Breakage caused by th-abstraction-0.3

[RyanGlScott]

I've just released th-abstraction-0.3, which features a pretty notable breaking API change:

 data DatatypeInfo = DatatypeInfo {
 ...
-  , datatypeVars      :: [Type]
+  , datatypeVars      :: [TyVarBndr]
 ...
 }

th-abstraction is present in a lot of packages' dependency graphs, so the potential for breakage is quite high here. As far as Hackage libraries go, http://packdeps.haskellers.com/reverse/th-abstraction shows that there 21 reverse dependencies of th-abstraction. Most of them already have upper version bounds of <0.3 put in place (ht to @phadej for helping me there), but there are still a handful of packages without the proper upper version bounds that ought to be checked:

• aeson-typescript
• matchable-th
• micro-recursion-schemes
• safecopy-migrate

[phadej]

I made revisions of th-lift. There is also a PR showing what broke: https://github.com/mboes/th-lift/pull/40

[hvr]

iirc I already fixed up some releases because of this of

• invariant
• deriving-compat
• text-show
• recursion-schemes

---

2019-04-26T19:18:38Z    HerbertValerioRiedel    invariant-0.4.3-r1
2019-04-26T19:18:29Z    HerbertValerioRiedel    invariant-0.5-r3
2019-04-26T19:15:24Z    HerbertValerioRiedel    deriving-compat-0.5.1-r1
2019-04-26T19:15:24Z    HerbertValerioRiedel    deriving-compat-0.5-r1
2019-04-26T19:15:23Z    HerbertValerioRiedel    deriving-compat-0.4-r1
2019-04-26T19:15:22Z    HerbertValerioRiedel    deriving-compat-0.4.2-r1
2019-04-26T19:15:22Z    HerbertValerioRiedel    deriving-compat-0.4.3-r1
2019-04-26T19:15:21Z    HerbertValerioRiedel    deriving-compat-0.4.1-r2
2019-04-26T19:03:35Z    HerbertValerioRiedel    recursion-schemes-5.0.1-r2
2019-04-26T19:03:32Z    HerbertValerioRiedel    recursion-schemes-5.0.2-r3
2019-04-26T18:56:29Z    HerbertValerioRiedel    text-show-0.4.1-r2
2019-04-26T18:56:28Z    HerbertValerioRiedel    text-show-0.4-r2
2019-04-26T18:56:26Z    HerbertValerioRiedel    text-show-0.3.1.0-r2
2019-04-26T18:56:24Z    HerbertValerioRiedel    text-show-0.3.0.0-r2
2019-04-26T18:53:26Z    HerbertValerioRiedel    text-show-3.7-r1
2019-04-26T18:53:25Z    HerbertValerioRiedel    text-show-3.7.3-r2
2019-04-26T18:53:24Z    HerbertValerioRiedel    text-show-3.7.2-r3
2019-04-26T18:53:23Z    HerbertValerioRiedel    text-show-3.7.1-r4
2019-04-26T18:53:22Z    HerbertValerioRiedel    text-show-3.6.2-r1

[RyanGlScott]

I've confirmed that the following three libraries are broken with th-abstraction-0.3 and require revisions:

