action | mux goes to read docs |
mux | goes to read docs |
dons | jcreigh: great. the focus bug seems to have gone. |
mux | it's a real pleasure to use gtk2hs the code is written quickly, as with Perl and the Gtk2 module, but you also get type-safety |
hpaste | mux annotated "Cairo and arithmetic" with "better formatting" at http://hpaste.org/1038#a3 |
Kobold_ | I have a goofy question about compiler implementation... (this seems like as good a place as any to ask) I have a really odd architecture that I have to deal with through all ASM code for about 30kLOC. It doesn't seem worth it to port one of the back ends of a real compiler to the arch. |
dolio | ?src liftM |
lambdabot | liftM f m1 = do { x1 <- m1; return (f x1) } |
quicksilver | dolio: thanks (let inside do) |
dolio | No problem. |
Kobold_ | But if I wanted to write a small compiler, would the fastest way just be to do an imperative language, and compile it to a stack machine? |
quicksilver | fastest in your time? or fastest in execution time? fastest in programmer time is probably a simple evaluator/interpreter... |
Kobold_ | Fastest in my time. I plan on generating piss poor code. :-P |
siti | haskell + parsec = dream |
mux | oh yes |
Kobold_ | I <3 parsec. This is a low level thing though, an interpreter would be a pain. |
siti | ok |
mux | Parsec is so great I'd like it to be named Parsex |
Kobold_ | mux: ++ |
mux | @karma+ Parsex |
lambdabot | Parsex's karma raised to 1. |
mux | to not be confused with `par` sex :D |
Kobold_ | I want to do a minimal functional language, like in wadler's implementing functional languages thing, but the price of admission seems to be higher. or rather, SPJ, not wadler |
quicksilver | Kobold_: rather depends how clever you're trying to be Kobold_: a 'naive' functional language implemented as a 'naive' graph reducer is very very easy Kobold_: most of SPJ's book is about making it less naive and more efficient |
Kobold_ | quicksilver: I'll have a real read through it. I've probably only spent an hour or two looking at it. Ideally I'd like the compiler to come in at under 1kLOC of haskell (which is a lot). |
quicksilver | Kobold_: an interpreter for lazy lambda calculus can be under 200 lines, including parsing and pretty-printing. Kobold_: add a little more if you have a complex type inference scheme in mind |
Kobold_ | quicksilver: na, typing was gonna be very simple. I think I will switch up to an interpreter though. The main problem is that as of now I don't have a real IO lib. :-D quicksilver: Essentially this is a project to build a modern computer infrastructure (up to a reasonably high language level) in a way comprehensible to an undergrad with a few mos. quicksilver: The processor is essentially a closer to modern ISA duct taped onto microcode via Verilog in an FPGA, duct taped onto an odd memory. :-P quicksilver: So I have to write routines even to write to the screen. |
narain | ?users |
lambdabot | Maximum users seen in #haskell: 336, currently: 303 (90.2%), active: 28 (9.2%) |
dolio | @quote qwe1234 |
lambdabot | qwe1234 says: just because it's inflammatory doesn't mean it's wrong |