00:00 jeapostrophe: (define-syntax-rule (enum sym ...) (begin (define sym 'sym) ...)))
00:00 jeapostrophe: is another simpler definition that is close to what you want
00:00 lewis1711: oh so that's what the begin form is
00:01 jeapostrophe: macros can only expand into one thing
00:01 jeapostrophe: but you want a sequence of defines
00:01 offby1: men only want one thing
00:01 jeapostrophe: so you can put them in a begin
00:01 jeapostrophe: (also, the top-level will splice in begins it finds... they don't create a new scope)
00:06 lewis1711: would integers be quicker to check than symbols? because otherwise that snippet you posted before is just about right
00:06 (join) vkz
00:06 jeapostrophe: no, they wouldn't be faster
00:06 jeapostrophe: symbol comparisons are integer comparisons
00:07 jeapostrophe: because symbols are pointers... they are not compared textually
00:07 lewis1711: ah
00:11 lewis1711: jeapostrophe: I will have a look at that book when I next try and learn macros
00:11 jeapostrophe: word
00:12 (quit) nilyaK: Quit: Leaving.
00:13 (quit) Kaylin: Read error: Connection reset by peer
00:13 (quit) vkz: Quit: vkz
00:14 (join) zussitarze
00:20 (quit) zussitarze: Quit: Page closed
00:31 (join) veer
00:42 (join) vkz
00:51 (quit) jrslepak: Quit: This computer has gone to sleep
00:52 neilv: hm. the racket reference says that "let*" ``evaluates the val-exprs one by one'', but unclear to me whether it *guarantees* left-to-right order of evaluation
00:53 jonrafkind: you could look at the expansion of it
00:55 neilv: i have a good idea what it currently does. i mean to ask what's specified
01:01 (quit) vkz: Quit: vkz
01:09 asumu: neilv: R6RS requires left-to-right evaluation http://www.r6rs.org/final/html/r6rs/r6rs-Z-H-14.html#node_idx_398
01:09 asumu: Racket isn't R6RS, but maybe it follows it.
01:11 neilv: aha, the racket reference specifies "let" as left-to-right, so "let*" specification inherits that
01:12 neilv: which is unfortunate, because every time i write a "let" of more than one val-expr, i fantasize about them being parallelized
01:13 neilv: well, my evening is ruined
01:15 (join) nilyaK
01:49 neilv: i wonder whether the contract overhead for the web-server response struct constructor is significant
01:52 (join) djcb`
01:52 jeapostrophe: neilv: i benchmarked it once
01:53 jeapostrophe: and so i optimized it
01:53 jeapostrophe: it is reasonable now
01:53 (quit) djcb: Ping timeout: 246 seconds
01:54 neilv: thanks, good to know
01:54 jeapostrophe: i'd still like a global contracts-off mode
01:54 neilv: absolutely
01:55 jeapostrophe: blake's ms work is designed to make that more possible
01:55 jeapostrophe: but his stuff is just the first step
02:05 (quit) jeapostrophe: Ping timeout: 252 seconds
02:07 (join) realitygrill
02:13 (quit) Diarmid: Quit: Leaving.
02:20 (quit) yoklov: Quit: computer sleeping
02:24 (join) Diarmid
02:27 (join) jao
02:28 (quit) djcb`: Remote host closed the connection
02:39 (join) cdidd
02:43 (quit) jonrafkind: Ping timeout: 272 seconds
02:48 (quit) veer: Quit: Leaving
02:56 (join) RacketCommitBot
02:56 RacketCommitBot: [racket] plt pushed 2 new commits to master: http://git.io/xW_sNA
02:56 RacketCommitBot: [racket/master] Remove no-longer used net/cgi exceptions. - Eli Barzilay
02:56 RacketCommitBot: [racket/master] Add to/from bytes functions to the json library. - Eli Barzilay
02:56 (part) RacketCommitBot
02:56 (join) hkBst
02:56 (quit) hkBst: Changing host
02:56 (join) hkBst
03:41 (join) nilyaK1
03:43 (quit) nilyaK: Ping timeout: 252 seconds
03:46 (join) RacketCommitBot
03:46 RacketCommitBot: [racket] plt pushed 1 new commit to master: http://git.io/bMRMHA
03:46 RacketCommitBot: [racket/master] Remove typed versions of the cgi exceptions. - Eli Barzilay
03:46 (part) RacketCommitBot
03:48 mrcarrot: how is it with racket, has it any way to automatically split the problem into threads? like in haskell where you pass the -threaded option to the compiler
03:51 neilv: racket has ways of using multiple CPUs, but i don't know of any current automatic parallelization. you could ask on the email list
03:59 (nick) nilyaK1 -> Kaylin
04:06 (quit) _AlbireoX: Ping timeout: 245 seconds
04:06 (join) AlbireoX
04:25 (join) kvda
04:34 (quit) kvda: Quit:  x___x
04:41 (quit) jao: Ping timeout: 260 seconds
04:43 neilv: bummer. drracket seems to have hanged, when i was doing an identifier rename with the on-the-fly check-syntax running
04:55 (join) itasyno
05:05 (join) noelw_away
05:05 Kaylin: legit question, when one says a program hangs, where does this term come from?
05:07 neilv: i don't know the etymology
05:09 neilv: might be similar to "leave you hanging"
05:10 Kaylin: like when we say something is hung up, snagged maybe
05:10 Kaylin: idk
05:11 (join) bas_
05:11 (nick) bas_ -> Skola
05:12 (join) untrusted
05:28 tim-brown: you can hang up a phone... that's a fairly physical description of
05:28 tim-brown: what happens when you put the receiver down.
05:28 tim-brown: (and the phone is mounted vertically -- or a candle-stick phone)
05:29 Kaylin: the call terminates gracefully =X
05:29 tim-brown: from there e.g. "I was talking to Jim and he hung up on me"
05:29 Kaylin: oh, depends which end you're on
05:29 Kaylin: I suppose
05:29 tim-brown: nah...  nothing graceful about Jim... something I said annoyed him
05:29 tim-brown: and he hung up on me
05:30 (join) mceier
05:30 Kaylin: :(
05:31 tim-brown: that's how I see it, personally. the program is at the other side of a conversation on a terminal
05:31 tim-brown: and if i upset it, it hangs up on me
05:32 tim-brown: (although I *do* deal with a lot of real devices on the end of real phone lines, so I'm actually being quite literal sometimes)
05:33 tim-brown: it's a nice word, though... it has the "dangling" aspect to it too
05:33 neilv: if your program hangs up on you, it might just be smoking POTS?
05:37 (part) untrusted: "ERC Version 5.3 (IRC client for Emacs)"
05:38 neilv: tough crowd for telephony jokes
05:38 Kaylin: is that stop in reverse?
05:38 neilv: "plain old telephone service"
05:38 Kaylin: :P
05:39 Kaylin: so not quite as bad as SMTP
05:39 neilv: http://en.wikipedia.org/wiki/Plain_old_telephone_service
05:42 (quit) dous: Ping timeout: 246 seconds
05:45 (join) chezduck
05:47 neilv: i hope nobody ever needs to use html-template with any charset other than utf-8, because i've broken all that
05:49 (join) gciolli
06:07 (join) masm
06:18 noelw_away: utf-8 ftw
06:18 (nick) noelw_away -> noelw
06:21 noelw: Educational software makes me sad.
