[Haskell-cafe] Ambiguous occurence error, despite use of DuplicateRecordFields (worked with ghc 8.10.7, not with 9.6.4)

Tom Ellis tom-lists-haskell-cafe-2023 at jaguarpaw.co.uk
Tue Apr 9 12:17:47 UTC 2024


I see, thanks.  So it does look like my wild guess was correct.  GHC
will no longer disambiguate ambiguous field selectors.  Perhaps you're
supposed to use OverloadedRecordDot for this (I'm not sure, I don't
use it myself).

Tom

On Tue, Apr 09, 2024 at 01:57:48PM +0200, Han Joosten wrote:
> The code can be found here:
> https://github.com/AmpersandTarski/Ampersand/blob/upgrade-ghc/src/Ampersand/Core/ParseTree.hs
> 
> The idea is that I have several data definitions each with a field named
> `pos`. I have a class that conveniently gets that field from whatever
> instance. This used to work, but now is broken:
> 
> data Foo = Foo
>    { pos :: !Origin,
>      ... other fields
>    }
> data Bar = Bar
>     { pos :: !Origin,
>      ... other fields
>    }
> 
> class Traced a where
>    origin :: a -> Origin
> 
> instance Traced Foo where
>   origin = pos
> 
> instance Traced  Bar  where
>   origin = pos
> 
> Of course I could rewrite the instance definitions something like
> 
> instance Traced Foo where
>   origin (Foo{pos = p} = p
> 
> Op di 9 apr 2024 om 12:38 schreef Tom Ellis <
> tom-lists-haskell-cafe-2023 at jaguarpaw.co.uk>:
> > Would it be possible to produce a minimal example that demonstrates
> > the problem?  Or at the very least share the code?  It's very hard to
> > know where to start diagnosing.  (Although perhaps someone else knows
> > without having to see code.)
> >
> > At a wild guess, have you been hit by this?
> >
> > > As of GHC 9.4.1, selector names have to be entirely unambiguous
> > > (under the usual name resolution rules), while for record updates,
> > > there must be at most one datatype that has all the field names
> > > being updated.
> >
> > https://ghc.gitlab.haskell.org/ghc/doc/users_guide/exts/duplicate_record_fields.html#extension-DuplicateRecordFields
> >
> > On Tue, Apr 09, 2024 at 12:18:14PM +0200, Han Joosten wrote:
> > > Hi all,
> > >
> > > I am currently migrating a Haskell project that uses ghc 8.10.7 to ghc
> > > 9.6.4.
> > >
> > > Using the new version, I get errors like
> > >
> > > ~~~.haskell
> > > Ambiguous occurrence ‘pos’
> > > It could refer to
> > >    either the field ‘pos’ of record ‘PClassify’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:1265:5
> > >        or the field ‘pos’ of record ‘PPurpose’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:1186:5
> > >        or the field ‘pos’ of record ‘P_ViewSegment’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:1129:5
> > >        or the field ‘pos’ of record ‘P_ViewD’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:1080:5
> > >        or the field ‘pos’ of record ‘P_IdentDf’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:1024:5
> > >        or the field ‘pos’ of record ‘P_BoxItem’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:981:9
> > >        or the field ‘pos’ of record ‘TemplateKeyValue’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:959:5
> > >        or the field ‘pos’ of record ‘BoxHeader’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:947:5
> > >        or the field ‘pos’ of record ‘P_SubIfc’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:934:9
> > >        or the field ‘pos’ of record ‘P_Interface’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:899:5
> > >        or the field ‘pos’ of record ‘P_Population’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:870:9
> > >        or the field ‘pos’ of record ‘P_Rule’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:806:5
> > >        or the field ‘pos’ of record ‘PairViewSegment’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:740:9
> > >        or the field ‘pos’ of record ‘P_NamedRel’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:591:5
> > >        or the field ‘pos’ of record ‘PAtomPair’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:467:5
> > >        or the field ‘pos’ of record ‘Pragma’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:415:5
> > >        or the field ‘pos’ of record ‘P_Relation’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:408:5
> > >        or the field ‘pos’ of record ‘Representation’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:339:5
> > >        or the field ‘pos’ of record ‘PConceptDef’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:268:5
> > >        or the field ‘pos’ of record ‘P_Pattern’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:209:5
> > >        or the field ‘pos’ of record ‘Role’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:185:5
> > >        or the field ‘pos’ of record ‘P_RoleRule’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:172:5
> > >        or the field ‘pos’ of record ‘P_Enforce’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:152:5
> > >        or the field ‘pos’ of record ‘MetaData’,
> > >           defined at
> > > /workspaces/AmpersandNamespace/src/Ampersand/Core/ParseTree.hs:136:5
> > >
> > > Previously, it was totally fine to have different data types each have a
> > > field with the same name, as long as you used `DuplicateRecordFields`
> > >
> > > I have searched the migration documentation but I couldn't find any clue
> > on
> > > why this has been changed or how to deal with it. I probably missed it.
> > > Any help is mostly appreciated!
> > >
> > > Thanks for reading
> > > Cheers,
> > > Han Joosten
> >
> > > _______________________________________________
> > > Haskell-Cafe mailing list
> > > To (un)subscribe, modify options or view archives go to:
> > > http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
> > > Only members subscribed via the mailman list are allowed to post.
> >
> > _______________________________________________
> > Haskell-Cafe mailing list
> > To (un)subscribe, modify options or view archives go to:
> > http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
> > Only members subscribed via the mailman list are allowed to post.

> _______________________________________________
> Haskell-Cafe mailing list
> To (un)subscribe, modify options or view archives go to:
> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
> Only members subscribed via the mailman list are allowed to post.



More information about the Haskell-Cafe mailing list