00:04 (quit) rbarraud: Read error: Connection reset by peer
00:08 (part) shofetim: "ERC Version 5.3 (IRC client for Emacs)"
00:28 (join) jonrafkind
01:27 (join) hellmage
01:30 (join) Rakko
01:36 (quit) emma: Ping timeout: 252 seconds
01:37 (join) emma
01:39 Rakko: Newbie question: is the command-line racket program supposed to be able to run scripts that start with #lang? I can only get such scripts to run in DrRacket.
01:41 (quit) hellmage: Ping timeout: 252 seconds
01:42 (quit) emma: Ping timeout: 252 seconds
01:48 jonrafkind: Rakko, yea you can do that
01:49 (join) emma
01:54 Rakko: jonrafkind: I can't get it to work.
01:54 (join) rbarraud
01:54 jonrafkind: can you give more details
01:55 jonrafkind: times up, time for sleep
01:56 Rakko: oh, damn
01:56 Rakko: was gonna give you a dl link
01:56 Rakko: night
01:56 Rakko: http://rakko.homeunix.com:8089/~eric/scmindent.scm -- for anyone interested
01:57 Rakko: if I try to run it either by just typing its name or by racket -r scmindent.scm, and feed it a scheme file to indent as input, it waits a second or so and then exits
01:57 Rakko: but if I take out #lang scheme at the beginning it works fine
01:58 Rakko: oh, I should add that it doesn't matter if I use #lang racket or #lang scheme
02:00 (quit) jonrafkind: Ping timeout: 264 seconds
02:38 (quit) Rakko: Quit: Rakko
03:07 (join) writer
03:30 (join) samth
03:30 samth: join #scheme
03:33 (join) stamourv
03:34 (join) hanDerPeder
04:44 (quit) rbarraud: Read error: Connection reset by peer
04:48 (join) rbarraud
05:19 (join) hellmage
06:00 (quit) hellmage: Ping timeout: 240 seconds
06:24 (join) hellmage
06:45 (quit) stamourv: Quit: ERC Version 5.3 (IRC client for Emacs)
06:49 (quit) emma: Remote host closed the connection
06:54 (quit) samth: Ping timeout: 255 seconds
06:58 (quit) rbarraud: Ping timeout: 245 seconds
07:06 (join) emma
07:12 (quit) hellmage: Ping timeout: 272 seconds
07:29 (quit) emma: Remote host closed the connection
07:30 (join) emma
07:37 (quit) emma: Ping timeout: 276 seconds
07:52 (quit) jao: Ping timeout: 265 seconds
07:52 (join) emma
08:15 (quit) emma: Ping timeout: 264 seconds
08:22 (join) samth
08:28 (join) emma
08:40 (join) chturne
08:47 (join) stamourv
08:52 (quit) emma: Ping timeout: 264 seconds
08:52 (join) emma
08:58 (quit) emma: Remote host closed the connection
09:06 (join) emma
09:10 (quit) emma: Read error: Connection reset by peer
09:11 (join) emma
09:16 (quit) emma: Ping timeout: 276 seconds
09:23 (join) emma
09:43 (join) b-man_
09:56 (join) jeapostrophe
10:01 samth: jeapostrophe, have you seen this kind of error before:
10:01 samth: require: namespace mismatch; reference (phase 1) to a module "/home/samth/sw/plt/collects/typed-scheme/typecheck/def-export.rkt" that is not available (phase level 1) in: renamer
10:01 samth: renamer is something i've defined
10:01 samth: but i've never gotten that error
10:02 jeapostrophe: i've never gotten it but i think i can decipher it
10:06 (join) hellmage
10:12 (quit) jeapostrophe: Quit: jeapostrophe
10:32 (quit) b-man_: Remote host closed the connection
10:32 (join) b-man_
11:01 samth: jay-mccarthy, you there?
11:01 samth: if you have a deciphering, that would be great
11:23 (quit) chturne: Remote host closed the connection
11:29 (join) jonrafkind
11:32 (quit) hanDerPeder: Quit: hanDerPeder
11:37 (join) chturne
11:39 jay-mccarthy: samth: i think it is saying that the def-export macro is used in a place where def-export.rkt is not required, because renamer is not actually bound inside the macro
11:39 samth: ah, ok
11:39 jay-mccarthy: i don't think define-for-syntax works that way
11:39 jay-mccarthy: (you basically want (for-template (for-syntax renamer))
11:39 samth: ah, that makes sense
11:40 jay-mccarthy: but since you can't write that, i think you need to put renamer in another file and get it for-template
11:54 samth: hmm, that didn't work
11:55 samth: now i'm getting the same error, but the second phase 1 is now phase 0
11:56 (join) jeapostrophe
11:57 (join) _ryanc_
11:58 (part) stamourv: "ERC Version 5.3 (IRC client for Emacs)"
11:58 jay-mccarthy: are you requiring for-template in def-export?