06:21 noelw: Has anyone ever used an online learning environment they like?
06:21 neilv: if they didn't want you to be sad, they would call it recreational, rather than educational
06:23 noelw: Recreation is supposed to be fun
06:24 (part) noelw
06:24 (join) noelw
06:26 (quit) Kaylin: Quit: Leaving.
06:32 (quit) rostayob: Ping timeout: 260 seconds
06:45 (join) rostayob
07:29 neilv: hm. drracket 5.2.1 got in an infinite loop on compiling, without hanging hard, no matter what i changed. i exited drracket and restarted, and it promptly gave me a correct compiler error
07:38 (quit) rostayob: Ping timeout: 246 seconds
07:39 lewis1711: is there anyway to do a (sequenceof number?) for contracts? typed racket has Sequenceof so I imagine there's *somethin*
07:42 (join) samth_
07:53 lewis1711: rudybot: (define (sequenceof type?) (lambda (seq) (sequence-andmap type? seq)))
07:53 rudybot: lewis1711: Done.
07:53 lewis1711 wonders if that will do the right thing
07:56 lewis1711: "contract violation, expected: ...ewis/std1711.rkt:4:27"  I remember reading about this problem somewhere
08:05 lewis1711: flatnamed contracts, that was it
08:10 (join) jrslepak
08:21 (quit) jrslepak: Quit: This computer has gone to sleep
08:21 (quit) bmp: Quit: Leaving...
08:23 neilv: if i have a utility procedure that's called from within syntax-case clauses, and i want this utility procedure to construct syntax with bindings from lexical context (or just toplevel) in which the utility procedure is defined... how do i do that?
08:23 neilv: it's too late at night to be reading reference/stxtrans.html
08:27 (join) RPR
08:28 RPR: samth_: Thx again for the assist on the typing.  Typed Racket is just the way to go.
08:29 samth_: RPR, no problem
08:29 samth_: happy to help
08:30 samth_: neilv, probably something like this: (datum->syntax #'here '(foo bar))
08:30 RPR: samth_: FWIW the 2 nits that I hit the most (bugwise) in TR 1) struct-copy not working. 2) providing struct: without itemizing every ctor,predicate and accessor etc.
08:31 samth_: RPR, ironically, i think those are the same bug
08:32 samth_: RPR, this works for me: (struct-copy Success (Success 3) [result 4])
08:33 RPR: nice.  I haven't tried it recently.
08:33 RPR: One down...
08:33 neilv: samth_: that's what i first tried, but #'here doesn't seem to come with lexical info in that context
08:34 samth_: neilv, do you want the *local* bindings from where you're creating the syntax?
08:34 samth_: RPR, this works too, in that file: (provide (struct-out Success))
08:35 neilv: samth_: even just racket/base would be nice to start with
08:36 samth_: #'here should work fine, but your problem may be that you need `(require (for-template racket/base))'
08:37 neilv: aha. i was using for-label
08:37 neilv: thanks
08:46 RPR: samth_: struct-out is working. :)  However struct-copy still only works for me in the defining module. :(
08:47 samth_: neilv, `for-label' is only for docs
08:47 samth_: RPR, even if both modules are typed?
08:47 RPR: yes
08:48 RPR: struct-copy: accessor name not associated with the given structure type at: Cell-x in: (struct-copy Cell (Cell "ray") (x "eve"))
08:48 RPR: Works fine in the module where struct: Cell is defined.
08:49 (quit) RPR: Quit: job thing ...
09:00 (join) bmp
09:10 (join) jeapostrophe
09:19 (join) rostayob
09:24 (quit) rostayob: Ping timeout: 246 seconds
09:37 (quit) kanak: Quit: Leaving.
09:41 (join) dnolen
09:47 (join) yoklov
09:52 (join) jrslepak
09:57 (quit) samth_: Ping timeout: 246 seconds
10:04 (join) rostayob
10:06 (join) dous
10:08 (quit) yoklov: Quit: computer sleeping
10:10 (quit) jrslepak: Quit: This computer has gone to sleep
10:16 (join) mye_
10:17 (join) gridaphobe
10:29 (join) kvda
10:31 (join) RacketCommitBot
10:31 RacketCommitBot: [racket] plt pushed 1 new commit to master: http://git.io/4j5K8A
10:31 RacketCommitBot: [racket/master] Fix sync/timeout hang due to place channels - Kevin Tew
10:31 (part) RacketCommitBot
10:32 (quit) kvda: Client Quit
10:33 (join) kvda
10:41 (join) RacketCommitBot
10:41 RacketCommitBot: [racket] plt pushed 1 new commit to master: http://git.io/uFylZQ
10:41 RacketCommitBot: [racket/master] Fix sync/timeout hang due to place channels - forgot to git add - Kevin Tew
10:41 (part) RacketCommitBot
10:42 neilv: i think this is about as fast as i can make a static html-template work with web-server. http://paste.lisp.org/display/128380
10:43 noelw: Looks good
10:43 noelw: Back on to GMT, I see. ;-0
10:45 neilv: for html-templates that aren't completely static, there will be two at least two different forms, basically distinguished by whether you want the content-length header determined for you, or you want to be able to do things like flush the output in the middle or send a huge file
10:46 chandler: does the latter form use chunked encoding?
10:48 neilv: all response/html-template will care about is whether it builds a list of bytes and then sends them, or it does a series of writes for each static and dynamic step as it goes through the dynamic ones in order
10:49 neilv: chunked encoding is up to web-server
10:49 chandler: ah
10:49 neilv: noelw: back on gmt as of this morning, looks like
10:50 (quit) kvda: Quit:  x___x
10:52 (join) yoklov
10:56 (quit) dnolen: Ping timeout: 260 seconds
11:01 tim-brown: tewk: thanks
11:02 tim-brown: is there a dialy build of windows (64) racket installer?
11:03 tim-brown: when is it built? (i want the places fix :-)
11:04 tim-brown: s'ok, I see the nightly installer page
11:05 (join) jrslepak
11:13 tim-brown: k
11:17 (nick) samth_away -> samth
11:18 samth: tim-brown: they're build around 4AM eastern time in the US
11:19 samth: mrcarrot: the -threaded option to GHC does not automatically parallelize your program
11:19 samth: mrcarrot: unless you've explicitly disabled places and futures at compile time (of the Racket runtime), Racket is always like GHC -threaded
11:19 (quit) cdidd: Remote host closed the connection
11:24 (join) kanak
11:28 neilv: my bad. i thought he was talking about a different feature of the haskell that i played with
11:29 (quit) hkBst: Quit: Konversation terminated!
11:29 (quit) jrslepak: Quit: This computer has gone to sleep
11:30 (join) asdfhjkl
11:36 tim-brown: samth: so about 9am GMT/UTC, I guess... looking forward to it!
