data/HashMap/Internal/List.o )
[2 of 9] Compiling Data.HashMap.Internal.Unsafe ( Data/HashMap/Internal/Unsafe.hs, /code/other-haskell/unordered-containers/dist-newstyle/build/x86_64-linux/ghc-7.10.3/unordered-containers-0.2.14.0/build/Data/HashMap/Internal/Unsafe.o )
[3 of 9] Compiling Data.HashMap.Internal.Array ( Data/HashMap/Internal/Array.hs, /code/other-haskell/unordered-containers/dist-newstyle/build/x86_64-linux/ghc-7.10.3/unordered-containers-0.2.14.0/build/Data/HashMap/Internal/Array.o )
[4 of 9] Compiling Data.HashMap.Internal ( Data/HashMap/Internal.hs, /code/other-haskell/unordered-containers/dist-newstyle/build/x86_64-linux/ghc-7.10.3/unordered-containers-0.2.14.0/build/Data/HashMap/Internal.o )

Data/HashMap/Internal.hs:536:10-48:
    Could not deduce (transformers-0.4.2.0:Data.Functor.Classes.Eq1
                        (HashMap k))

as Eq1 instance is given only for base-4.9.

I'd suggest just dropping GHC-7 support. It's the simplest option.

sjakobi · 2021-10-31T23:26:53Z

I'd suggest just dropping GHC-7 support. It's simplest option.

Dropping support for GHC < 8 sounds good to me. If it turns out that some users really need continued support for GHC-7, we can consider a more complex solution.

treeowl · 2021-10-31T23:33:53Z

I'm missing something. What's the purpose of this change? And how does an Eq superclass for Hashable relate to Eq1 for HashMap k?

phadej · 2021-10-31T23:39:24Z

analogously Eq1 is a superclass of Hashable1. HashMap is Hashable1, but not always (GHC <8) Eq1.

The proposed change is just restrict the unordered-containers' supported GHC versions so Eq1 (HashMap k) always exists.

phadej · 2021-10-31T23:41:15Z

Alternatives are:

adding a dependency on transformers (and transormers-compat) to have Eq1 always.
removing Hashable1 instances (breaking change)

sjakobi · 2021-10-31T23:48:00Z

I've started working on removing the compat code for GHC < 8. So far, this looks like a very nice cleanup. We can even get rid of an entire module: https://github.com/haskell-unordered-containers/unordered-containers/blob/master/Data/HashMap/Internal/Unsafe.hs

treeowl · 2021-10-31T23:50:03Z

I'm not using 7.x anymore. If y'all think other people won't mind, let's go ahead.

treeowl · 2021-10-31T23:50:42Z

And getting rid of hideously dangerous code is always good.

This prepares u-c for supporting hashable-1.4, as discussed in #318.

sjakobi · 2021-11-01T12:59:13Z

Support for GHC < 8 was removed in #323. #324 will bump our bound on hashable.

phadej · 2021-11-07T14:11:08Z

Is there an ETA for a release?

sjakobi · 2021-11-07T19:19:48Z

@phadej I'll upload a release once #327 is merged.

So, probably Monday or maybe Tuesday.

phadej mentioned this issue Oct 31, 2021

Allow hashable-1.4 ekmett/semigroups#102

Merged

sjakobi added a commit that referenced this issue Nov 1, 2021

Drop support for GHC < 8

af2c7ea

This prepares u-c for supporting hashable-1.4, as discussed in #318.

sjakobi mentioned this issue Nov 1, 2021

Drop support for GHC < 8 #323

Merged

sjakobi added a commit that referenced this issue Nov 1, 2021

Drop support for GHC < 8 (#323)

d90ed09

This prepares u-c for supporting hashable-1.4, as discussed in #318.

sjakobi closed this as completed Nov 1, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Will anything break if Hashable got Eq as a superclass? #318

Will anything break if Hashable got Eq as a superclass? #318

phadej commented Aug 30, 2021 •

edited

Loading

treeowl commented Aug 30, 2021

phadej commented Aug 30, 2021

treeowl commented Aug 30, 2021 via email

treeowl commented Aug 30, 2021 via email

phadej commented Aug 30, 2021 •

edited

Loading

treeowl commented Aug 30, 2021 via email

phadej commented Oct 31, 2021

phadej commented Oct 31, 2021 •

edited

Loading

sjakobi commented Oct 31, 2021

treeowl commented Oct 31, 2021

phadej commented Oct 31, 2021 •

edited

Loading

phadej commented Oct 31, 2021

sjakobi commented Oct 31, 2021

treeowl commented Oct 31, 2021

treeowl commented Oct 31, 2021

sjakobi commented Nov 1, 2021

phadej commented Nov 7, 2021

sjakobi commented Nov 7, 2021

Will anything break if Hashable got Eq as a superclass? #318

Will anything break if Hashable got Eq as a superclass? #318

Comments

phadej commented Aug 30, 2021 • edited Loading

treeowl commented Aug 30, 2021

phadej commented Aug 30, 2021

treeowl commented Aug 30, 2021 via email

treeowl commented Aug 30, 2021 via email

phadej commented Aug 30, 2021 • edited Loading

treeowl commented Aug 30, 2021 via email

phadej commented Oct 31, 2021

phadej commented Oct 31, 2021 • edited Loading

sjakobi commented Oct 31, 2021

treeowl commented Oct 31, 2021

phadej commented Oct 31, 2021 • edited Loading

phadej commented Oct 31, 2021

sjakobi commented Oct 31, 2021

treeowl commented Oct 31, 2021

treeowl commented Oct 31, 2021

sjakobi commented Nov 1, 2021

phadej commented Nov 7, 2021

sjakobi commented Nov 7, 2021

phadej commented Aug 30, 2021 •

edited

Loading

phadej commented Aug 30, 2021 •

edited

Loading

phadej commented Oct 31, 2021 •

edited

Loading

phadej commented Oct 31, 2021 •

edited

Loading