[EMAIL PROTECTED] (Kevin Hammond) writes:
It sounds like if we allow empty declarations, then we have to fix the
layout rule. Is that the case?
I think the layout rule is still fine (there are already empty declarations
in some cases, e.g. class).
Empty declarations in classes are
Regarding my previous mail:
The problem is caused by the definition of catchall that uses the
operation catch of the IO monad rather than the more general class
operation trap of ErrorMonad.
The previous definition is:
catchall c1 c2 = catch c1 (\_ - c2)
which should be:
--
X-Sun-Data-Type: text
X-Sun-Data-Description: text
X-Sun-Data-Name: text
X-Sun-Charset: us-ascii
X-Sun-Content-Lines: 27
While experimenting with multiple parameter type classes, I
introduced an ErrorMonad. Compiling this with GHC 3.00, the
following happened:
Computation.hs:16:
Marc van Dongen= writes:
While compiling some happy output with 3.00 I came
across the following:
NOTE: Simplifier still going after 4 iterations; bailing out.
NOTE: Simplifier still going after 4 iterations; bailing out.
panic! (the `impossible' happened):
fun_result_ty:
Simon Marlow [EMAIL PROTECTED] writes
Best way around it is to use a newtype:
newtype ListBlah = ListBlah [Blah]
instance Show ListBlah where
Ok, it's a bit of a pain to put the extra constructors everywhere, but
at least you don't lose any efficiency.
Is that `True'? Does the
Ralf Hinze [EMAIL PROTECTED] writes:
newtype ListBlah = ListBlah [Blah]
instance Show ListBlah where
Ok, it's a bit of a pain to put the extra constructors everywhere, but
at least you don't lose any efficiency.
Is that `True'? Does the compiler spot identity functions like
--
X-Sun-Data-Type: text
X-Sun-Data-Description: text
X-Sun-Data-Name: text
X-Sun-Charset: us-ascii
X-Sun-Content-Lines: 27
The following works fine with GHC 3.00:
class Variable v where
updVar :: v a - (a - IO (a,b)) - IO b
applyVar :: Variable v = v a - (a - a) - IO a
At 11:17 am 5/2/98, Simon Marlow wrote:
Ok, that sounds like a reasonable fix. I'd be happier if the
semicolon was constrained to being one that had been inserted by the
layout system, though.
That just needs you to use different tokens for the two kinds of semicolon
(plus a production for the
While experimenting with multiple parameter type classes, I
introduced an ErrorMonad. Compiling this with GHC 3.00, the
following happened:
Computation.hs:16: Warning:
`ErrorMonad' mentioned twice in export list
I'll look into this.
panic! (the `impossible' happened):
The following class causes the generated C Code to cause an error:
class (OMSObjectC o,OMSObjectC o',OMSLinkC l,OMSLnkAttrC a,OMSAttrValC av) =
OMSLnkAttrAppC l o o' a av where
lAGet :: IO (a l o o') - IO av
lAGet' :: (a l o o')- IO av
lASet :: IO (a l o o') - av -
panic! (the `impossible' happened):
fun_result_ty: 6 GHC.Int#{-3e-}
- GHC.Int#{-3e-}
- b_trKC
- PolyParse.HappyState{-rq9-} b_trKC c_trKD
- [PolyParse.HappyState{-rq9-} b_trKC c_trKD]
-
I know that ghc performs hardly any optimisation if called without -O or -O2.
However, is it necessary that just adding a single function definition to a
program *without* calling this function increases execution time by 50%?
(ghc-2.08)
Example:
{-
fromStepTo :: Int - Int - Int - [Int]
The following works fine with GHC 3.00:
class Variable v where
updVar :: v a - (a - IO (a,b)) - IO b
applyVar :: Variable v = v a - (a - a) - IO a
applyVar v f = updVar' v (\x - let x' = f x in (x',x'))
By changing the definition of applyVar
Nice point! For the class decl
class (C a, D a) = E a where {...}
we use to generate superclass selectors:
scsel_E_C :: E a - C a
scsel_E_D :: E a - D a
But now there can be class decls like yours:
class (C a, C b) = E a b where {...}
and our naming
14 matches
Mail list logo