• aeson-typescript-0.1.1.0 (on GHC 8.6.5)

  ``` [4 of 5] Compiling Data.Aeson.TypeScript.TH ( src/Data/Aeson/TypeScript/TH.hs, dist/build/Data/Aeson/TypeScript/TH.o ) src/Data/Aeson/TypeScript/TH.hs:200:71: error: • Couldn't match type ‘TyVarBndr’ with ‘Type’ Expected type: [Type] Actual type: [TyVarBndr] • In the second argument of ‘fmap’, namely ‘datatypeVars’ In the second argument of ‘($)’, namely ‘fmap getFreeVariableName datatypeVars’ In the first argument of ‘zip’, namely ‘(catMaybes $ fmap getFreeVariableName datatypeVars)’ | 200 | let subMap = M.fromList $ zip (catMaybes $ fmap getFreeVariableName datatypeVars) templateVarsToUse | ^^^^^^^^^^^^ src/Data/Aeson/TypeScript/TH.hs:201:66: error: • Couldn't match type ‘Type’ with ‘TyVarBndr’ Expected type: [TyVarBndr] Actual type: [Type] • In the ‘datatypeVars’ field of a record In the expression: (datatypeInfo {datatypeVars = templateVarsToUse, datatypeCons = fmap (applySubstitution subMap) datatypeCons}) In an equation for ‘fullyQualifiedDatatypeInfo’: fullyQualifiedDatatypeInfo = (datatypeInfo {datatypeVars = templateVarsToUse, datatypeCons = fmap (applySubstitution subMap) datatypeCons}) | 201 | let fullyQualifiedDatatypeInfo = (datatypeInfo {datatypeVars = templateVarsToUse | ^^^^^^^^^^^^^^^^^ src/Data/Aeson/TypeScript/TH.hs:210:53: error: • Couldn't match type ‘TyVarBndr’ with ‘Type’ Expected type: [Type] Actual type: [TyVarBndr] • In the second argument of ‘fmap’, namely ‘datatypeVars’ In the first argument of ‘mkInstance’, namely ‘(fmap getDatatypePredicate datatypeVars)’ In the expression: mkInstance (fmap getDatatypePredicate datatypeVars) (AppT (ConT ''TypeScript) (foldl (\ x y -> AppT x y) (ConT name) datatypeVars)) [otherGetTypeFn] | 210 | return [mkInstance (fmap getDatatypePredicate datatypeVars) (AppT (ConT ''TypeScript) (foldl (\x y -> AppT x y) (ConT name) datatypeVars)) [otherGetTypeFn]] | ^^^^^^^^^^^^ src/Data/Aeson/TypeScript/TH.hs:210:131: error: • Couldn't match type ‘TyVarBndr’ with ‘Type’ Expected type: [Type] Actual type: [TyVarBndr] • In the third argument of ‘foldl’, namely ‘datatypeVars’ In the second argument of ‘AppT’, namely ‘(foldl (\ x y -> AppT x y) (ConT name) datatypeVars)’ In the second argument of ‘mkInstance’, namely ‘(AppT (ConT ''TypeScript) (foldl (\ x y -> AppT x y) (ConT name) datatypeVars))’ | 210 | return [mkInstance (fmap getDatatypePredicate datatypeVars) (AppT (ConT ''TypeScript) (foldl (\x y -> AppT x y) (ConT name) datatypeVars)) [otherGetTypeFn]] | ^^^^^^^^^^^^ src/Data/Aeson/TypeScript/TH.hs:304:45: error: • Couldn't match expected type ‘Type’ with actual type ‘TyVarBndr’ • In the first argument of ‘getTypeAsStringExp’, namely ‘typ’ In the expression: getTypeAsStringExp typ In the first argument of ‘ListE’, namely ‘[getTypeAsStringExp typ | typ <- vars]’ | 304 | let typeNames = ListE [getTypeAsStringExp typ | typ <- vars] | ^^^ ```

• matchable-th-0.1.0.0 (on GHC 8.6.5)

  ``` [1 of 1] Compiling Data.Matchable.TH ( src/Data/Matchable/TH.hs, dist/build/Data/Matchable/TH.o ) src/Data/Matchable/TH.hs:52:44: error: • Couldn't match type ‘TyVarBndr’ with ‘Type’ Expected type: [Type] Actual type: [TyVarBndr] • In the second argument of ‘(<$>)’, namely ‘dtVars’ In the first argument of ‘reverse’, namely ‘(removeSig <$> dtVars)’ In the expression: reverse (removeSig <$> dtVars) | 52 | tyA : rest' = reverse (removeSig <$> dtVars) | ^^^^^^ src/Data/Matchable/TH.hs:227:50: error: • Couldn't match type ‘TyVarBndr’ with ‘Type’ Expected type: [Type] Actual type: [TyVarBndr] • In the second argument of ‘(<$>)’, namely ‘dtVars’ In the first argument of ‘reverse’, namely ‘(removeSig <$> dtVars)’ In the expression: reverse (removeSig <$> dtVars) | 227 | tyB : tyA : rest' = reverse (removeSig <$> dtVars) | ^^^^^^ ```

• micro-recursion-schemes (on GHC 8.4.4)

  ``` [3 of 3] Compiling Data.Functor.Foldable.TH ( Data/Functor/Foldable/TH.hs, dist/build/Data/Functor/Foldable/TH.o ) Data/Functor/Foldable/TH.hs:128:37: error: • Couldn't match type ‘TyVarBndr’ with ‘Type’ Expected type: [Type] Actual type: [TyVarBndr] • In the second argument of ‘map’, namely ‘vars’ In the fourth argument of ‘makePrimForDI'’, namely ‘(map toTyVarBndr vars)’ In a stmt of a 'do' block: makePrimForDI' rules (variant == Newtype) tyName (map toTyVarBndr vars) cons | 128 | (map toTyVarBndr vars) cons | ^^^^ ```

[hvr]

For convenience, here's also the links to the matrix-ci build reports clearly showing they were all green previously, and turned broken (i.e. red) in the report for the most recent index-state:

• https://matrix.hackage.haskell.org/package/matchable-th (cc @viercc)
• https://matrix.hackage.haskell.org/package/aeson-typescript (cc @thomasjm)
• https://matrix.hackage.haskell.org/package/micro-recursion-schemes (cc @vmchale)

[phadej]

reviewed aeson-typescript, matchable-th and micro-recursion-schemes

[RyanGlScott]

Thanks, @phadej! I believe that should be the full extent of the Hackage-related breakage caused by th-abstraction-0.3's release, so I'll go ahead and close this issue. I am mistaken and there is more left to be done, please feel free to re-open this issue.