11:37 (quit) gciolli: Ping timeout: 240 seconds
11:41 (quit) jeapostrophe: Ping timeout: 255 seconds
12:05 (join) nilyaK
12:18 (quit) Shviller: Ping timeout: 276 seconds
12:18 (join) Shviller
12:21 (quit) Skola: Ping timeout: 264 seconds
12:22 (join) vkz
12:25 mrcarrot: samth: really? take a look at http://shootout.alioth.debian.org/u64q/benchmark.php?test=all&lang=racket&lang2=ghc
12:25 rudybot: http://tinyurl.com/7heyv39
12:25 (nick) noelw -> noelw_away
12:25 mrcarrot: samth: take a look at the haskell source and you will see that there is no code for making concurency, and still it is using all cores
12:25 (quit) noelw_away: Quit: noelw_away
12:26 mrcarrot: samth: the only times racket is using several cores is when they do it in source
12:28 mrcarrot: haskell looks like it is all the time using some threading...
12:29 nilyaK has some sheets that have many threads.
12:29 bremner: well, for GC anyway
12:33 mrcarrot: what is GC?
12:33 bremner: garbage collection
12:33 (join) nilyaK1
12:33 bremner: since ghc, 6.x there is parallel gc
12:34 mrcarrot: ah, garbage collector
12:35 (quit) nilyaK: Ping timeout: 246 seconds
12:36 (nick) nilyaK1 -> Kaylin
12:37 (join) jonrafkind
12:37 (quit) Kaylin: Client Quit
12:38 mrcarrot: out of curiousity, is that something on the way to come to racket?
12:39 mrcarrot: the performance of racket seems to otherwise be really okay. not fully as good as haskell, but very much good enough
12:39 mrcarrot: (if microbenchmarks are to be trusted at all)
12:41 (join) zussitarze
12:44 samth: mrcarrot: that's not correct about haskell
12:45 samth: mrcarrot: if you look at http://shootout.alioth.debian.org/u64q/program.php?test=spectralnorm&lang=ghc&id=4 you can see the calls to `forkIO`
12:45 rudybot: http://tinyurl.com/7toeo5b
12:48 (join) MayDaniel
12:53 bremner: oh, C style parallelism ;)
12:54 tim-brown: mrcarrot and samth: where did this conversation about parallelism come from?
12:54 tim-brown: i can really follow it back to the start in my irc log
12:54 tim-brown: can't
13:02 mrcarrot: tim-brown: yes, some are doing that... but take a look at some of the other examples there and you can see that it is not always the case
13:02 mrcarrot: still the little gain it has over racket comes from threading
13:03 tim-brown: mrcarrot: ?
13:03 mrcarrot: fasta, for example
13:03 mrcarrot: http://shootout.alioth.debian.org/u64q/benchmark.php?test=all&lang=racket&lang2=ghc
13:03 rudybot: http://tinyurl.com/7y4fk7s
13:04 mrcarrot: tim-brown: fasta source is here: http://shootout.alioth.debian.org/u64q/program.php?test=fasta&lang=ghc&id=1
13:04 rudybot: http://tinyurl.com/7lj7pkf
13:05 mrcarrot: tim-brown: but you are right that they use many times similarities to futures in haskell
13:05 (quit) realitygrill: Quit: realitygrill
13:06 tim-brown: mrcarrot: pretty sure you're not talking to *me*
13:06 mrcarrot: in the fasta example, they pass the -threaded argument and get the load spread as 18% 69% 17% 2%
13:06 mrcarrot: ah, it was samth i mean all that too... sorry
13:07 tim-brown: np, just confused me... is all
13:08 mrcarrot: tim-brown: this discussion came because i asked if somebody knows if racket has something similar like -threaded in ghc to automatically create some concurency without doing anything in the source
13:08 mrcarrot: tim-brown: i just recently found out about racket and i am curious how it compares to haskell... that is why i am asking a lot of strange questions
13:09 tim-brown: ah... i'm playing with places myself, atm
13:09 bremner: last time I looked at parallelism in haskell, you still had to annotate with "par" to get parallelism, so it wasn't really automatic.
13:09 tim-brown: and anyway, racket raises a lot of strange questions!
13:10 tim-brown: from what i've seen, places seem quite heavy, for small units of work, and futures seem quite limiting in what work they'll handle
13:11 mrcarrot: bremner: not in that "fasta" code... it is doing some parallelism there automatically and winning racket (with allthough a small marginal)
13:11 tim-brown: although (once i've got a fix downloaded) i'm using a place farm for parallelism
13:11 tim-brown: which seems to do the trick as long as there's not too much inter-place communication
13:11 bremner: mrcarrot: ok, then I guess there was some progress in the last few years. I remember there were projects along those lines, but not integrated into ghc
13:12 mrcarrot: bremner: probably it got integrated since then
13:13 (quit) zussitarze: Quit: Page closed
13:14 mrcarrot: all i can conclude from my questions so far is that both racket and haskell are good languages
13:17 (join) jeapostrophe
13:18 mrcarrot: bremner: how much did you use haskell?
13:19 bremner: not much. I only wrote one non-toy program.
13:19 mrcarrot: ah, okay
13:20 mrcarrot: had you been using racket before you begun with haskell?
13:20 bremner: no
13:20 mrcarrot: okay, so what did get you to switch over to racket?
13:20 (join) gciolli
13:21 bremner: mrcarrot: I'm teaching and I find racket better for that.
13:21 mrcarrot: hmm, that was what i concluded too.. that if i am teaching programming, racket would be better for keeping the interest up
13:22 mrcarrot: at least from the material i have quickly been looking at
13:22 rostayob: bremner: "par" is a very restricted view of parallelism in Haskell
13:22 bremner: ok
13:22 (part) tim-brown: "Leaving"
13:23 bremner: all I know is what is in Real World Haskell ;)
13:23 bremner: oh, I guess I forgot STM
13:23 rostayob: well in general you have excellent support for concurrent programming in Haskell
13:23 asumu: mrcarrot: the teaching benefit is that full Haskell is too complex and has poor error messages. In theory a teaching subset of Haskell might be nice, but it's hard to implement such a thing.
13:23 asumu: (while Racket provides tools to do so)
13:24 bremner: rostayob: we were talking about automatic parallelism, or or least I was.
13:24 mrcarrot: yes
13:24 rostayob: asumu: "full" haskell is not big (haskell98)
13:24 rostayob: if with "full" haskell you meant GHC haskell... that's not haskell :P
13:24 mrcarrot: excuse me if i am asking very strange question. i am just trying to figure out in what situations racket is better and in what haskell...
13:25 mrcarrot: for teaching, i think racket is winning.
13:25 asumu: rostayob: It's bigger than Beginning Student Language.
13:25 rostayob: bremner: oh ok, even there there is a lot of interesting stuff recently (e.g. data parallell haskell)
13:25 asumu: (much bigger)
13:25 bremner: rostayob: right, it seems I lost touch
13:25 rostayob: asumu: what's your ideal "beginning student language"? I can agree that it'd be better without typeclasses
13:26 asumu: rostayob: I'm referring to the actual BSL language in Racket that you can use. I don't know what my ideal counterpart in Haskell would be, but probably it wouldn't involve using monads for I/O and graphics.
13:26 asumu: (that would come probably in an advanced student language)
13:27 mrcarrot: i have been hunting for some really easy to use graphic library for haskell... but not found any so far
13:27 rostayob: asumu: ah. I had no idea there was such a thing.
13:27 rostayob: mrcarrot: I already suggested gloss I think. that's dead easy.
13:27 rostayob: asumu: still my lecturer used haskell without ever mentioning typeclasses and things went really smoothly.
