00:01 (quit) Shambles_: Ping timeout: 245 seconds
00:03 (quit) antithesis: Quit: yes leaving
00:03 (quit) replsos: Quit: ChatZilla 0.9.88.2 [Firefox 12.0/20120423130206]
00:16 neilv: if you think you found a bug in syntax-parse, unlikely as it seems, probably you did not
00:34 (quit) kvda: Quit: Computer has gone to sleep.
00:37 (join) kvda
01:00 (join) RacketCommitBot
01:00 RacketCommitBot: [racket] plt pushed 1 new commit to master: http://git.io/O6lZsw
01:00 RacketCommitBot: [racket/master] [Distributed Places] a few of Eli's documentation suggestions - Kevin Tew
01:00 (part) RacketCommitBot
01:19 eli: Shambles_1: if you really want to get errors for redefinitions of, say, bindings in a `#lang racket' file, then add (require racket) to your code.
01:20 eli: But like jamessan said, there is really no point in that.  It's perfectly fine to use any name you want -- that's following a very long scheme tradition of not having any "holy" names, and being able to redefine any name you want.
01:21 eli: yoklov: The threads that you get with `thread' are green threads; they still have a very valid use case -- for any kind of IO, they are extremely easy to deal with (especially with events etc).
01:22 eli: For OS threads you can use either futures which is a very lightweight but limited form of OS threads (share everything, but can be dangerous, and can't do IO etc), of places which are the opposite approach and are one step before starting a separate process.
01:23 eli: bremner: The reason that rudybot allows redefinitions is that it uses a toplevel-like namespace, where such redefinitions are allowed (with the usual equivalence to a `set!')
01:24 neilv: that said, if you enumerate *all* the holy names, then the universe ends
01:25 eli: That's platform^H^H^H^H^H^H^H^Hreligion-dependent...
01:26 (quit) Fare: Ping timeout: 248 seconds
01:27 neilv: just to be safe, i'm not going to be adding test cases for that
01:30 eli: neilv: http://en.wikipedia.org/wiki/Prayer_beads#Islam
01:33 neilv: http://en.wikipedia.org/wiki/Nine_Billion_Names_of_God
01:37 (quit) gerardpaapu: Ping timeout: 252 seconds
01:38 (quit) dnolen: Quit: ERC Version 5.3 (IRC client for Emacs)
01:42 (quit) ambrosebs: Ping timeout: 252 seconds
02:00 (join) ambrosebs
02:02 (quit) yoklov: Quit: bye!
02:08 (join) mye
02:19 (join) mye_
02:22 (join) hkBst
02:23 (quit) mye: Ping timeout: 260 seconds
02:24 (quit) mye_: Ping timeout: 245 seconds
02:32 (join) jesyspa
02:46 (quit) jonrafkind: Read error: Operation timed out
02:52 (quit) mithos28: Quit: mithos28
02:54 Shambles_1: eli: In what way are Futures dangerous?  Aren't they just single assignment variables, so they dodge the need for locking?
02:54 (quit) Shambles_1: Quit: Leaving.
02:54 (join) Shambles_
03:11 Shambles_: friscosam: Hey, that's pretty neat (the shadowing warnings thing).  Thanks.
03:18 (join) bluezenix
03:47 (quit) bluezenix: Quit: Leaving.
03:48 (join) bluezenix
03:49 (quit) kvda: Quit: Computer has gone to sleep.
03:51 (quit) ambrosebs: Ping timeout: 256 seconds
04:23 neilv: my new, lighter unit test syntax is making it just a little bit easier to enumerate all the test cases i need. http://i.imgur.com/J2FQG.png
04:24 (quit) jesyspa: Quit: Lost terminal
04:26 (join) noelw
04:26 (join) jesyspa
04:56 neilv: cheerios
04:56 (quit) neilv: Quit: Leaving
04:57 (quit) jesyspa: Quit: leaving
04:59 (join) jesyspa
05:04 (quit) jyc: Read error: Connection reset by peer
05:14 (quit) nilyaK1: Ping timeout: 276 seconds
05:28 (join) surrounder
05:28 (join) nilyaK
05:46 (join) masm
06:14 (join) netrino
06:19 (join) nilyaK1
06:21 (quit) nilyaK: Ping timeout: 252 seconds
06:29 (quit) bluezenix: Read error: Connection reset by peer
06:32 (join) bluezenix
06:32 (join) bitonic
06:37 eli: Shambles_: No, futures can lead to the obvious races.
