[project @ 2001-08-23 10:53:46 by simonmar]

[ghc-hetmet.git] / ghc / tests / specialise / COMMENTS

diff --git a/ghc/tests/specialise/COMMENTS b/ghc/tests/specialise/COMMENTS
deleted file mode 100644 (file)
index 3ff1339..0000000
--- a/ghc/tests/specialise/COMMENTS
+++ /dev/null
@@ -1,43 +0,0 @@
-CHAR I/O:
-*********
-
-In clausify ...
-
-The unifier would like to propogate use of Char#s all the way to the
-readChan and appendChan. However these have explicit [Char] arguments
-so we must explicitly coerce the Chars as we extract them.
-       clause produces [Char#]s
-       parse reads [Char] and builds Sym Char#
-       disp takes [Char#]s and produces [Char]
-
-COMMENTS:
-* The extent of this unboxification is quite surprising and possibly
-  unexpected.
-* Coersion when constructing or extracting from unboxed structures can
-  be a pain. Where this occurs defines the extent of the unboxedness.
-
-OVERLOADING CHARS:
-
-Might want to introduce versions of I/O operations which read/write
-[Char#]. Use a type class to capture either boxed or unboxed Chars.
-
-class Char a 
-  toChar :: a -> Char
-  fromChar :: Char -> a
-
-instance Char Char
-  toChar   = id
-  fromChar = id
-
-instance Char Char#
-  toChar c#  = MkChar c#
-  fromChar c = case c of MkChar c# -> c#
-
-Now rather than specifying type as
-   ... Char ...
-We use
-   Char c => ... c ...
-
-Now just need a specialised versions I/O operations which deal with [Char#]
-
-The Char class is very similar to the overloading of numeric constants.