13:28 asumu: rostayob: http://docs.racket-lang.org/htdp-langs/beginner.html
13:28 jeapostrophe: rostayob: the problem is that students don't know what they should not use or that they are using it.
13:29 jeapostrophe: rostayob: a smaller language turns student's mistakes and typos into language errors
13:29 rostayob: I agree, I agree. haskell 98 is still a small language.
13:29 asumu: Yep, what jea said. Students will gleefully use state and mutation everywhere if they can find it.
13:29 asumu: Even in Racket student languages they will find it (require racket)
13:29 jeapostrophe: rostayob: a classic example is "price * quantity = total_cost" in C
13:30 jeapostrophe: rostayob: a natural thing for a student to write, but you get an incomprehensible message about pointer arithmetic
13:30 rostayob: that's not easy to do in Haskell either eh :)
13:30 (join) djcb
13:30 mrcarrot: rostayob: yes, gloss seems to be what i have been searching for for haskell
13:30 mrcarrot: rostayob: can gloss also handle mouse events?
13:30 mrcarrot: i mean to make small simple games
13:31 rostayob: mrcarrot: yes.
13:31 rostayob: mrcarrot: http://hackage.haskell.org/packages/archive/gloss/1.6.2.1/doc/html/Graphics-Gloss.html#v:play
13:31 rudybot: http://tinyurl.com/7jpo5r7
13:33 (quit) Fulax: Read error: Connection reset by peer
13:39 (join) dirKat
13:39 (quit) dirKat: Client Quit
13:42 (quit) rostayob: Read error: Operation timed out
13:42 (quit) neilv: Ping timeout: 246 seconds
13:45 (join) anRch
13:46 (quit) yoklov: Quit: computer sleeping
13:55 (join) RacketCommitBot
13:55 RacketCommitBot: [racket] plt pushed 2 new commits to master: http://git.io/J7WQ5w
13:55 RacketCommitBot: [racket/master] Fixes PR12638 - Jay McCarthy
13:55 RacketCommitBot: [racket/master] Adding #lang web-server/base - Jay McCarthy
13:55 (part) RacketCommitBot
13:57 (join) yoklov
14:01 (quit) gridaphobe: Remote host closed the connection
14:12 samth: mrcarrot: that program does not use any of haskell's parallel facilities, and if you look at the timings, you can see that all the work on the smaller sizes is done on one core
14:12 (join) dzhus
14:12 samth: the largest input gets some use of other cores, which i believe is due to the GC
14:13 samth: mrcarrot: again, haskell does not automatically parallelize your program
14:13 (join) Skola
14:14 samth: mrcarrot: note that the one-core version of that benchmark (change u64q to u64 in the url) runs in the same amount of time
14:14 mrcarrot: i never checked that...
14:14 mrcarrot: interesting
14:16 (quit) yoklov: Quit: computer sleeping
14:21 (quit) gciolli: Ping timeout: 244 seconds
14:22 (quit) itasyno: Quit: To iterate is human, to recurse divine
14:37 (join) rostayob
14:40 (quit) Skola: Quit: Lost terminal
14:58 (part) bmp
14:59 (join) bmp
15:02 (join) yoklov
15:03 (quit) anRch: Quit: anRch
15:25 (quit) mye_: Ping timeout: 248 seconds
15:29 (quit) vkz: Quit: vkz
15:42 (join) vkz
15:45 (join) Gertm
16:04 (quit) vkz: Quit: vkz
16:22 (quit) Gertm: Ping timeout: 260 seconds
16:33 (quit) asdfhjkl: Quit: Leaving
16:35 (join) gridaphobe
16:48 (join) vkz
17:03 offby1: Idle question: is there a way to make compilation fail if the racket version isn't what I want?
17:03 offby1: i.e., (require racket/5.2.1)
17:03 offby1: perl has something like that.
17:07 chandler: rudybot: eval (define-syntax (check-version stx) (syntax-case stx () ((_ v1 ...) (member (version) (syntax->datum #'(v1 ...))) #'(void))))
17:07 rudybot: chandler: your sandbox is ready
17:07 rudybot: chandler: Done.
17:07 chandler: rudybot: eval (check-version "5.2.1")
17:07 rudybot: chandler: Done.
17:07 chandler: rudybot: eval (check-version "5.2.0")
17:07 rudybot: chandler: error: #<undefined>:1:0: check-version: bad syntax in: (check-version "5.2.0")
17:08 offby1 rubs chin
17:08 offby1: can't say I understand that.  At all :-|
17:09 rapacity: just write a macro to check the version at compile time
17:10 chandler: rudybot: that's what the macro I wrote above does
17:10 rudybot: chandler: ACTION holds palm a foot above floor
17:10 chandler: gah. that was for rapacity
17:10 chandler: offby1: OK; are you familiar with `syntax-case'?
17:10 offby1: vaguely.
17:10 rapacity: yeah I know, I meant that for the person that wasn't able to understand that
17:11 offby1: it's the lowest-level way to make a macro, yes?
17:11 chandler: I'm not sure about "lowest"... but it's needed here, since the purpose is to check what (version) returns at compile time.
17:11 sid0: hm, given a symbol 'x, is there an easy way to use it in a macro (foo x) without using eval?
17:11 offby1: so your macro expands to (void) if (version) is a member of the required version ... and fails cryptically otherwise.
17:11 chandler: Yup.
17:11 offby1: which is pretty much what I want, modulo the cryptography :)
17:12 offby1 ponders
17:12 chandler: sid0: I'm afraid I don't really understand that question at all.
17:12 chandler: What are you trying to do?
17:12 sid0: chandler: sorry, let me elaborate
17:12 offby1: oh, I was thinking I'd force rudybot to fail compilation if the version is < 5.2.1, since there's a fix in that version that it requires.
17:13 chandler: offby1: that was for sid0, sorry
17:13 chandler: I figured you were trying to do something like that
17:13 sid0: chandler: I have a macro called foo which accepts syntaxes of the form (foo a b c)
17:14 (join) cdidd
17:14 sid0: chandler: and I have a symbol that's passed in to a function, and I'd like to apply that macro on the symbol without the '
17:15 chandler: rudybot: eval (define-syntax (check-version stx) (syntax-case stx () ((_ v1 ...) (member (version) (syntax->datum #'(v1 ...))) #'(void)) ((_ v1 ...) (raise-syntax-error #f (format "racket version ~a is totally borken and unacceptable to me" (version)) stx)))
17:15 rudybot: chandler: error: #<undefined>:1:0: read: expected a `)' to close `('
17:15 chandler: Gah.
17:15 rapacity: (version) returns a string
17:15 sid0: so one way to do it would be something like (define (apply-foo x) (eval `(foo ,x ,x ,x)))
17:15 chandler: rudybot: eval (define-syntax (check-version stx) (syntax-case stx () ((_ v1 ...) (member (version) (syntax->datum #'(v1 ...))) #'(void)) ((_ v1 ...) (raise-syntax-error #f (format "racket version ~a is totally borken and unacceptable to me" (version)) stx))))
17:15 rudybot: chandler: Done.
17:15 chandler: rudybot: eval (check-version "5.2.2")
17:15 rudybot: chandler: error: #<undefined>:1:0: check-version: racket version 5.2.1 is totally borken and unacceptable to me in: (check-version "5.2.2")
17:16 offby1 lights up
17:16 offby1: now we're talking!