07:14 (join) antithesis
07:15 (quit) nilyaK1: Quit: Leaving.
07:23 Shambles_: eli: I'm trying to understand where the problem is, since what 'futures' means in most implementations is safe, and that seems to be what Racket means by the term too.  The closest thing I can come up with is what the reference mentions about using set! ?  The problem not being with futures themselves but other things that are not thread safe.
07:24 eli: Shambles_: I don't know what you mean.  Specifically, I don't know about such meaning of safety.
07:24 eli: The bottom line is that if you run (set! x (+ x 1)) in
07:25 eli: 1. two threads -- you might get 2 as a result, but unlikely so because the system tends to swap threads on io or larger computations
07:25 eli: 2. two futures: you can definitely get 2 (starting from 1, btw)
07:26 (quit) bluezenix: Read error: Connection reset by peer
07:26 eli: 3. two places: that won't work since you can't mutate one place from another.
07:27 Shambles_: eli: My understanding of futures is they're implemented as single assignment variables, which avoids the need for locking, and any potential for lock contingency, but means computation forms a 'tree', where the root has to wait for the other computations to finish before it can read the final result, so it's not good for keeping a UI from 'hanging', like message passing concurrency (as in places).
07:27 (join) bluezenix
07:29 Shambles_: eli: The two threads don't have their own stack space or something?  I don't see why set would necessarily cause problems across threads.  It seems like it should only happen if it's done to a reference to shared state (a global variable or whatnot)
07:29 eli: That sentence doesn't make any sense to me...  Futures are parallel computations that are implemented usually via OS threads; variables are something completely different.
07:30 (join) plobzik
07:31 Shambles_: eli: Generally speaking anything 'held onto' by a thread's stack can be manipulated safely *so long as* it doesn't exist in another thread's stack.  If this was C and I passed a pointer to a global variable named x it would be unsafe.  But if x was inside the thread's own stack frames it should be safe.
07:31 Shambles_: eli: So I'm trying to figure out how the behavior of futures is 'surprising'.
07:32 eli: Sorry, I still don't know what you're talking about.
07:32 eli: "Stack" is an implementation detail.
07:32 eli: The main point is lexical scope.
07:32 (quit) antithesis: Quit: yes leaving
07:32 (join) antithesis
07:32 Shambles_: eli: Or perhaps you think of 'variables are something completely different' in a confusing way.  A future usually refers to a result that will start running when you create it, and only block when you try to read it.  A future is a kind of variable, or at least that's how it's usually described.
07:33 eli: "A future is a kind of a variable" --???
07:34 Shambles_: eli: Parallelism (and the opportunity for race conditions) is a implementation detail too.  Trying to understand the 'why' of "futures are bad, umkay?" involves me trying to figure out how futures behave differently than I expect, and futures /usually/ aren't 'bad'.  I assume from what you said Racket's futures aren't real futures somehow.
07:34 (quit) plobzik: Read error: Connection reset by peer
07:35 eli sighs
07:35 eli: It doesn't help to refer to things that are even less defined like "real futures".
07:35 eli: Also, I didn't say that they were bad in any way, just that the responsibility to keep out of the obvious problem is on you.
07:36 Shambles_: Theoretically a future represents the value produced by a computation.  You make them, you store them, you pass them around.  As soon as you try to read one, it will either return the value (if the computation is already finished), or 'hang' until the computation finishes.
07:37 eli: The last part is correct, but it explains why the first part is wrong -- futures don't represent values.
07:37 Shambles_: If you do this with single assignment variables, and you use a atomic (to the hardware) operation to update the "computation is finished, result should be safe to read now" flag, there should be no need for locking, and no surprising behavior.