11:59 samth: `renamer' actually occurrs on the RHS of a define-syntax
12:00 samth: so it should normally just be required `for-syntax'
12:01 samth: i tried requiring it at phase 0 but that didn't help
12:03 jay-mccarthy: but it appears in a syntax template for a define-syntax rhs
12:03 jay-mccarthy: so from the perspective of def-export it is using it as a template
12:03 jay-mccarthy: maybe i'm crazy
12:03 samth: adding the for-template require didn't help
12:04 jay-mccarthy: requiring it normally doesn't help either?
12:04 jay-mccarthy: and in the other file is it define'd or define-for-syntax'd
12:05 samth: in the other file it's defined
12:06 samth: ok, i got it to work, by requiring "renamer.rkt" from typed-racket itself
12:06 samth: I think I can now construct a small example of the confusing behavior
12:07 samth: thanks for your help
12:12 (quit) jeapostrophe: Quit: jeapostrophe
12:13 _ryanc_: samth, how were you inserting the reference to renamer?
12:17 jay-mccarthy: _ryanc_: http://github.com/plt/racket/blob/master/collects/typed-scheme/typecheck/def-export.rkt
12:17 (notice) rudybot: http://tinyurl.com/3a465dd
12:18 (quit) samth: Ping timeout: 240 seconds
12:19 (join) hanDerPeder
12:20 _ryanc_: I seem to recall this error coming up before... I wonder if there was a regression in the module instantiation code.
12:21 jay-mccarthy: maybe it is a bug in the lazy instantiation?
12:22 _ryanc_: IIRC, it had to do with the order that modules were instantiated in... but that might have been a different error ("used before definition"?)
12:23 _ryanc_: anyway, sounds like samth is boiling it down to something smaller
12:32 jonrafkind: _ryanc_, should this file exist? collects/syntax/private/stxparse/runtime-prose.rkt
12:32 jonrafkind: i just did a pull --rebase and its missing from my tree
12:33 (quit) writer: Quit: writer
12:34 _ryanc_: jonrafkind, I moved it to syntax/parse/private/runtime-report.rkt
12:35 jonrafkind: hm, why am i getting an error about it not being found
12:35 jonrafkind: i ran make install..
12:35 jonrafkind: ill rebuild the syntax collects
12:37 _ryanc_: had you changed anything in syntax/*?
12:38 jonrafkind: nah
12:39 jonrafkind: bleh same error..
12:39 jonrafkind: it would be nice if the load handler told me who invoked it
12:39 _ryanc_: grep :)
12:40 jonrafkind: ok it was in my own file somehow
12:42 _ryanc_: ah... maybe you had changed current-fail-handler to debug something at some point?
12:42 jonrafkind: oh yea, that was probably why
12:42 jonrafkind: ok i erased the occurences in my own source, but the problem persists
12:43 _ryanc_: did you re-run setup?
12:43 jonrafkind: well i just erased all the compiled directories instead
12:43 _ryanc_: grep :)
12:43 jonrafkind: yea.. grepping the whole tree now
12:43 jonrafkind: nothin
12:44 _ryanc_: re-run setup for real this time
12:44 jonrafkind: ok
12:49 (join) uygur
12:53 uygur: Hello out there, I am from Java camp and played with Clojure a bit. I also found HOP web developement. Is there anything like HOP in Racket? I am also interested using Racket for scripting. Any guidance is appreciated.
13:02 jonrafkind: hop = higher order functions?
13:03 fjl: uygur:  the ubiquitous package repository for racket lives at http://planet.plt-scheme.org/, maybe you'll find something there...
13:03 clklein: jonrafkind: probably hop.inria.fr
13:03 fjl: jonrafkind: HOP is Bigloo's web framework thing: http://hop.inria.fr/
13:04 jonrafkind: well thats an awful choice for a name.. I thought hop was pretty standard for "higher order programming"
13:05 fjl: uygur: but as far as i know, nothing as comprehensive as HOP exists for Racket
13:07 fjl: uygur: if you're into web stuff, you should really try working with the web server that comes with Racket (http://docs.racket-lang.org/web-server/index.html).