17:16 rapacity: rudybot: (version)
17:16 rudybot: rapacity: your sandbox is ready
17:16 rudybot: rapacity: ; Value: "5.2.1"
17:16 (quit) jonrafkind: Ping timeout: 264 seconds
17:16 sid0: chandler: so I was wondering if there's a way to do it without using eval
17:16 chandler: sid0: I'm afraid you're mixing phases here. The macro doesn't really exist by the time the function is called.
17:16 sid0: chandler: it does work inside eval, though.
17:17 chandler: Yes, you can construct a new evaluator to do that, but I'm curious as to whether the task at hand really requires that.
17:18 sid0: well, I guess I could dump the macros entirely and use expressions everywhere
17:18 chandler: What is it you're actually trying to do, if I could ask?
17:19 sid0: I'm working on integrating SMT-LIB2 with Racket
17:19 chandler: offby1: So, basically what you need to do now is replace that test using `member' with something that returns #t if the version is acceptable to you.
17:19 offby1: yep
17:19 offby1: which means figuring out if the version strings sort chronologically the same as they sort lexicographically.
17:19 offby1: eli: what say you?
17:19 chandler: You might also want to accept only one version argument to `check-version' in that case, and adjust the bit in `syntax->datum' appropriately.
17:20 sid0: chandler: so I was hoping I could use the standard SMT-LIB2 forms (assert (= a b)) etc, which is why I wrote everything as macros
17:20 offby1: sure sure
17:20 sid0: chandler: but I'd like to substitute runtime values at certain places
17:21 chandler: sid0: I'd have to do a lot more digging into this than I'm prepared to do right now to determine what the right way is to do this, sorry.
17:21 sid0: since I'd like the evaluation to be at least partially driven by runtime input
17:21 offby1: Oy Gott does "make install" take forever.
17:21 offby1 twiddles thumbs
17:21 sid0: chandler: fair enough
17:22 sid0: I guess I'll just use eval for now :)
17:22 chandler: You may want to look into namespaces, if you're working with symbols that represent top-level definitions.
17:23 sid0: I'm using namespaces already
17:23 sid0: the a b etc live in the example above live a separate namespace
17:23 sid0: er, live in
17:23 sid0: ugh, sigh
17:23 sid0: the a b etc in the example above live in a separate namespace
17:25 chandler: Oh, and you're using two-argument `eval' with that namespace?
17:25 sid0: chandler: inside the macro, effectively, yeah
17:26 chandler: Doesn't `namespace-variable-value' effectively substitute for eval in that case?
17:27 sid0: Sure, but that isn't the issue, though. That part works fine
17:27 chandler: OK. I guess I'm not following, and I'm afraid I don't really have the concentration to at the moment.
17:27 chandler leaves this one for someone else
17:28 sid0: What I'd like to do is be able to write (assert (= a x-which-is-determined-at-runtime))
17:29 eli: offby1: ?
17:29 sid0 tries something
17:30 eli: is this about comparing versions?
17:31 (join) neilv
17:32 (quit) neilv: Changing host
17:32 (join) neilv
17:32 (quit) neilv: Client Quit
17:32 offby1: yes -- I'm not sure if they sort lexicographically the same as chronologically, if you see what I mean
17:32 offby1: in other words: if (string->=?(version) "5.2.1"), does that mean I'm sure to be running a version newer than 5.2.1 ?
17:33 rapacity: there are a few version checking functions in version/utils
17:33 samth: offby1: there's a whole lib for this
17:33 samth: or, what rapacity said
17:34 lewis1711: ohh people are here *rushes to paste problem*
17:34 eli: offby1: version/utils
17:34 offby1: thanks folks
17:34 eli: It's intended for that because they're don't sort lexicographically if you go back to the plt-scheme days.
17:35 lewis1711: rudybot: (define (sequenceof type?) (lambda (seq) (sequence-andmap type? seq)))
17:35 rudybot: lewis1711: your sandbox is ready
17:35 rudybot: lewis1711: Done.
17:35 sid0: chandler: Aha, got it working
17:35 lewis1711: this contract doesn't print nicely when it's broken, how can I fix that? flat-named-contract seems to rely on it ending in a question mark
17:36 sid0: quasiquote/unquote fwi
17:36 sid0: *ftw
17:36 sid0: eli: hi, what do you think of https://github.com/plt/racket/pull/87 ?
17:37 sid0: do you think that's reasonable?
17:38 eli: sid0: I get to that code yet...
17:38 sid0: sorry?
17:39 (quit) dzhus: Remote host closed the connection
17:40 eli: sid0: I didn't get to read the changes, yet.
17:41 sid0: eli: oh, all right
17:49 (join) DanBurton
17:50 DanBurton: does PLT stand for anything?
17:50 DanBurton: jeapostrophe is telling me that it doesn't stand for anything
17:50 DanBurton: I don't know if I believe him
17:54 (join) jrslepak
17:54 jeapostrophe: it doesn't
17:54 lewis1711: I thought it was Programming Language Theory
17:54 jeapostrophe: nor did it
17:54 jeapostrophe: lewis1711: nope
17:55 lewis1711: permuting lexical types
17:56 jeapostrophe: PLT = PLT Language Tower
17:56 jeapostrophe: It's a tower of languages with PLT all the way down
17:56 DanBurton: so it does stand for something
17:56 DanBurton: it's just recurive
17:56 DanBurton: *recursive
17:56 jeapostrophe: no
17:56 DanBurton: but you just said. PLT = PLT Language Tower
17:57 DanBurton: or did you just make that up?
17:57 lewis1711: jeapostrophe: http://secretgeek.net/lisp_truth.asp that reminds me of this :D
17:57 jeapostrophe: ya :)
17:57 DanBurton: jeapostrophe: that should be its backronym then
17:57 jeapostrophe: no way
17:58 jeapostrophe: it's a enigma
17:58 (join) kvda
17:58 DanBurton: but everyone will assume it's Programming Language Theory
17:58 jeapostrophe: and they are wrong
17:59 jeapostrophe: so it is a teaching opportunity
18:03 (quit) jeapostrophe: Ping timeout: 245 seconds
18:11 (quit) DanBurton: Quit: leaving
18:13 stamourv: sid0: Re SMT-LIB2: IIRC, ianj may have tried getting it to work with Racket.
18:14 (join) ShenIsBetter
18:15 ShenIsBetter: hello, dear community! What's up?
18:15 ozzloy: hola
18:18 ShenIsBetter: when the next version of (buggy) drracket will come out?
18:19 chandler: eli: This is getting tiring.
18:20 ShenIsBetter: hi eli!
18:20 ShenIsBetter: :)
18:20 samth: ShenIsBetter: new version of Racket are released every 4 months
18:21 ShenIsBetter: oh, so!
18:21 samth: ShenIsBetter: and you're close to being banned
18:21 samth: *versions
18:21 ShenIsBetter: samth, please don't be nazzy ragazzo
18:22 ozzloy: just fyi, there's a #shen about shen the programming language
18:22 ShenIsBetter: nice to know, ozzy
18:22 ShenIsBetter: i'm glad you check this out hehehe
18:24 ozzloy: me too!  are you in that channel?  i don't see you
18:24 ozzloy: by another name, perhaps?
18:24 ozzloy: has anyone written their resume in scribble?