07:37 (join) mye
07:37 eli: And single-assignment is a machine-code-generation thing, so I don't see how it's related here.
07:38 eli: Here's an example:
07:38 eli: (let ([x 0]) (define (inc!) (set! x (+ x 1))) (for-each touch (map future (list inc! inc!))) x)
07:38 Shambles_: eli: Forgive me if I try to understand programming in terms of what it does in the machine.  That's how I view programming.  It's a way to make the machine do useful or interesting things.  If understanding the implementation details bothers you, then just focus on explaining why futures should be more surprising than places.
07:39 eli: What should that return?
07:40 Shambles_: It looks like x exists outside the future, so I would believe it is shared, and thus not safe to mutate within the future.  That's not a shortcoming of futures.  The same thing should happen with message passing concurrency, or any other kind where you pass it a reference to global data.
07:40 Shambles_: It's akin to "pass a pointer to a global variable to the thread" in C, but of course that's probably considered implementation details.
07:41 eli: OK, the danger is that you *can* do that, and with places you cannot.
07:42 (join) mceier
07:42 Shambles_: How do places keep me from ever passing a reference to global data?  I thought it just provided message channels between threads?
07:43 eli: There is no shared memory since you initialize a place with its own module (and therefore namespace).
07:43 eli: And you can only pass restricted kinds of data across places.
07:43 Shambles_: You can still send data to...
07:43 eli: Specifically, no mutable data.
07:43 Shambles_: Alright, that sounds like the key.
07:43 Shambles_: Alright, that's what I needed to hear.  You can only pass constants then?
07:44 eli: Roughly, IIUC.
07:44 eli: That's what I meant about the difference between them and about places being almost like two processes.
07:45 Shambles_: That's interesting.  I take it it's made possible by a 'flag' set on the value in Racket that is passed around with the value?
07:45 (join) ahinki
07:47 Shambles_: What I'm getting at (yes, this is implementation details too, likely) is how does Racket 'know' whether it's a constant or not, and what happens if I try to pass something mutable?
07:50 eli: Some values are constants, and some are not, there's no flag other than the implicit type tag (if you want to talk about implementation).
07:51 eli: But note that I'm not getting into such implementation issues in the above -- for example, there is no relevance to whether `x' in the above is allocated on the stack or on the heap.
07:53 Shambles_: eli: I'm looking at the documentation and it explains what's allowed, and I have a guess as to how it's working now.  One thing I can't discern from the documentation is what happens with large amounts of data.
07:53 Shambles_: eli: Say I make a immutable byte string out of a image, and I'm working on a image processing program.  I aught to be able to pass this to a thread running as a place.  Is it going to actually create a copy of all that, or is it going to read the reference since it's immutable?
07:53 Shambles_: eli: Because it makes a great big difference in performance.
07:54 eli: You mean with places?  -- If so, then you'll need to ask on the list about such details...  I just know about the highlevel.
07:55 Shambles_: eli: Okay.  Maybe someone else on here will know later today.  I try to stay off mailing lists.
07:57 Shambles_: eli: Thanks for clearing up the statement about futures.  I aught to be able to use both safely since I would have expected the behavior you described with futures, but I can see why places would be considered safer now.
07:58 eli: I don't think that "safer" is the right term, "lightweight" seems more descriptive.
07:59 eli: As for asking, tewk is sometimes active here, but the mailing list is definitely a better place -- if you try to avoid mailing list load, you can subscribe to the list without mail delivery, and then you can see replies via a web interface (like gmane, google groups, or mail-archive.com).
08:06 Shambles_: eli: It's mostly the mailing list load, yes.  It also often takes a very long time to get a reply.  If we tried to have the conversation we just had on a mailing list it would probably have taken about a week due to all the back-and-forth.
08:07 Shambles_: As for safer vs lightweight.  I would only use the term lightweight to describe something more efficient and/or less featureful.  Safer would be what I would expect for something harder to shoot yourself in the foot with.