13:20 (join) spidermario
13:24 (quit) hanDerPeder: Quit: hanDerPeder
13:28 (join) samth
13:28 (part) hellmage
13:29 samth: _ryanc_, you still around?
13:38 _ryanc_: samth, yes
13:39 samth: so, what was happening is that I had a macro which expanded to a use of renamer
13:39 samth: and a use of that macro was in the output of TR's #%module-begin
13:40 samth: and then I changed TR's module-begin to dynamic-require its implementation
13:40 samth: and then i got that error
13:40 samth: which i fixed by having the same module that defined #%module-begin require the def of renamer
13:41 samth: _ryanc_, hope that makes sense
13:42 _ryanc_: aha
13:43 _ryanc_: I suspected something of the sort, but I couldn't figure out how you were disconnecting the syntax from the require chain.
13:44 _ryanc_: Why dynamic-require?
13:44 samth: i would really like to have a way to say "this for-syntax computation doesn't need to be run at runtime"
13:44 samth: so that I can avoid loading the TR implementation at runtime
13:44 samth: right now, #lang typed/racket/base 1 is an order of magnitude slower than #lang racket/base 1
13:45 samth: even after precompiling
13:45 _ryanc_: when run as a script via racket, or run in drracket, or what?
13:46 samth: as a script on the command line
13:46 _ryanc_: lazy instantiation is supposed to help with this... apparently not enough?
13:46 samth: right, not nearly enough
13:46 _ryanc_: hmm, I wonder why not
13:47 samth: part of the problem is that YWIW works against lazy instantiation
13:47 samth: since you're inserting compile-time code that runs when the program runs
13:47 samth: so everything at phase 1 gets loaded
13:47 _ryanc_: I'm not sure exactly what you mean, but I think I disagree :)
13:48 _ryanc_: no, the idea is to only evaluate the phase 1 code if phase 1 is needed (by eval, for example)
13:49 samth: my understanding from reading the docs is that if you execute any phase 1 code, it loads all the phase 1 modules
13:49 (join) jao
13:49 _ryanc_: but "racket myprog.rkt" shouldn't execute any of the phase 1 code of myprog.rkt if it's already compiled (I think)
13:50 samth: ah, hmm, you seem to be right
13:50 samth: so then i'm not sure why this is happening
13:51 _ryanc_: TR has essentially no runtime support beyond contracts, right?
13:51 samth: right
13:51 _ryanc_: in other words, the goal should be "as fast as racket + racket/contract"
13:51 samth: no, as fast as racket/base + racket/contract
13:51 samth: which is a little faster
13:51 _ryanc_: yes, sorry
13:52 samth: but I'd take the former in a heartbeat right now
13:52 _ryanc_: perhaps you have unnecessary requires that cause things that are only needed at phase 1 to be loaded at phase 0 too
13:53 _ryanc_: we need a tool for inspecting and pruning module require graphs
13:53 samth: yeah, the module browser is ok but not great
13:53 samth: and really slow
13:54 _ryanc_: IIRC, it's not good at showing phases, and it doesn't help at all for detecting useless requires
13:55 samth: right
13:57 _ryanc_: maybe I can come up with something
13:59 samth: i think there are some places where i have excess requires
14:01 samth: i have a suspicion that there are some places where i'm doing excess requires to make the tests work
14:02 _ryanc_: actually, this may be a good fit with Jay's work
14:02 samth: yes, i think so
14:03 _ryanc_: There are two ways to approach the code to analyze what gets used: watch expansion using the macro-stepper tools, or inspect the compiled code. The first might have benefits, but I'm not sure yet if they actually help in this task.
14:06 _ryanc_: well, they might. Suppose a program used match, which expands into pure racket code (pretend match:error doesn't exist). If a macro doesn't leave any run-time references in a module, the macro's module might get marked as useless by a compiled-code analyzer.
14:07 samth: yeah
14:07 _ryanc_: although watching expansion has blind spots too, like side-effects
14:08 samth: although, i want to make sure that laziness is doing enough for me, so comparing the too might be instructive
14:10 _ryanc_: I'll look into it...