18:25 ozzloy: anyone in here, reading this, that is
18:25 ShenIsBetter: it's not easy to be unloved, samth... :(
18:27 ozzloy: i have my resume written in LaTeX and every once in a while i think about rewriting it in something else
18:27 (quit) yoklov: Quit: computer sleeping
18:27 stamourv: ozzloy: I'm thinking about it. My website is in scribble, though.
18:27 ozzloy: oh cool
18:27 stamourv: It's really convenient.
18:27 ozzloy: at least one other person is crazy, or this idea isn't that crazy
18:28 stamourv: But it kinda looks like the Racket docs. I guess a bit of css could fix that.
18:28 ShenIsBetter: if you have your resume already written in Latex, then what's the point in conversion to scribble?
18:28 stamourv: jeapostrophe's is in sctibble to, IIRC.
18:28 ozzloy: ShenIsBetter, partially to learn scribble by way of project
18:28 ShenIsBetter: you mean: "ugly like Racket docs?"
18:28 ozzloy: ShenIsBetter, partially because i'm hoping it's easier to work with than LaTeX
18:28 samth: ozzloy: our course web pages are written in scribble: http://www.ccs.neu.edu/course/cs2510h/
18:29 samth: but i still use tex for my cv
18:29 ShenIsBetter: oh, what a nice web page, samth!
18:29 samth: ShenIsBetter: glad you like it
18:29 stamourv: ozzloy: It certainly is nicer to work with.
18:29 asumu: lewis1711: BTW there is a sequence/c in unstable/contract.
18:30 ozzloy: samth, much difference between TeX and LaTeX?
18:30 stamourv: I write my research papers with scribble.
18:30 samth: no, and i meant latex anyway
18:30 ozzloy: ic
18:31 ozzloy: samth, do you make the page look that way entirely in scribble or do you use some external css too?
18:31 ShenIsBetter: I'm fantasizing about Taylor Swift... i think i'm in love... :)
18:31 samth: ozzloy: we wrote a custom css file
18:32 samth: but scribble is nice for including that sort of thing
18:32 ozzloy: samth, icic
18:32 samth: try view-source on it, and you can see all the hooks for css
18:33 (part) bmp
18:33 asumu has been planning to redo his website in scribble/html...
18:33 asumu: I just generate HTML right now out of sexpressions.
18:33 ozzloy: man, is there any way you people can do a udacity-style class?  or team up with udacity?
18:35 samth: my website uses scribble/html (sadly still undocumented)
18:35 ozzloy: i'm a huge fan of the format and delivery of udacity.  the content could be vastly improved though
18:35 samth 's website is at http://www.ccs.neu.edu/home/samth/
18:35 ozzloy: well maybe *vastly* is overstating
18:36 asumu: Well, I saw their PL course seems kinda iffy.
18:36 ozzloy: so you have a stylesheet link to 2510H.css, i'm guessing that's the external css you used?
18:36 asumu: In that it covers something about HTML.
18:36 samth: ozzloy: yes
18:36 samth: but all the classes are natively created by scribble
18:37 ShenIsBetter: samth, please, persuade PLT-friends to make racket's logo change to Taylor Swift picture on her birthday!
18:37 asumu: ozzloy: how does udacity recruit teachers anyway?
18:37 samth: ShenIsBetter: seriously -- behave yourself
18:37 ShenIsBetter: :(
18:37 ozzloy: asumu, i'm guessing it's people sebastian knows personally
18:37 asumu: ozzloy: I like the idea of udacity, but maybe they are expanding too quickly to courses where they don't have the expertise...
18:38 asumu: I mean, it's great for robots where Sebastian has tons of experience, presumably.
18:38 ozzloy: indeed
18:38 ozzloy: he does.  he has google cars and darpa grand challenge under his belt
18:39 ShenIsBetter: A few days ago eli criticized udacity
18:39 ozzloy: i've started watching the mit lectures on SICP
18:39 ozzloy: but they're an hour long each lecture
18:39 ozzloy: and there's no feedback
18:39 ShenIsBetter: it's time that PLT-team offer something better hehehe
18:40 ozzloy: is realm-of-racket supposed to be like that?
18:40 ozzloy: anyone in here know?
18:40 ShenIsBetter: realm-of-racket is a clone of Land of Lisp
18:41 ozzloy: land of lisp is a book, realm of racket is a website, isn't it?
18:41 ShenIsBetter: realm of racket is soon to be published book
18:41 ShenIsBetter: clone of LoL
18:41 ShenIsBetter: made by felleisen's students
18:42 ozzloy: but then there's http://realmofracket.com/videos.html
18:42 ozzloy: which makes me think something in addition to book
18:42 ShenIsBetter: oh, it's just commercials
18:43 ozzloy: so you've seen the unreleased videos?
18:43 ShenIsBetter: yeah
18:43 ozzloy: is it taylor swift videos?
18:43 ozzloy: like her latest single about racket or something?
18:44 ShenIsBetter: http://stackoverflow.com/questions/9344564/commercials-for-a-new-racket-still-unpublished-book
18:44 rudybot: http://tinyurl.com/7ltvbvg
18:44 ShenIsBetter: im deeply in love
18:45 ozzloy: "this video is private" T_T
18:45 ozzloy: for both
18:45 ozzloy: mirror?
18:45 ShenIsBetter: few day ago it worked
18:45 ShenIsBetter: i've seen those videos
18:45 lewis1711: asumu: why is it unstable? :/
18:45 ShenIsBetter: notimg special
18:45 ShenIsBetter: just a bunch of geek
18:46 ozzloy: ShenIsBetter, ololo, good one
18:46 (join) jeapostrophe
18:47 ShenIsBetter: ??
18:47 asumu: lewis1711: Well, it's there because it hasn't been used/tested as much probably.
18:47 offby1: I'm doing some stuff with sqlite, and have gotten "commit-transaction: the database file is locked".  Is this the exception referred to in the docs for sqlite-connect?
18:49 ShenIsBetter: ozzloy, do you like Taylor?
18:49 ozzloy: "just a bunch of geek" doesn't add any information.  we're already in the context of discussing upcoming books on programming.
18:50 ShenIsBetter: you have read LoL, do you?
18:50 samth: ShenIsBetter: come back tomorrow if you can behave better
18:51 ShenIsBetter: pleaseee
18:52 (join) samthYoureTheOne
18:53 chandler: samth: You have a lot more patience than I do.
18:53 samthYoureTheOne: but i didn't do anything wrong....
18:54 samth: chandler: MINSWAN
18:56 samth hopes that his web server doesn't go down
18:56 chandler: Heh. I had to google that.
19:00 (nick) samth -> samth_away
19:10 (quit) MayDaniel: 
19:18 Diarmid: I'm hoping Realm of Racket will go over some of the more typically-unfamiliar things, like macros, call/cc, and maybe some Racket-specific stuff like the contracts.
19:24 lewis1711: ohh there's a racket book coming out?
19:24 Diarmid: Yes.
19:24 Diarmid: http://realmofracket.com/
19:24 lewis1711: Diarmid: yeah I hope so too. I mean I can figure out the stuff that's like other languages except with brackets
19:24 Diarmid: Sadly there's no real information about it up.
19:24 (join) MarkSoric
19:25 MarkSoric: Hello! Just wondering if someone could help me out… I have a function that contains a (read) and recurses until EOF. Is there a way to supply a file to the function so I don't have to constantly type in umpteen integers?
