<<03-11-2012>>
00:00:57dom96But it will get done :)
00:06:24*ryderblue quit (Remote host closed the connection)
00:48:12*q66 quit (Quit: Quit)
01:12:36*Trixar_za is now known as Trix[a]r_za
02:12:31*netrino_ quit (Quit: Ave!)
02:46:41*netrino quit (Ping timeout: 252 seconds)
06:15:17*Araq__ quit (Read error: Connection timed out)
06:17:30*Araq_ joined #nimrod
06:18:54*Trix[a]r_za is now known as Trixar_za
06:35:56*Araq_ quit (Read error: Operation timed out)
06:37:31*Araq_ joined #nimrod
07:37:53Trixar_zaWell, let's see if I can port this: http://byuu.org/programming/beat/
08:40:02AraqTrixar_za: why do you want to port it?
08:42:03Araqreactormonk: you can implement a lisp-like condition system if you want to; now that we have closures it's not hard; do you want to? I can guide you how to do it
08:44:40Trixar_zaBoredom. Also it's Linux code is written in C++
08:46:38Araqwell ... we surely have lots of things left to do that you could do instead
08:48:31Araqfor instance a webkit wrapper would be nice :-)
08:52:02*Araq_ quit (Read error: Operation timed out)
08:53:30*Araq_ joined #nimrod
09:15:27Trixar_zaThis is going to be painful
09:27:51Trixar_zahome/brenton/nimrod/c2nim/c2nim.nim(11, 41) Error: cannot open 'llstream'
09:27:57Trixar_zaSometime tells me I have to update Nimrod
09:28:02Trixar_zaSomething*
09:43:31Trixar_zaOo
09:43:37Trixar_zaIt's building faster than it used to
10:00:09AraqTrixar_za: strange error ... everything right with your paths?
10:07:43Trixar_zaNah, i just use the release copy to compile the development version's c2nim
10:07:50Trixar_zaused*
10:08:06Trixar_zaDownloaded and bootstrapped the one off github
10:08:09Trixar_zaAnd it worked fine
10:08:14Araqalright, good.
10:08:25Trixar_za... but I'm probably not going to do the webkit header
10:08:27Trixar_zaIt's huge
10:08:37Trixar_zaAnd in several different files
10:08:48Araqthere is a gtk wrapper around it that is way smaller
10:09:13Trixar_za163 files
10:10:07Trixar_za173 if you include the JavaScriptCore
10:10:08Trixar_za:P
10:19:09dom96good morning
10:25:26Araqhi dom96
10:41:51*Araq_ quit (Read error: Connection timed out)
10:42:31*Araq_ joined #nimrod
12:14:45*Trixar_za is now known as Trix[a]r_za
13:21:57*q66 joined #nimrod
13:36:06Araqso ... I want to have 1 release before christmas ;-)
13:39:15Araqany feature desparately needed?
13:39:49q66butts
13:41:17q66Araq, global type inference! :p
13:41:59Araqwe have clay-like type propagation :P
13:42:08Araqsomebody should test it though ...
13:42:26dom96continuations!
13:42:42dom96well perhaps not desperately needed
13:42:53q66continuations would be neat.
13:45:01Araqhrm ... continuations are expensive but I guess people don't care ...
13:48:11*Boscop quit (Ping timeout: 256 seconds)
13:51:14Araqkind of ironic how bad C's support for stack control is
13:51:58Araqthe reason is of course that C has not been designed as a "low level" language :P
13:52:24*Araq_ quit (Read error: Connection timed out)
13:53:33*Araq_ joined #nimrod
13:55:34*Boscop joined #nimrod
15:14:29*Araq_ quit (Read error: Connection timed out)
15:15:30*Araq_ joined #nimrod
17:16:26Araqso dom96 ...
17:16:32Araqyou promised improved docs :P
17:17:39dom96I'm kind of busy with nimbuild :P
17:34:02*Araq_ quit (Read error: Connection timed out)
17:35:29*Araq_ joined #nimrod
18:04:53*comex` joined #nimrod
18:06:17*ccssnet quit (*.net *.split)
18:06:18*comex quit (*.net *.split)
18:22:24*Araq_ quit (Read error: Connection timed out)
18:23:31*Araq_ joined #nimrod
18:46:49reactormonkAraq, sure, but not this week
18:47:18Araqtoo bad then :P
18:47:26Araqcause it looks like 3 lines of code to me
18:47:32Araqand I feel like doing it :P
18:47:51reactormonkAraq, but the main point is to define conditions for the errors in the stdlib - having a feature without stdlib support is a useless feature
18:48:23Araqhm what do you have in mind?