08:11 eli: Shambles_: tewk is the guy who implement{ed,s} places, and you're likely to get faster reply on the mailing list than trying to catch him.
08:12 Shambles_: eli: I see.  Thanks.
08:13 (join) snearch
08:45 (join) kanak
08:50 (quit) kanak: Client Quit
08:51 (join) zyoung
08:59 (quit) karswell: Remote host closed the connection
09:03 (quit) bluezenix: Quit: Leaving.
09:05 (join) bluezenix
09:09 (quit) bluezenix: Client Quit
09:09 (join) karswell
09:10 (join) bluezenix
09:21 (join) Fare
09:21 (quit) bluezenix: Quit: Leaving.
09:22 (join) bluezenix
09:37 (quit) lem_e_tweakit: Ping timeout: 244 seconds
09:38 (quit) getpwnam: Ping timeout: 256 seconds
09:38 (quit) hash_table: Ping timeout: 256 seconds
09:46 (join) ynniv
09:50 (join) Skola
09:54 (join) ynniv_
09:54 (quit) ynniv: Read error: Connection reset by peer
09:54 (nick) ynniv_ -> ynniv
10:18 (join) tyson1
10:24 (quit) Skola: Quit: leaving
10:24 (join) ynniv_
10:27 (join) Skola
10:30 (quit) ssbr: Remote host closed the connection
10:35 (join) ssbr
10:44 (join) getpwnam
10:45 (join) lem_e_tweakit
10:45 (join) hash_table
10:45 stamourv: danking: Great, I'll have a closer look!
10:45 stamourv: danking: As for `#%kernel', that's what's provided directly by the runtime.
10:46 stamourv: The very lowest-level parts of Racket (e.g. the implementation of `racket/base') are written in `#%kernel'.
10:47 stamourv: I don't recommend using it if you can avoid it.
10:47 (quit) ahinki: Quit: ChatZilla 0.9.88.2 [Firefox 13.0/20120509070325]
10:50 (quit) ynniv_: Quit: ynniv_
10:52 (part) tyson1
10:55 (join) tyson1
10:58 (quit) antithesis: Quit: yes leaving
11:05 (quit) Skola: Read error: Operation timed out
11:17 (part) tyson1
11:19 anonus: can racket use regexps in pattern-matching ?
11:21 samth: anonus: yes
11:21 anonus: can you gimme and example ?
11:21 samth: see the `regexp' form here: http://docs.racket-lang.org/reference/match.html?q=match#%28form._%28%28lib._racket/match..rkt%29._match%29%29
11:21 rudybot: http://tinyurl.com/cfkfojd
11:23 anonus: awesome, thanks
11:26 (join) ASau
11:28 (quit) hkBst: Quit: Konversation terminated!
11:51 (join) aalix
11:52 (quit) jrslepak: Quit: This computer has gone to sleep
11:53 (join) jonrafkind
11:53 (quit) bluezenix: Quit: Leaving.
11:55 (quit) aalix: Client Quit
11:57 (join) aalix
11:59 (quit) Shvillr: Quit: bye
11:59 (join) dnolen
11:59 (part) dnolen
11:59 (join) Shviller
12:03 friscosam: Shambles_: iirc this paper has more details on the implementation of places: http://www.eecs.northwestern.edu/~robby/pubs/papers/dls2010-tsffd.pdf
12:05 (join) kanak
12:12 Shambles_: friscosam: Thanks.
12:13 (join) nilyaK
12:17 (join) anRch
12:18 (join) jrslepak
12:19 (quit) jrslepak: Remote host closed the connection
12:28 (join) jacius
12:35 ChibaPet: A quick question... Is there anything in r6rs or r7rs dealing with starting and communicating with processes, or is that all implementation-specific?
12:36 ChibaPet: I'm searching but nothing is jumping out at me just yet.
12:48 (join) ASau`
12:49 (quit) dented42: Quit: Computer has gone to sleep.