19:25 lewis1711: LOL why is the logo a depressed looking teenaged boy
19:25 Diarmid: lewis1711: From screwing around with EMACS in the past I managed to pick up more than that, but the way macros are handled in Scheme isn't the same as other Lisps, and frankly, some of that stuff is a bit hard to get a handle on for me.
19:26 Diarmid: lewis1711: Got me. :P
19:26 MarkSoric: I guess what I mean is: is it possible to (read) from some kind of input file?
19:27 Diarmid: lewis1711: I quite like Design by Contract, so will probably end up using that.  Given Python makes you addicted to lots of itty namespaces I may end up using that a lot too.
19:27 chandler: MarkSoric: You can either open the file as a port with `open-input-file', then read until EOF and close the port, or you can perahps just use `file->list'.
19:27 chandler: `file->list' reads all of the expressions from a given file and returns them as a list.
19:27 MarkSoric: So would I still call (read) then ?
19:27 MarkSoric: And would I get rid of the recursion?
19:27 lewis1711: Diarmid: like rackets contracts? Yeah i Like them too. though I wonder if they're slow
19:28 chandler: MarkSoric: If `file->list' works for you, then you wouldn't call `read' yourself. It recurses through the file automatically for you.
19:28 MarkSoric: chandler: Hmmm… I'll give it a shot.
19:28 MarkSoric: Thanks
19:29 Diarmid: MarkSoric: http://docs.racket-lang.org/guide/ports.html  <-- if I understand what I read from Land of Lisp right, you probably want to use call-with-input-file to make sure file handles are released and whatnot
19:29 lewis1711: it'd be really handy if you could do type annotations where needed and they could function as contracts. I know there's typed rackets but that's very typed
19:30 Diarmid: At least that seems to be the equivalent to Common Lisp's with-open-file.
19:30 chandler: Yes, that'd be the direct equivalent, but my spidey sense is telling me that `file->list' is actually the right tool in this case.
19:31 MarkSoric: Diarmid:  and chandler: Awesome - thanks guys
19:31 Diarmid: lewis1711: I use Python for a lot of stuff.  Racket won't seem slow to me after that. ;)  Also, I have more patience for waiting on the computer than I do fighting hard to find bugs.  Further, I suspect there's a way to 'compile them out' if you really want to.  Most contract support lets you, though the wisdom of doing so is questionable.
19:32 lewis1711: oh a compile them out option would be great
19:33 lewis1711: Diarmid: I use ruby mostly on a computer with less than a gig of ram and 1.8hhz process from almost a decade ago. fun times :D
19:34 Diarmid: lewis1711: The typed Racket stuff shouldn't be scary if you have familiarity with C or something similar (Java or C# for you younguns :P ).  It apparently does type inference so you don't even have to expend much effort, and if you're overly concerned about speed, I suspect it helps lots.
19:34 Diarmid: lewis1711: At least it doesn't bear much resemblance to ML/Haskell type systems to me.  I had no trouble understanding what little reading I did on it.
19:34 lewis1711: I find the error messages very confusing
19:34 lewis1711: It works ok on little functions
19:36 lewis1711: http://rosettacode.org/wiki/Power_set#Racket if anyone can typed/racket this I'd be curious (though the code itself may be garbage)
19:36 Diarmid: lewis1711: I haven't messed with it much, but it looks like the errors amount to stuff like "expected Integer, got String"  and a location, which didn't seem too bad.  It might be using the "arg1 * arg2 -> result" notation in places, which can be confusing at first unless you tried to blunder through Haskell for a while.  Maybe it gets scarier.
19:37 (quit) kvda: Quit:  -___-
19:38 Diarmid: I'm not sure how you type the dog staring at the gramophone symbol (lambda). :P
19:40 jamessan: there's a shortcut for it in DrRacket, but I forget what it is
19:40 lewis1711: ctrl-\ ?
19:41 lewis1711: yeah that's it
19:41 lewis1711: in vim it's ctr-k, l, *
19:41 lewis1711: I have a macro that renames it to "fn" for the repl
19:42 Diarmid: I must be the only person that doesn't type APL any more. ;_;
19:42 Diarmid: Does it actually work the same as if you'd typed "lambda"?
19:42 (quit) MarkSoric: Quit: MarkSoric
19:43 lewis1711: yes
19:44 lewis1711: Diarmid: http://pastebin.com/4iGDS3Rm that as a type error makes no sense to me, for example
19:44 lewis1711: For one thing there isn't even a make-sequence in there
19:46 Diarmid: It's from a macro expansion, most likely.
19:47 Diarmid: I'm not sure, but my *guess* at the problem is it wanted two things like lists or vectors, and one of them is a set.  A set doesn't have a order, at least in most languages I'm aware of.
19:48 Diarmid: lewis1711: Have you looked at it in DrRacket's macro debugger?
19:51 (join) Gertm
19:53 (quit) gridaphobe: Remote host closed the connection
20:04 (quit) m4burns: *.net *.split
20:04 (quit) dsp_: *.net *.split
20:05 (join) dsp_
20:05 Diarmid: lewis1711: This gives some tips on how to use Racket's contracts, including compiling them out: http://stackoverflow.com/questions/5671662/
20:06 (join) m4burns
20:07 (join) dnolen
20:14 Diarmid: I probably shouldn't ask, but what's the deal with the Shen troll?  Is it the developer of Shen, or some famous programming troll like Xah Lee, or somebody else?
20:29 asumu: Diarmid: Nobody knows.
20:29 asumu: I wouldn't expect Shen's developer to be that immature, so I doubt that.
20:32 Diarmid: asumu: Thanks for answering.  Well, they seem to have just shown up in the last week or so.  Maybe they'll get bored and go away.
20:32 Diarmid: For what it's worth I appreciate Racket, and this channel has been nicer to me than most.
20:32 asumu: Diarmid: They go back a bit longer. A few months ago he/she was trolling on the mailing list.
20:33 asumu: Also, good to hear. I think it's great that Racket is perceived as a nice community.
20:33 Diarmid: asumu: Someone mentioned the mailing list earlier.  How is the signal to noise ratio on that, and is there a searchable archive somewhere, in case I can dredge a answer I need off it?
20:35 asumu: Diarmid: I find the signal to noise ratio is decent, in that most of the posts are actually about Racket. You can browse it here: http://dir.gmane.org/gmane.comp.lang.racket.user
20:35 asumu: (I find the gmane interface nice, but there are others listed on the website)
20:37 asumu: lewis1711: It shouldn't be too hard to convert that to TR, I think. There are typed versions of the for macros.
20:37 lewis1711: Diarmid: yeah I am sure I could solve it by going through the macro stepper thing. but to be honest if that's the price of static typing in racket it's not really worth it
20:38 lewis1711: (for me)
20:38 (quit) gf3: Excess Flood
20:38 (join) gf3
20:39 Diarmid: lewis1711: It's probably nice for high performance, but Python has convinced me that when your code is slow, you usually need to do things a different way.  Unfortunately, there are corner cases where sometimes it's just plain old slow or ugly.  That's how I ended up here.
