On 07/11/2009, at 03:10, Paolo Losi wrote:
Don Stewart wrote:
I'd be happy to talk more about the design of the library, if you like.
Don,
I would be personally grateful if you could talk about the design of the library and/or point to some comprehensive documentation.
Can you confirm that uvector is going to stay almost api compatible with dph, and that the knowledge investment is going to be "reusable" on dph?
uvector has (almost) nothing in common with DPH's API. It is forked off the flat sequential array layer which DPH uses internally and which the users aren't supposed to even know about. Also, the fork happened quite a while ago, DPH has changed a lot since then and is going to change a lot more in the future. My plan is to eventually use my vector package to replace those flat arrays but sadly I don't have a lot of time to work on it (although vector is quite usable by now and even implements recycling which should improve DPH's performance by quite a bit). The fact that everything DPH depends on will have to be distributed with GHC doesn't help, either, since adding a new package into the mix is a pretty big step. Roman