#haskell - Mon 19 Feb 2007 between 11:10 and 11:34

NY Lost Funds



EvilTerranwell, module Monad (Control.Monad(..)) where import Control.Monad , but yeah.
ndmEighty: http://darcs.haskell.org/packages/haskell98/Monad.hs - thats the actual soruce code
allbery_bmodule Monad (module Control.Monad) where import Control.Monad -- I think
EvilTerranok, there's a bit more too it than i thought.
*to
ndmnot a whole lot more!
EvilTerranbut it's the same concept.
ndmits just restricting the export list
Eightywhy should i use Control.Monad instead then? will Monad go away? isn't hoogle just a bot in here?
EvilTerran@where hoogle
lambdabothttp://www.haskell.org/hoogle
ndm@seen dons
lambdabotdons is in #haskell. I last heard dons speak 1h 22m 23s ago.
EvilTerrannot just a bot. but, that aside, heirarchical modules are The Way Forward.
allbery_bControl.Monad will be in Haskell' I think
maskddoes anyone have solutions to "write yourself a scheme in 48 hours" exercises? some of them are rather difficult
allbery_bthink of the old namespace (things like Monad) as backward compatibility with ancient versions of Haskell, most of which no longer exist in the real world :)
Eightyokay
boegeldons: great work on nofib!
or nobench, rather :-)
ndmdons: none of the Yhc developers have working copies of nhc - it doesn't work on 64 bit, the distro in our department, or windows - which pretty much covers us all...
Eightyndm: you made hoogle? could you make a text link "next" in addition to the arrows? then fast-forwarding would work in opera
Igloondm: Why do you think yhc is faster, then?
ndmEighty: doesn't opera use the link type="next" field?
Igloo: because way back when we were able to benchmark it, on one developers machine (before he got a 64bit processor), we were reliably 20% faster - the code is sufficiently similar that we know why we are faster
Eightyi don't know... fast-forwarding doesn't work as it is. adding "next" there made it work.
ndmEighty: yes, i did make hoogle, and i'll add the appropriate link tags when i get a chance - probably for hoogle 4 though since i've pretty much broken hoogle 3, and can't rebuild working copies...
Igloo: we share the same front end, and hte same basic code generator - we just use less instructions, and each instruction is simpler - there is no where we could potentially be slower (in theory...)
Eightyokay
dino-Do the arrow functions like (>>>), (<+>), (***), (&&&) have names? Like how we call (>>=) "bind"
I'm finding them hard to read and discuss with people out loud without names.
donsndm, i'll do a full build of yhc in the morning, and give you the log?
ndmdons: that would be handy, and let us know when you have a complete benchmark set with yhc type=release done, so we can check it
dons: Tom thinks the numbers you are giving suggests you have got -O2 specified - but he'd like to see all the comparative numbers
donsall i've done is a scons type=release; scons install. so should i have set other flags?
ndmdons: that should be plenty, you might need scons fullclean or scons clean beforehand if its an existing repo
dons: are the numbers on the web now with type=release?
vincenzanyone know how hard it is to build ghc under cygwin?
kombinatordino-, (>>>) is serial composition, (***) is parallel composition for instance
donsndm, nope. let me copy across the results so far

Page: 1 8 15 22 29 36 43 50 57 64 71 78 85 

IrcArchive

NY Lost Funds