#3480: Easily make Typeable keys pure, so that Typeable can be handled
efficiently across communications
-----------------------------+----------------------------------------------
Reporter: guest | Owner:
Type: task | Status: new
Priority: normal | Milestone: 6.14.1
Component: libraries/base | Version:
Resolution: | Keywords: Typeable, efficiency
Difficulty: Unknown | Os: Unknown/Multiple
Testcase: | Architecture: Unknown/Multiple
Failure: None/Unknown |
-----------------------------+----------------------------------------------
Changes (by simonpj):
* failure: => None/Unknown
Comment:
I talked to Simon M about the question of tidying up GHC's `TypeRep`
mechanism. A good start would be the following:
1. Generate a fingerprint from the `String` (including the fully-
versioned module) of the `TyCon` (eg "`base-4.3.2.2:GHC.Arr.Array`")
2. At each `App` node in the `TypeRep` generate a fingerprint from the
fingerprints of the component pieces. We do this all the time when
generating fingerprints in interface files. (This would retain constant-
time comparison for `TypeRep`.)
This would completely deal with the single-program case, because in any
one program there is definitely only one type constructor
`base-4.3.2.2:GHC.Arr.Array`. If we wanted a stronger, cross-program,
story, then we'd need a stronger fingerprint under (1), based on the type
structure as outlined in an earlier comment. That is quite doable too,
but would probably require us to inject a binding for each `TyCon`
fingerprint only ''after'' they'd been computed in `MkIface`. A bit more
plumbing, that's all.
This would be a straightforward project for someone to do. Since the
fingerprinting for `TypeRep` must be done in `Data.Typeable`, which is in
the `base` package, we'd need to add fingerprinting technology to `base`.
Not hard, since GHC itself already has it, so we can just copy it over.
But we don't want to make `base` depend on the `binary` package, so
something a bit less general than GHC's fingerprinting interface is
wanted.
Any volunteers?
Simon
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/3480#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