12:52 (quit) ASau: Ping timeout: 248 seconds
13:01 (join) bluezenix
13:04 (join) sstrickl
13:04 (quit) sstrickl: Changing host
13:04 (join) sstrickl
13:16 (quit) bitonic: Quit: WeeChat 0.3.7
13:23 (quit) Nisstyre: Ping timeout: 246 seconds
13:28 (quit) nilyaK: Read error: Connection reset by peer
13:32 (quit) lem_e_tweakit: Ping timeout: 246 seconds
13:33 (quit) hash_table: Ping timeout: 246 seconds
13:33 (quit) getpwnam: Ping timeout: 246 seconds
14:09 asumu: ChibaPet: No, I really doubt there is. Why?
14:09 asumu: (that is, why do you need to use R6 or R7?)
14:10 ChibaPet: Oh, the notion of doing some conversions from Perl to Lisp came up, and I'm trying to see how different Lisps do things I want to do. In particular, Perl makes it trivial to say, for instance, open $query, "$wmctrl -l -G|", and then pull program output out of the $query filehandle.
14:11 ChibaPet: And so, I was wondering if there's anything in any Lisp-related standards that talks about that sort of thing, but it seems like everything (CL and Scheme) is all very implementation-specific.
14:12 asumu: I see. Well, in that regard Perl has it much easier since it's just one implementation essentially.
14:12 asumu: As opposed to Scheme where you may as well consider them all separate languages.
14:12 asumu: Racket has a pretty good process API.
14:13 asumu: http://pre.racket-lang.org/docs/html/reference/subprocess.html
14:13 ChibaPet: Yeah, that looks pretty flexible.
14:14 ChibaPet: Hey, does Racket support asychronous socket I/O out of the box?
14:16 ChibaPet: As in, what I'd get from select in C/Perl?
14:16 Shambles_: Is there a way to define contracts for things you don't want exported within the module, to help reduce debugging effort?
14:19 (quit) anRch: Quit: anRch
14:19 (join) antithesis
14:19 asumu: ChibaPet: As in Unix sockets? I don't think there is, because Racket's API is all cross-platform. You can do async io with things like threads and async-channels though.
14:19 (quit) antithesis: Remote host closed the connection
14:19 ChibaPet: I'll have to look at async-channels and see what that's about. I despise threading.
14:20 asumu: rudybot: (define/contract integer? "foo")
14:20 rudybot: asumu: your sandbox is ready
14:20 rudybot: asumu: error: #<undefined>:1:0: define/contract: bad syntax in: (define/contract integer? "foo")
14:20 asumu: rudybot: (define/contract x integer? "foo")
14:20 rudybot: asumu: error: x: self-contract violation, expected: integer?, given: "foo" <NEWLINE>   contract from: (definition x), blaming: (definition x) <NEWLINE>   contract: integer? <NEWLINE>         at: #<undefined>:1.17
14:20 asumu: Shambles_: ^
14:20 chandler: rudybot: doc sync
14:20 rudybot: chandler: your sandbox is ready
14:20 rudybot: chandler: http://docs.racket-lang.org/reference/sync.html#(def._((quote._~23~25kernel)._sync))
14:20 chandler: ChibaPet: Have a look at that.
14:21 chandler: `sync' is the closest Racket equivalent to select().
14:22 ChibaPet: Hm, was looking at http://pre.racket-lang.org/docs/html/reference/async-channel.html but it seems suspiciously like it might be threading under the hood if it's not using select - which it seems not to, since it describes async-channel-get as blocking.
14:23 bremner: if only the source were available ;)
14:23 ChibaPet: sync looks similar, except that it's explicitly blocking
14:23 ChibaPet: I know, right?
14:24 Shambles_: asumu: Thanks.  I see that in the documentation now.  I'm guessing it works fine for most purposes?  I could use it for anything defined?
14:28 (join) lem_e_tweakit
14:28 (join) getpwnam
14:29 (join) hash_table
14:32 Shambles_: asumu: I wonder how that would look for attaching checks to a members/methods on a internal class.  The documentation is sort of confusing.  It seems very oriented toward only doing things at module boundaries.  There's a mention of using define-struct/contract.  Is that what I'd need?