14:10 _ryanc_: after lunch
14:11 samth: ok
14:11 samth: it's after dinner here, so hurry up :)
14:12 samth: (i'm in the netherlands at ifl)
14:12 samth: if there was just a way to print every visited module, that would be helpful
14:13 _ryanc_: that should be easy to predict, harder to monitor
14:14 samth: but monitoring would be the really useful bit
14:16 _ryanc_: only if the prediction is wrong :)
14:16 _ryanc_: although I suppose monitoring might also provide a way of timing module instantiation, which would be useful on its own
14:17 _ryanc_: anyway, I'm off to lunch now
14:26 samth: maybe i can use `current-load'
14:29 (join) carleastlund
14:59 jay-mccarthy: samth: did neil do okay?
15:00 jay-mccarthy: _ryanc_: we already prune non-0 module references during the batch compiling
15:00 samth: his talk isn't for 2 days
15:01 jay-mccarthy: i thought it was on wednesday
15:01 jay-mccarthy: nevermind then :)
15:01 samth: jay-mccarthy, will the batch compiler work on TR?
15:01 jay-mccarthy: it already does
15:01 samth: :)
15:02 jay-mccarthy: (as much as it works on anything else which is about 90% probability on any given zo)
15:02 clklein: Can it demodularize itself yet?
15:03 samth: afaict, TR is always in the 10% for all tools
15:03 jay-mccarthy: oh ya
15:03 clklein: yay!
15:03 jay-mccarthy: well, we are currently working on a bug involving doubly cyclic closures
15:03 clklein: doubly cyclic?
15:04 clklein: captures two references to itself?
15:04 jay-mccarthy: a cyclic closure that contains another cyclic closure
15:04 clklein: ah
15:05 clklein: And when you say "closure" you mean the strangely named thing whose defining property is that it does NOT capture anything, right?
15:05 jay-mccarthy: ya :)
15:07 clklein: Robby insisted that we rename those things in the model.
15:14 (quit) emma: Remote host closed the connection
15:20 (join) emma
15:22 (join) shofetim
15:32 samth: btw, carleastlund, john woo is an uncredited director of many of the talks here
15:36 (quit) emma: Ping timeout: 240 seconds
15:38 (join) emma
15:40 _ryanc_: jay-mccarthy, I know you eliminate phase non-0 references, but I suspect the issue here is actual requires at phase 0 that aren't used. Perhaps you can prune those too, but it would be nicer to clean up the code; that's why an analysis tool would be useful.
15:40 samth: _ryanc_, try the following to see the problem
15:40 samth: [samth@hermes:~/tmp] cat testtr.rkt
15:40 samth: #lang typed/racket/base
15:40 samth: 1
15:41 samth: [samth@hermes:~/tmp] cat /home/samth/.racketrc
15:41 samth: (let ([loader (current-load)])
15:41 samth: (current-load (lambda args
15:41 samth: 		  (displayln args)
15:41 samth: 		  (apply loader args)
15:41 samth: 		  (displayln "done"))))
15:41 samth: [samth@hermes:~/tmp] r
15:41 samth: Welcome to Racket v5.0.1.4.
15:41 samth: > (require "testtr.rkt")
15:41 samth: you'll see a lot of output
15:42 _ryanc_: ah, that's very different from running "racket testtr.rkt", though
15:43 samth: no, i think it's the same thing
15:44 samth: if you do the same for a file written in racket/base, there's no loading other than the directly-requireed file
15:44 carleastlund: samth: bang, bang.
15:45 samth: carleastlund, i was killed multiple times
15:45 carleastlund: 'Bout time.
15:48 samth: oh, and _ryanc_, switching to dynamic-require of the implementation of #%module-begin reduces that output considerably
15:53 jay-mccarthy: _ryanc_: i eliminate all unused top levels which can mean removing all of a require's top levels
15:57 _ryanc_: samth, jay-mccarthy, after playing with the example Sam sent, I realized (and Matthew confirmed) that only *instantiation* happens lazily; all of the modules are still read in.
15:58 _ryanc_: that's why the load handler is printing everything out
15:58 samth: why does the load handler not print anything for something written in plain racket?
15:59 _ryanc_: because it's already loaded
15:59 samth: since that requires a bunch of files (namely, the implementation of `racket')
15:59 samth: ah, of course
15:59 _ryanc_: you might see it if you create a fresh namespace
16:02 _ryanc_: So the baseline "#lang racket/base + racket/contract" is wrong, unless you want to play creepy namespace tricks.
16:03 _ryanc_: The real baseline is "#lang racket/base + racket/contract + (for-label typed/racket/base)".
16:04 samth: why is that?
16:04 _ryanc_: I get testtr.rkt: 1.07s, baseline.rkt: 0.74s, testrb.rkt: 0.16s.
