Sean Seefried <sseefried@cse.unsw.edu.au> writes:
Wow, what a crazy idea. Is the basic idea that you have a program that never exits the TH compile loop and when dynamic loading is required it reads in a source program, places it inside reification brackets and then tries to splice it in? Sure this might work, but what about unloading such code?
Ok, I have no idea how to reload existing code. Shadow existing definitions and compile more?
Surely this is an important part of hs-plugins? This reminds me of getting C++ templates to do things they should be doing - it might work but it's surely not the most elegant or user friendly solution.
True, but I can't think of another solution that carries around the entire typechecker. Lauri Alanko mentioned that hs-plugins is only as type-safe as Dynamics, this was my proposed improvement. This also beats the problem of having two different type-tables, one static and one dynamic, and never the twain shall meet. Even so, Don's recent solution of dynloading everything on a tiny static core is better there. Ok, maybe it's not such a great idea :-) So, next crazy TH idea. Can I use today's version of rType[1] to dump a TH-evaluable type representation that I can send across a network connection? In essence, can I make mobile types by using TH to dump the type description, and hs-plugins to compile and load those type descriptions on another running RTS that's never seen this type before? Even if that works, I'm still not sure how to transport functions to actually work on the new datatypes... [1] see rType in this post, http://www.haskell.org/pipermail/template-haskell/2003-September/000172.html -- Shae Matijs Erisson - Programmer - http://www.ScannedInAvian.org/ "I will, as we say in rock 'n' roll, run until the wheels come off, because I love what I do." -- David Crosby