14:35 asumu: Shambles_: class as in using racket/class?
14:36 asumu: rudybot: (define point% (class/c (field [x integer?])) (class object% (super-new) (init-field [x 0])))
14:36 rudybot: asumu: error: #<undefined>:1:0: define: bad syntax (multiple expressions after identifier) in: (define point% (class/c (field (x integer?))) (class object% (super-new) (init-field (x 0))))
14:37 asumu: rudybot: (define point% (class/c (field [x integer?])) (class object% (super-new) (init-field [x 0]))))
14:37 rudybot: asumu: [racket/master] Add #:opaque keyword to class/c. - Asumu Takikawa
14:37 asumu: rudybot: (define/contract point% (class/c (field [x integer?])) (class object% (super-new) (init-field [x 0])))
14:37 rudybot: asumu: Done.
14:38 asumu: rudybot: (new point% [x "wrong"])
14:38 rudybot: asumu: ; Value: #(struct:object:point% "wrong")
14:38 asumu: Uhh.
14:38 asumu: Oh.
14:38 asumu: rudybot: (define/contract point% (class/c (init-field [x integer?])) (class object% (super-new) (init-field [x 0])))
14:38 rudybot: asumu: Done.
14:38 asumu: rudybot: (new point% [x "wrong"])
14:38 rudybot: asumu: error: point%: contract violation, expected: integer?, given: "wrong" <NEWLINE>   contract from: (definition point%), blaming: program <NEWLINE>   contract:  <NEWLINE>     (class/c <NEWLINE>      (init (x integer?)) <NEWLINE>      (field (x integer?))) <NEWLINE>   at: #<undefined>:1.17
14:38 asumu: Shambles_: ^
14:38 Shambles_: asumu: yes, as in using racket class.
14:39 Shambles_: asumu: Thanks for all the help.
14:46 (quit) karswell: Remote host closed the connection
14:49 (quit) mye: Read error: Connection reset by peer
14:49 (join) mye
14:55 (quit) noam: Ping timeout: 252 seconds
14:55 (join) nilyaK
14:56 (join) karswell
15:02 (join) bitonic
15:03 (join) noam
15:05 (quit) lem_e_tweakit: Ping timeout: 244 seconds
15:05 (quit) getpwnam: Ping timeout: 248 seconds
15:06 (quit) hash_table: Ping timeout: 248 seconds
15:09 (join) wand
15:09 (quit) wand: Client Quit
15:13 (join) wand
15:21 (quit) bluezenix: Quit: Leaving.
15:24 (join) getpwnam
15:25 (join) hash_table
15:30 (join) jonathansizz
15:34 (join) bluezenix
15:46 (quit) acarrico: Ping timeout: 248 seconds
15:50 (join) jao
15:50 (quit) jao: Changing host
15:50 (join) jao
15:56 (quit) nilyaK: Quit: Leaving.
15:57 (join) RacketCommitBot
15:57 RacketCommitBot: [racket] plt pushed 1 new commit to master: http://git.io/JCLXVg
15:57 RacketCommitBot: [racket/master] Add \bot as alias for \perp in DrRacket (both are supported in LaTeX). - Sam Tobin-Hochstadt
15:57 (part) RacketCommitBot
16:08 (join) stamourv`
16:09 (quit) stamourv`: Client Quit
16:18 (join) anRch
16:28 (join) yoklov
16:34 (join) RacketCommitBot
16:34 RacketCommitBot: [racket] plt pushed 1 new commit to master: http://git.io/TEcKVA
16:34 RacketCommitBot: [racket/master] Use coerce-contract when building interfaces with contracts. - Stevie Strickland
16:34 (part) RacketCommitBot
16:34 (join) Nisstyre
16:55 (quit) anRch: Quit: anRch
16:57 (join) anRch
17:26 (quit) anRch: Quit: anRch
17:28 (join) gerardpaapu
17:32 (quit) gerardpaapu: Client Quit
17:42 (quit) sstrickl: Quit: sstrickl
17:43 (join) jrslepak
17:43 (quit) bluezenix: Quit: Leaving.
