Prepare for change of Data.Text.Array to type synonym (Support text-2.1)

[TeofilC]

In an upcoming version of text, Array will become a type synonym for Data.Array.Byte.ByteArray, which means that importing its constructors will no longer bring the ByteArray constructor into scope.

This change gives a backwards compatible way to avoid the build failing when that change happens.

[Bodigrim]

text-2.1 is out now: https://hackage.haskell.org/package/text-2.1/docs/Data-Text-Array.html

[TeofilC]

I've added a commit to relax the text bounds to allow 2.1

[andreasabel]

I reported the CI timeout here:

• https://github.com/haskell/aeson/issues/1061

[andreasabel]

@phadej : Would it be possible to release this patch now?

That would help me testing text-2.1 using a Haskell CI constraint set, in projects that also use aeson.
I tried this:

constraint-set text-2.1
  ghc: >= 8.2
  constraints: text ^>= 2.1
  tests: True
  run-tests: True

raw-project
  allow-newer: text

But it fails because allow-newer: aeson:text is causing a build failure of aeson: https://github.com/haskell-hvr/microaeson/actions/runs/6034329194/job/16372500106?pr=14

[phadej]

@phadej : Would it be possible to release this patch now?

I don't see any urgency. GHC-9.8 will come soon (this month, i.e. September) with plenty of major version bumps, also there's tasty-1.5 coming, and maybe something else (I haven't checked if GHC-9.8 changes will cause cascade effect somewhere, as GHC-9.6 did with bifunctors e.g.). I have no time too spend too early on that kind of work.

AFAICT, text-2.1 becomes relevant only when GHC-9.8 is released, so I'll wait until then.

Yes, there are alpha releases so we can test stuff before actual release, but the latest GHC-9.8.1-alpha3 doesn't even have text-2.1 in it, nor bytestring-0.12 (will it even be included?), also Cabal is still at development 3.11 version. So these alpha-previews feel to be quite far away from release-quality, so I don't spend time on testing with them too much. (I would if they included some new feature I'm interested in, but these are not relevant for aeson)

[TeofilC]

Another alternative for testing things before 9.8 is released should be head.hackage. This patch should also be there. Not sure if haskell-ci supports that though

[phadej]

Another alternative for testing things before 9.8 is released should be head.hackage. This patch should also be there. Not sure if haskell-ci supports that though

It does.

[Bodigrim]

(AFAICT GHC 9.8 will ship with bytestring-0.11, and an upgrade to bytestring-0.12 is postponed until GHC 9.10)

[phadej]

(AFAICT GHC 9.8 will ship with bytestring-0.11, and an upgrade to bytestring-0.12 is postponed until GHC 9.10)

That's silly. The most visible breaking change in bytestring-0.12 and text-2.1 are the same one: Data.Array.Byte stuff.

[Bodigrim]

@phadej not me calling the shots on this.

[phadej]

I opened a GHC issue https://gitlab.haskell.org/ghc/ghc/-/issues/23926

[phadej]

@phadej not me calling the shots on this.

@Bodigrim, you say you don't, but you do: https://github.com/haskell/text/issues/532#issuecomment-1658988764

[Bodigrim]

@phadej not me calling the shots on this.

@Bodigrim, you say you don't, but you do: haskell/text#532 (comment)

If GHC developers prefer to stick to an older release for whatever reason, it's entirely up to them. Equally, it's entirely up to me whether to spend my unpaid time on backports and/or get grumpy about unplanned work.

[andreasabel]

@phadej: With the GHC 9.8.1 RC1 out, may I renew my request to release this patch please?

This would make it so much easier to test text-2.1 on various GHCs, see e.g. my recent breakage of a constraint-set at https://github.com/haskell-hvr/http-io-streams/actions/runs/6352652447/job/17255929726 .

New facts are: GHC 9.8 will be released both with bytestring-0.12 and text-2.1, see https://downloads.haskell.org/ghc/9.8.1-alpha4/docs/users_guide/9.8.1-notes.html#included-libraries.

[TeofilC]

Thanks!

[phadej]

https://hackage.haskell.org/package/aeson-2.2.1.0 is on Hackage

[andreasabel]

Great thanks!

Would be excellent if you could you also release attoparsec-aeson-2.2.1.0 to address the issue there:

Building library for attoparsec-aeson-2.2.0.0...
...
[2 of 5] Compiling Data.Aeson.Internal.Text ( src/Data/Aeson/Internal/Text.hs, dist/build/Data/Aeson/Internal/Text.o, dist/build/Data/Aeson/Internal/Text.dyn_o )

src/Data/Aeson/Internal/Text.hs:30:50: error: [GHC-88464]
    Data constructor not in scope:
      ByteArray :: ghc-prim-0.11.0:GHC.Prim.ByteArray# -> Array
    Suggested fix:
      Add ‘ByteArray’ to the import list in the import of
      ‘Data.Text.Array’ (at src/Data/Aeson/Internal/Text.hs:11:1-60).
   |
30 |   let !(SBS.SBS arr) = SBS.toShort bs in T.Text (ByteArray arr) 0 len
   |