Hi all, I have a quick question about ArrayArray#. Is it safe to store *both* an ByteArray# and ArrayArray# within the *same* ArrayArray#? For instance: - at index 0 of an ArrayArray# I store a different ArrayArray#, - at index 1 of that same ArrayArray# I store a ByteArray#. It seems to me that this should be safe/supported from the point of view of the runtime system: - both ArrayArray# and ByteArray# have the same kind/runtime representation, - the arrays have a header that tells rts/GC what they are/how to handle them. (But I, as a user, would be responsible for using the right primop with the right index to read them back) Is this correct? Thanks a lot! - Michal
I think Ed’s structs package explicitly makes use of this :)
On Wed, May 2, 2018 at 7:31 AM Michal Terepeta
Hi all,
I have a quick question about ArrayArray#. Is it safe to store *both* an ByteArray# and ArrayArray# within the *same* ArrayArray#? For instance: - at index 0 of an ArrayArray# I store a different ArrayArray#, - at index 1 of that same ArrayArray# I store a ByteArray#.
It seems to me that this should be safe/supported from the point of view of the runtime system: - both ArrayArray# and ByteArray# have the same kind/runtime representation, - the arrays have a header that tells rts/GC what they are/how to handle them. (But I, as a user, would be responsible for using the right primop with the right index to read them back)
Is this correct?
Thanks a lot!
- Michal _______________________________________________ ghc-devs mailing list ghc-devs@haskell.org http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
On Thu, May 3, 2018 at 2:40 PM Carter Schonwald
I think Ed’s structs package explicitly makes use of this :)
Oh, interesting! Thanks for the pointer! Looking at Ed's code, he's seems to be doing something similar to that I'm also interested in: having a SmallArray# that at one index points to another SmallArray# and at another one to a ByteArray#. (my use case involves multiple small arrays, so I'd rather use SmallArray# than ArrayArray#): https://github.com/ekmett/structs/blob/master/src/Data/Struct/Internal.hs#L1... So I guess my second question becomes: is anyone aware of some rts/GC invariants/expectations that would be broken by doing this? (ignoring the issue of getting every `unsafeCoerce#` right :) Thanks! - Michal
participants (2)
-
Carter Schonwald -
Michal Terepeta