17:44 (join) bluezenix
17:45 (quit) bluezenix: Client Quit
17:49 (quit) wand: Remote host closed the connection
17:50 (join) bluezenix
17:52 (quit) kanak: Ping timeout: 240 seconds
17:53 (quit) bluezenix: Client Quit
17:57 (quit) mye: Ping timeout: 240 seconds
17:57 (join) jeapostrophe
17:57 (quit) jeapostrophe: Changing host
17:57 (join) jeapostrophe
18:06 (quit) netrino: Quit: Ave!
18:06 (quit) zyoung: Ping timeout: 248 seconds
18:09 danking: stamourv: Yeah, then I'm not sure how to create a more direct integration of multiple auto-values. The implementation of auto-values seems to depend on `make-struct-type' which, AFAICT, is `require'd from #%kernel.
18:11 danking: So, we need to hack #%kernel so that the fifth argument to make-struct-type expects a list of [Maybe Any] that corresponds to the auto-values for the fields.
18:14 (quit) jeapostrophe: Ping timeout: 256 seconds
18:15 (quit) snearch: Quit: Verlassend
18:18 (join) jyc
18:26 Shambles_: I had to go away a while.  I take it racket/async-channel is what is used to do async i/o in racket?  Someone was asking about TCP/IP earlier.
18:34 (join) jeapostrophe
18:34 (quit) jeapostrophe: Changing host
18:34 (join) jeapostrophe
18:39 friscosam: async-channel is for messaging between racket threads asynchronously
18:40 friscosam: async i/o is built on top of however the OS does async i/o
18:40 friscosam: so select or poll or whatnot
18:40 friscosam: iirc
18:42 Shambles_: friscosam: I think the conclusion was that there was no 'select', but 'sync' could be used to do something similar.  I was trying to figure out where Racket hid away the 'good' file and networking i/o.  My best guess was there.
18:45 friscosam: I'm not sure I get what you mean by 'good'
18:46 Shambles_: friscosam: By good I mean asynchronous. ;)  It's nice to be able to say you want a file 'eventually', carry on with your computations, and when you run out of stuff to do, then try to read the result and if it isn't finished it blocks until it is.
18:48 friscosam: yeah you could do that with sync and channels
18:52 Shambles_: friscosam: Is there any high level documentation on this.  I haven't finished the guide yet but got curious after the topic came up.
18:55 samth_away: Shambles_: you should use threads
18:55 Shambles_: samth_away: Any particular reason for that?  Also, which API? 'threads' threads?
18:56 samth_away: rudybot: doc thread
18:56 rudybot: samth_away: your sandbox is ready
18:56 rudybot: samth_away: http://docs.racket-lang.org/reference/threads.html#(def._((quote._~23~25kernel)._thread))
18:56 samth_away: that threads
18:56 samth_away: that's how you do asynchronous processing in racket
18:57 Shambles_: samth_away: So basically just use regular blocking i/o in a separate green thread?
19:06 (join) dyoo
19:07 dyoo: Shambles_: my impression was that C's select and Racket's sync or sync/timeout had equivalent behavior; where do they differ?
19:08 Shambles_: dyoo: The discussion earlier seemed to indicate they were equivalent.  Now I'm rather confused about what one should do to read/write files and sockets asynchronously though, since there seems to be differing opinions.
19:09 Shambles_: (2:23:05 PM) ChibaPet: sync looks similar, except that it's explicitly blocking
19:11 dyoo: Shambles_: hmm... but doesn't select C's block as well, unless we provide it a non-NULL timeout argument?  Racket's sync/timeout will also not block if given a timeout of 0.
19:13 Shambles_: dyoo: I think they only found 'sync' not 'sync/timeout', so that may be the reason for the comment.  It just got me thinking about how to do async I/O for working with files and the network.
