Remove strict compose versions

[sjakobi]

…as discussed in
haskell-unordered-containers/unordered-containers#299 (comment).

[sjakobi]

Once this is merged, I think it would be good to make a release in time for GHC 9.0, so users of compose are less likely to encounter the differences in strictness.

[treeowl]

Do you know the deadline?

…

On Thu, Sep 10, 2020, 5:34 AM Simon Jakobi ***@***.***> wrote: Once this is merged, I think it would be good to make a release in time for GHC 9.0, so users of compose are less likely to encounter the differences in strictness. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#745 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOOF7OQYUIHX3Y3LQZPGMTSFCMS5ANCNFSM4REUVONA> .

[sjakobi]

Do you know the deadline?

Unfortunately I don't. https://gitlab.haskell.org/ghc/ghc/-/issues/18216#note_298165 seems to indicate that the release preparations are ongoing.

@bgamari When would a containers release need to happen so it could be included in GHC 9.0.1?

[treeowl]

Do we have enough map/mapMaybe fusion rules to make the old behavior cheap to recover?

[sjakobi]

Do we have enough map/mapMaybe fusion rules to make the old behavior cheap to recover?

No idea.

I was wondering how to best convert a lazy map to a strict one by forcing all the values. Is there anything better than *.Strict.map id?

[sjakobi]

Let's not discuss this in a closed PR though.

[treeowl]

`map id` is the best way I know.

…

On Fri, Sep 11, 2020, 5:50 PM Simon Jakobi ***@***.***> wrote: Let's not discuss this in a closed PR though. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#745 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOOF7J3QCM7GX5DSKK3TTLSFKLQTANCNFSM4REUVONA> .

[treeowl]

Actually, that's wrong.... The nf package has the right stuff for doing it efficiently. Or a strictifying fold.

…

On Fri, Sep 11, 2020, 6:05 PM David Feuer ***@***.***> wrote: `map id` is the best way I know. On Fri, Sep 11, 2020, 5:50 PM Simon Jakobi ***@***.***> wrote: > Let's not discuss this in a closed PR though. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#745 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAOOF7J3QCM7GX5DSKK3TTLSFKLQTANCNFSM4REUVONA> > . >

[sjakobi]

Actually, that's wrong.... The nf package has the right stuff for doing it
efficiently. Or a strictifying fold.

What does that look like in practice? Could you give an example?

[treeowl]

forceValues :: Map k v -> () forceValues = foldl' (flip seq) () Looks like I was wrong about nf. I thought it had a type like this: newtype Shallow a = Shallow a instance NFData (Shallow a) where rnf !_ = () With that, forceValues = rnf . L.map Shallow where L.map Shallow should be rewritten to a coercion.

…

On Fri, Sep 11, 2020, 6:10 PM Simon Jakobi ***@***.***> wrote: Actually, that's wrong.... The nf package has the right stuff for doing it efficiently. Or a strictifying fold. What does that look like in practice? Could you give an example? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#745 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOOF7MCDURYQ6ZE32E5M2LSFKN3XANCNFSM4REUVONA> .

[sjakobi]

So a toStrict function could be defined as

toStrict :: Map k v -> Map k v
toStrict m = foldl' (flip seq) () m `seq` m

Cheers!

[treeowl]

Yes, but fusing that definition ... not so easy! Maybe we should actually offer that toStrict function so we can attach RULES to it.

…

On Fri, Sep 11, 2020, 6:41 PM Simon Jakobi ***@***.***> wrote: So a toStrict function could be defined as toStrict :: Map k v -> Map k v toStrict m = foldl' (flip seq) () m `seq` m Cheers! — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#745 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAOOF7I3G3QAJAAH7O7MRKLSFKRTNANCNFSM4REUVONA> .