18:57:44reactormonkgo through every 'raise' statement in the stdlib and see if there's a condition that would fit
18:57:52reactormonkerr, not condition, but restart
18:58:54Araqwell hm
19:01:15reactormonkotherwise we have the car that runs on 5% of the streets
19:02:40reactormonkand a default if the condition is uncaught - one that you can change. CLI apps want a different one than webapps than GUI ...
19:02:56reactormonkand warnings don't abort if uncaught
19:05:12*Araq_ quit (Read error: Operation timed out)
19:07:33*Araq_ joined #nimrod
19:09:13Araqwell the stdlib hardly has any UI support
19:09:33Araqand if 'write' fails with an IO exception I'm not sure we need a 'retry' mechanism for that
19:09:51reactormonkwrite to a file?
19:09:57Araqyeah
19:10:28reactormonkwrite to a different fd, open a new fd and write to that one
19:10:44reactormonkactually just a) would be enough, you can do b) yourself then
19:10:55Araqwhen was the last time you needed to do that? :P
19:11:19reactormonkyou might have a point there
19:12:17Araqwell I guess I should simply explain how I'm about to implement the feature:
19:12:59Araqcurrently a 'try' statement produces a hidden 'SafePoint' object on the stack
19:13:36Araqa 'raise' statement then jumps to that (uses C's setjmp/longjmp)
19:14:11AraqI will add a closure field to the SafePoint data structure
19:14:38Araqand 'raise' will check if it's set and execute that instead of longjmp'ing
19:14:48Araqwhich means 'raise' will not unroll the stack then
19:15:35Araqit will depent on the type of exception being raised though
19:16:11Araqthere will be a built-in proc named 'handle' to set this closure field
19:16:17Araqso it will look like:
19:16:18Araqtry:
19:17:08Araq handle(EIO, proc () = WriteToDifferentFile(mydata))
19:17:16Araq WriteToFile(myData)
19:17:20Araqfinally:
19:17:26Araq close(x)
19:18:01Araqsince every 'raise' is potentially not unwinding the stack, I don't need to go through every 'raise' statement in the stdlib
19:19:49Araqeven nicer would be if a 'handle' section would part of the 'try' statement but that's much more work to implement
19:20:11Araqand would hide the closure construction which is a bit troublesome in a systems programming language
19:20:54Araq(and yeah my example is silly)
19:34:42reactormonkso you don't offer restarts from a raise?
19:35:09reactormonkI propose a syntax like
19:35:12reactormonktry:
19:35:16reactormonk WriteToFile(myData)
19:35:19reactormonkhandle EIO:
19:35:40reactormonk ChangeFD(something_else)
19:35:43reactormonkfinally:
19:35:49reactormonk close(x)
19:36:51Araqwell syntax aside, where's the difference?
19:37:13Araqyou can 'restart' in the callback
19:37:33reactormonkthere is none :-)
19:38:07Araqyour syntax is much nicer, I agree
19:38:13Araqit's also more work :P
19:38:32reactormonkimho both should be possible
19:38:48Araqand 'handle' can't become a keyword easily I think
19:38:56Araqwindows' api is full of handles
19:38:58reactormonkgo with 'catch' then
19:39:33Araqtoo confusing :P
19:39:37reactormonkI have to admit I prefer the ruby terminology with begin/rescue and throw/catch
19:40:02reactormonkwhy? if you don't call a restart, do a stack unroll
19:50:47*Araq_ quit (Read error: Connection timed out)
19:51:17Araqwell if you don't call 'handle' in my proposal, you'll get a stack unroll
19:51:33*Araq_ joined #nimrod
19:52:11Araqcan't see what's wrong with it
20:01:56*fowl quit (Ping timeout: 248 seconds)
20:02:10reactormonknow way to merge them?
20:03:38reactormonks/now/no/
20:04:06Araqmerge what?
20:04:15reactormonkthe two keywords - catch and handle
20:13:41*fowl joined #nimrod
20:16:44Araqhu? currently nimrod has neither
20:17:34reactormonkoh.
20:17:46reactormonkhow do you rescue/catch/whatever raises then? :-)
20:22:32Araqwith a built-in 'handle' proc
20:22:39Araqno keyword :P
20:29:41*Araq_ quit (Read error: Connection timed out)
20:31:31*Araq_ joined #nimrod
20:37:21*ccssnet joined #nimrod
20:44:01*Araq_ quit (Read error: Operation timed out)
20:45:29*Araq_ joined #nimrod
21:08:39*Araq_ quit (Read error: Connection timed out)
21:09:32*Araq_ joined #nimrod
23:02:15*Araq__ joined #nimrod
23:03:23*Araq_ quit (Ping timeout: 260 seconds)
23:34:15*comex` is now known as comex