19:13 dyoo: Shambles_: yup.  Let me check something.. I'm reading through http://docs.racket-lang.org/reference/port-lib.html
19:14 dyoo: which might have some content about treating ports as events
19:14 (quit) jacius: Remote host closed the connection
19:14 dyoo: ... yeah, you might find the content starting around http://docs.racket-lang.org/reference/port-lib.html#(def._((lib._racket/port..rkt)._read-bytes-evt)) relevant
19:17 dyoo: for the writing side of things: http://docs.racket-lang.org/reference/Byte_and_String_Output.html?q=write%20evt#(def._((quote._~23~25kernel)._write-bytes-avail-evt))
19:17 rudybot: http://tinyurl.com/72978g5
19:17 dyoo: ah, thank you rudybot
19:20 (part) dyoo
19:20 Shambles_: dyoo: So synchronization events and sync are the usual way to do async i/o in Racket.
19:21 (quit) yoklov: Quit: computer sleeping
19:23 (join) yoklov
19:23 (join) dyoo
19:23 dyoo: shambles_: yup!  That's why I was getting confused from reading the IRC logs.  :)
19:24 Shambles_: dyoo: Okay, thanks for clearing that up. :)
19:26 (quit) mceier: Quit: leaving
19:27 (part) dyoo
19:36 (join) mye
19:36 (quit) jonrafkind: Ping timeout: 240 seconds
19:41 (join) nilyaK
19:51 (quit) jeapostrophe: Ping timeout: 265 seconds
19:57 (quit) bitonic: Quit: WeeChat 0.3.7
20:05 (join) mye_
20:05 (join) kvda
20:07 (join) dnolen
20:08 (quit) mye: Ping timeout: 250 seconds
20:08 (join) jacius
20:23 (quit) aalix: Ping timeout: 260 seconds
20:39 (quit) masm: Quit: Leaving.
20:57 (quit) dnolen: Ping timeout: 276 seconds
21:25 (quit) Fare: Ping timeout: 245 seconds
21:27 (join) zyoung
21:38 (join) Fare
21:43 ozzloy: i just installed neil's sicp package which gave some errors during the install.  i'd like to reinstall so i can file a bug report.  how do i completely uninstall it?
21:51 asumu: ozzloy: which OS?
21:51 (join) mye__
21:52 asumu: (on Unix it's in ~/.racket, on Windows "Racket" under "Application Data", and... not sure on Mac)
21:52 (join) jonrafkind
21:55 (quit) mye_: Ping timeout: 246 seconds
22:19 ozzloy: asumu, ubuntu
22:20 ozzloy: asumu, you mean i can just rm -rf ~/.racket/5.2.1/ ?
22:23 (quit) jesyspa: Quit: leaving
22:29 Fare: looking for speakers in Boston in June, July, etc.
22:31 offby1: You can probably get them at any decent stereo place
22:31 offby1 flees
22:33 eli: Fare: How about a more realistic two month schedule?
22:35 Fare: can work, too
22:35 Fare: but I don't have a speaker for july, either.
22:36 eli: I can't think of a concrete potential speaker though.
22:36 eli: My try to get some of the old-timers to talk?
22:37 eli: s/My/Maybe/
22:40 (quit) kvda: Quit: Computer has gone to sleep.
22:44 (join) mithos28
22:51 Fare: yup - that's my current plan for june
22:54 (quit) jonathansizz: Ping timeout: 240 seconds
22:54 (quit) getpwnam: Ping timeout: 260 seconds
22:54 (quit) hash_table: Ping timeout: 244 seconds
22:56 asumu: ozzloy: You don't want to do that, just delete the planet package directory.
22:59 asumu: (and more specifically just the sicp package)
23:07 (quit) yoklov: Remote host closed the connection
23:07 (join) yoklov
23:07 (quit) yoklov: Remote host closed the connection
23:14 (quit) zyoung: Remote host closed the connection
23:17 (join) zyoung
23:34 (quit) zyoung: Remote host closed the connection
23:39 (join) kvda
23:43 (quit) flying_rhino: Quit: KVIrc 4.0.4 Insomnia http://www.kvirc.net/