16:05 samth: also, changing the implementation of #%mb to be dynamic-required saves loading about 50 modules
16:05 _ryanc_: Because you have to load the entire implementation of TR even if you don't instantiate it. (Again, unless you do tricks like dynamic-require.)
16:10 samth: so, for me i get TR: 0.54 Baseline: 0.36 RB: 0.08
16:10 samth: that's with large parts of the implementation dynamic-required
16:10 samth: unfortunately, other parts of the implementation are required
16:11 samth: since they are harder to change
16:11 (join) tcoppi_
16:14 (quit) tcoppi: Ping timeout: 276 seconds
16:20 _ryanc_: samth, I talked with Matthew about your dynamic-require hack and the possibility of another flavor of require that makes it safer and more official.
16:21 _ryanc_: The summary is that he's thought about it and hasn't thought of a combination that is both safe and useful (in cases such as yours).
16:21 samth: the hell with safe, i want fast! :)
16:21 _ryanc_: On the other hand, he sees the dynamic-require hack as reasonable, as long as you're aware of the sharp edges and the issues with eg creating executables for distribution.
16:22 _ryanc_: (At least, reasonable as filtered through my limited understanding of what you're doing :) )
16:22 samth: so, programs written in TR should still be turnable into executables
16:22 samth: just not TR itself, right?
16:23 samth: the trouble is that there are portions that are harder to turn into dynamic-require
16:23 _ryanc_: maybe... something like that...
16:23 _ryanc_: like what?
16:24 samth: typed-scheme/private/base-env.rkt
16:24 samth: or (somewhat more simply) typed-scheme/private/base-types.rkt
16:25 samth: which expands into roughly:
16:25 samth: (begin (define-syntax Type (\ (stx) error)) (dict-set! type-table #'Type (make-a-type ...)))
16:26 samth: and the catch is that TR needs to provide Type
16:26 samth: but `make-a-type' depends on some substantial portion of the implementation
16:28 _ryanc_: does it need to?
16:28 samth: does which need to?
16:28 _ryanc_: can you minimize the dependencies of 'make-a-type'?
16:29 samth: (make-a-type ..) is just a stand-in for an arbitrary expression
16:29 samth: it doesn't exist as a particular function
16:30 _ryanc_: then can you minimize the dependencies of those expressions? eg just have a module of structure definitions that they refer to?
16:31 samth: not really
16:32 (part) fjl
16:32 _ryanc_: or change the dictionary from id => Type to id => (type-factory -> Type)?
16:37 _ryanc_: This is the problem with final-algebraic representations, you know. (Or perhaps, non-intial-algebraic representations.)
16:38 (quit) carleastlund: Quit: carleastlund
16:51 samth: unfortunately, that doesn't even help
16:51 samth: if I want to maintain the structure of the file, the relevant identifiers need to be required into it
16:51 samth: so even if i delay the computation, the require has already happened
16:52 samth: i've tried adding delays - it didn't help
17:16 _ryanc_: samth, the point was to eliminate the require and get the machinery currently coming from the require from the type-factory
17:19 (quit) samth: Ping timeout: 272 seconds
17:25 (join) anRch
17:42 (nick) tcoppi_ -> tcoppi
17:50 (quit) _ryanc_: Quit: Leaving
17:51 (quit) spidermario: Read error: Connection reset by peer
18:19 (part) shofetim: "ERC Version 5.3 (IRC client for Emacs)"
18:21 (quit) anRch: Quit: anRch
18:53 (join) tym
18:54 (quit) tym: Client Quit
19:14 (join) rbarraud
19:26 (join) jeapostrophe
19:37 (quit) b-man_: Remote host closed the connection
19:43 (join) hanDerPeder
19:48 (quit) emma: Ping timeout: 240 seconds
20:03 (join) emma
20:30 (quit) jeapostrophe: Quit: jeapostrophe
20:34 (join) jeapostrophe
21:03 (quit) hanDerPeder: Quit: hanDerPeder
21:19 (quit) rbarraud: Ping timeout: 255 seconds
21:26 (quit) jeapostrophe: Quit: jeapostrophe
21:56 (join) jeapostrophe
21:59 (quit) Lajla: Read error: Connection reset by peer
22:00 (join) Lajla
22:11 (join) rbarraud
22:25 (join) asumu
22:43 (quit) rbarraud: Ping timeout: 245 seconds
22:43 (join) rbarraud
23:02 (join) Rakko
23:32 (quit) rbarraud: Remote host closed the connection