Giant list of missing/changed functions

[RyanGlScott]

After scanning through the GHC and base changelogs, I made a list of functions that were introduced or changed in later versions of base that would be good candidates for base-compat. I intentionally skipped the GHC.-prefixed modules for now, since they tend to be rather volatile.

base-4.3.0.0: (Edit: these are no longer needed due to dropping support for base-4.2.0.0 and earlier)

• mfilter added to Control.Monad
• partitionEithers (in Data.Either) was made lazier
• swap added to Data.Tuple
• castCUCharToChar, castCharToCUChar, castCSCharToChar, and castCharToCSChar added to Foreign.C.String
• unsafeLocalState added to Foreign.Marshal
• hGetBufSome added to System.IO

base-4.4.0.0:

• gcd 0 0 now returns 0
• forkIOWithUnmask, forkOn, and forkOnWithUnmask, added to Control.Concurrent
• allowInterrupt added to Control.Exception
• Control.Monad.ST.Unsafe and Control.Monad.ST.Lazy.Unsafe modules added
• inits, tails, intersperse, and intersectBy (in Data.List) are lazier
• Foreign.ForeignPtr.Unsafe and Foreign.Marshal.Unsafe modules added
• Data.String reexports String, lines, unlines, and unwords
• maybeNew (in Foreign.Marshal.Utils) had its type generalized from (a -> IO (Ptr a)) -> Maybe a -> IO (Ptr a) to (a -> IO (Ptr b)) -> Maybe a -> IO (Ptr b). (Make sure to check this with typediff.)
• catchIOError and tryIOError added to System.IO.Error
• unsafeDupablePerformIO added to System.IO.Unsafe
• peekFilePath, puts, and c_safe_open added to System.Posix.Internals
  • On non-Windows platforms, peekFilePathLen added to System.Posix.Internals

base-4.5.0.0:

• fixIO (in System.IO) is now thread-safe
• unsafeFixIO added to System.IO.Unsafe

base-4.5.1.0:

• writeChan (in Control.Concurrent.Chan) fixed to avoid potential race condition
• hGetBufSome (in System.IO) fixed to avoid blocking on Windows

base-4.6.0.0:

• forkFinally added to Control.Concurrent
• mkWeakMVar, modifyMVarMasked, and modifyMVarMasked_ added to Control.Concurrent.MVar
• bitDefault, testBitDefault, and popCountDefault added to Data.Bits
• modifyIORef' and atomicModifyIORef' added to Data.IORef
• modifySTRef' added to Data.STRef
• getExecutablePath added to System.Environment (requires peekFilePath, so this will need System.Posix.Internals.Compat to work)

base-4.7.0.0:

• byteSwap16, byteSwap32, and byteSwap64 added to Data.Word. Their implementations require functions in ghc-prim not available to earlier versions of GHC, but it should be possible to define equivalent implementations without using ghc-prim.
• eNOTSUP added to Foreign.C.Error (Edit: This is isn't possible to backport without involving a configure script, which isn't desirable.)
• showFFloatAlt and showGFloatAlt added to Numeric

base-4.7.0.2:

• inits (in Data.List) had a performance bug fixed
• trace (in Debug.Trace) fixed to handle null bytes properly

base-4.8.0.0:

• toIntegralSized added to Data.Bits (due to its use of bitSizeMaybe in its implementation, however, this can only be backported back to base-4.7.0.0)
• nub and nubBy's implementations (in Data.List) were changed such that the arguments to == and eq, respectively, are swapped
• fillBytes added to Foreign.Marshal.Utils

base-4.9.0.0

• Data.Ratio.{denominator,numerator} have no Integral constraint anymore
• Generalize Debug.Trace.{traceM, traceShowM} from Monad to Applicative
• Generalise forever from Monad to Applicative
• Generalise filterM, mapAndUnzipM, zipWithM, zipWithM_, replicateM, and replicateM_ from Monad to Applicative
• Data.Functor.Const module added
• errorWithoutStackTrace function added to Prelude (it's just the old behavior of error, I believe)
• forkOSWithUnmask function added to Control.Concurrent

base-4.10.0.0

• fromLeft and fromRight added to Data.Either
• gcoerceWith added to Data.Type.Coercion
• plusForeignPtr added to Foreign.ForeignPtr
• asProxyTypeOf's type has been generalized in Data.Proxy
• minimumBy/maximumBy are now implemented with foldl instead of foldr

base-4.11.0.0

• Semigroup((<>)) added to the Prelude
• (<&>) added to Data.Functor
• iterate' added to Data.List
• showHFloat to Numeric
• Type.Reflection.withTypeable is now polymorphic in the RuntimeRep of its result

base-4.12.0.0

• Data.Functor.Contravariant is now in base
• $! and throw are levity polymorphic

base-4.13.0.0

• MonadFail(..) added to the Prelude and Control.Monad (and Monad.fail removed)

base-4.14.0.0

• isResourceVanishedError, resourceVanishedErrorType, and isResourceVanishedErrorType added to System.IO.Error

base-4.15.0.0

• hGetContents', getContents', and readFile' added to System.IO
• singleton added to Data.List and Data.List.NonEmpty

base-4.16.0.0

• readBin and showBin added to Numeric
• oneBits added to Data.Bits

base-4.17.0.0

• pattern TypeRep added to Type.Reflection
• (.^.), (.>>.), (.<<.), (!>>.), and (!<<.) added to Data.Bits

base-4.18.0.0

• liftA2 added to Prelude
• traceWith, traceShowWith, and traceEventWith added to Debug.Trace
• inits1 and tails1 added to Data.List.NonEmpty
• minusNaturalMaybe added to Numeric.Natural
• applyWhen added to Data.Function
• mapAccumM and forAccumM added to Data.Traversable
• heqT added to Data.Type.Equality

base-4.19.0.0

• unzip added to Data.Functor
• (!?) added to Data.List
• unsnoc added to Data.List
• getSolo exported from Data.Tuple
• decTypeRep, decT, and hdecT added to Data.Typeable/Type.Reflection

base-4.20.0.0

• foldl' added to Prelude
• permutations and permutations1 added to Data.List.NonEmpty
• sortOn added to Data.List.NonEmpty
• List added to Data.List

base-4.21.0.0 (moved to #97)

• compareLength added to Data.List and Data.List.NonEmpty
• inits1 and tails1 added to Data.List
• firstA and secondA added to Data.Bitraversable

[bergmark]

What are your thoughts about the GHC modules? I'm missing for GHC.Stack.prettySrcLoc for example.

[RyanGlScott]

I would happily accept patches that added functionality from the GHC.* namespace. I haven't set out to do it myself since:

1. The GHC.* namespace is quite enormous.
2. The API tends to be quite volatile, so I imagine lots of CPP would be required to support it.
3. I personally don't have a need for it.

[sol]

@bergmark if you want to deal with call stack in a backward compatible way, you may want to look at https://github.com/sol/call-stack.

[RyanGlScott]

Now that base-compat has dropped support for pre-8.0 versions of GHC (see #96), the only unresolved items on this issue involve pre-8.0 GHCs and the as-of-yet-unreleased GHC 9.12. Going forward, I think it would be better to open individual issues for specific GHC versions. I'll opt to close this and track the GHC 9.12–related tasks separately in #97.