>>>>> Simon Marlow <[EMAIL PROTECTED]> writes:
>> Another issue that I've encountered on the same BLAS project:
>> is there an officially sanctioned way to get at the ByteArray#
>> or MutableByteArray# underlying the UArray and STUArray types?
>> When I import ArrayBase, getting at the ByteArray# of a UArray
>> is no problem. However, things are not as simple for an
>> STUArray. I have resorted to using my own slightly hacked
>> version of MArray in order to get at the MutableByteArray#. Is
>> there some better method to do all this?
Simon> I'm not sure why you say things aren't as simple for
Simon> STUArray - the definition of this datatype is exported from
Simon> ArrayBase the same as UArray.
My mistake. I threw in the whole `kitchen sink' in trying to solve my
problems , and I neglected to narrow down my imports to determine what
would be really necessary.
In tracking this down, I did notice that the IOUArray data constructor
is not exported by MArray (and that is why I produced my own version
that does export the data constructor). What would be the recommended
method to convert an STUArray into a IOUArray (without copying)?
Essentially, I have a function that returns ST s (STUArray s _ _), and
I want to take the returned value into the IO monad as an IOUArray.
Simon> Anyway, you're right that there's no easy way to pass these
Simon> arrays to foreign functions, like you can with the old
Simon> MutableArray and ByteArray types (at least for "unsafe" FFI
Simon> calls, 'cause the array lives in the Haskell heap). Now
Simon> that MArray/IArray have settled down, it's probably worth
Simon> adding this support to the compiler.
For my concerns, that would be great.
Cheers,
Martin