20:40 lewis1711: I mainly prefer types for contract reasons. which racket contracts provides and for less effort. though ideally I would just like types at top level, even within modules. I think that can be done
20:40 Diarmid: lewis1711: Luckily they are exceedingly rare.  I just ran into a patch of code where I really wanted deep (non-tail) recursion, and lambdas, which is sort of a odd experience given I haven't ever had that happen before and had messed with it for ages.
20:41 lewis1711: asumu: I will have a look for the typed for macros
20:41 Diarmid: lewis1711: I think they're fairly different notions.  Types can either be so the computer can figure out what kind of storage and instructions to use for good speed, or as a form of error checking for passing the wrong 'kind of thing' to a function.
20:42 lewis1711: both function as a form of error checking which is mainly what I'm after
20:43 (quit) mceier: Quit: leaving
20:43 Diarmid: lewis1711: Contracts seem to be all about values, instead of kinds of things.  While you can try to bludgeon value restrictions into a 'kind of thing' mold, it doesn't work very well.  You'll end up doing a lot of nonsense 'conversions', and it's hard to say "I need the variable to have this property at (only) this point in the program", like a assert would give you.
20:44 Diarmid: asumu: Thanks.
20:44 lewis1711: I don't understand the distinction you are drawing between kind of thing and values
20:47 (join) realitygrill
20:51 (quit) dnolen: Ping timeout: 272 seconds
20:54 Diarmid: lewis1711: While I could conceivably make types for every possible combination of traits I want, working with them would be unpleasant.  "non-zero" isn't so bad.  "power of 2", by itself isn't so bad either.  But when we get into "non-zero power of 2 less than or equal to 1024" (e.g. a dimension for a texture in a game system) things are getting a bit silly.
20:54 lewis1711: oh right
20:55 lewis1711: I guess maybe contracts fit more into my style of thinking then
20:55 lewis1711: because to me "values" are more important
20:55 lewis1711: and flexible
20:55 Diarmid: lewis1711: I already mentioned that you can force everything into a type system, but I find having preconditions, postconditions, (and in OO) inherited invariants (lets you know if the object got corrupted, and why), (in imperative code) loop variants/invariants (lets you know if the loop 'hung' and why), and assertions (lets you say you want a trait at just a particular point, like checking for zero before division) to be more
20:56 jamessan: Diarmid: you were cut off at "to be more"
20:56 Diarmid: "to be more natural"
20:57 lewis1711: also Diarmid: I saw that stackoverflow about compiling out contracts. seems irritating though, I'll just leave them in there for now, and maybe unit test instead if a contracted function proves too slow
20:58 Diarmid: lewis1711: Yes, I feel the same way, which is why I'd like something that does some of the things 'fancy' type systems do (e.g. I think CQual is really jiffy http://www.cs.umd.edu/~jfoster/cqual/ ), I think most of them have gone about it the wrong way.  But that's just an opinion.  Plenty of people like ML (and its variants), Haskell, and the various dependent-typed languages.
20:59 lewis1711: haskell is by far the best static type system I've used
20:59 lewis1711: it really makes other static type systems seem irritating in comparison
21:02 Diarmid: lewis1711: It wasn't for me I guess.
21:04 lewis1711: oh? what sort of static typing do you like?
21:04 Diarmid: lewis1711: BTW, if contracts and unit testing interest you, you might want to consider using both.  The Cobra language encourages that.  The idea is to use the unit tests to 'perturb' the code, and the contracts to isolate the bugs where they originate, so you don't spend as much time in the debugger.
21:05 lewis1711: call me close minded but I can't really be bothered with "pragmatic" scripting languages. I like them to have a theme, like pure OOP (ruby, smalltalk) or everythings a list (racket)
21:05 Diarmid: lewis1711: I haven't liked anything that went beyond what was necessary for generating decent assembly.  I do think C/C++ could be improved upon some there, but it'd take doing it in a different language since it would break things.  For instance, things old Fortran 77 couldn't do (like aliasing via pointers) allow it to be more heavily optimized.  Heavy annotation could let something /like/ C do a better job.
21:05 lewis1711: I always thought if you're going to abstract that much why bother reimplementing BASIC or some kind of dynamic BASIC
21:06 lewis1711: sorry not dynamic Basic, dynamic... uh
21:06 lewis1711: Algol
21:06 lewis1711: so you're just not a fan of static typing in general?
21:07 Diarmid: lewis1711: I used to think that.  I do think there are too many Fortran + call stack + classes /but with new syntax!/ in the world.  Some 'genres' of programming are basically dead (e.g. dataflow), despite being rather nice for certain tasks.
21:07 asumu: lewis1711: I think there's a bug in TR that makes it hard to type-check your powerset code... so let me investigate that.
21:08 lewis1711: asumu: woohoo, inadvetently discovering bugs with my noobishness
21:09 Diarmid: lewis1711: In any case I've found that 'purity' doesn't work out so well.  Sometimes (notably when working with math) what you really need *is* a function, not a class.  I also like being able to write most of my code in a ... well, peculiar, style.  I won't call it functional, because it's not, but I rather like operations on aggregates, and method chaining.
21:09 asumu: (namely (Setof x) is not a (Sequenceof x) for some reason)
21:10 lewis1711: Diarmid: my sweet spot is generally close to an ideal, but with some compromises. like ruby, where *almost* everything is an object
21:10 Diarmid: lewis1711: So I like being able to use very imperative code for I/O, and have the functional-oid, or whatever you want to call it, code separate, but I don't want to deal with multiple languages for that.  Looks like Lisp and it's macros will let me do that without much pain.  Lisp records and set give you OO.  Most of the rest is functional.  You can build the rest on as you want it.
21:10 lewis1711: but I have really gone off sort of "wish-washy" multiple paradigm stuff. I
21:10 asumu: lewis1711: BTW, Racket is definitely trying to shake off Lisp's "everything is a list" which is probably a flaw and not a feature.
21:11 lewis1711: asumu: what!? how will racket even be racket if the whole code isn't a big a list?
21:11 asumu: Racketeers frequently use structs, hashtable, etc. The syntax itself is represnted as syntax objects.
21:11 asumu: Which technically aren't lists (though they wrap them).
21:12 lewis1711: ohh, no Id on't mean list is the only data structure. I just mean the homoiconity
21:12 lewis1711: and everythign being inside of it
21:12 lewis1711: I am a big fan of sets, as you may have guessed :)
21:12 asumu: Well, I've never thought homoiconicity is a well-defined term. I mean, does Racket have that? After all, the syntax is not equal to quoted s-expressions.
21:12 lewis1711: Diarmid: racket has an OOP system
21:12 asumu: You have to use datum->syntax to get that.
21:13 asumu: But I get your point.
21:13 lewis1711: I think this stuff is over my head. tbh I just like the uniform syntax, it's a big feature for me
21:13 Diarmid: asumu: I hope it doesn't lose too much in the way of Lisp syntax and semantics.  The only syntax I like better is assembly and Forth, because it's all I can think of with less syntax. :P  Blame dealing with C++ too long.  I hate complex syntax.  I also like what Lisp's syntax lets you do with macros and the like, so they aren't just a way to obfuscate your code like they tend to be with cute preprocessor macros and template meta
21:13 Diarmid probably got cut off there somewhere.
21:14 lewis1711: template meta... (programming I guess)
21:14 asumu: Diarmid: I don't think you need to worry about that. I was really just talking about representation, not so much how macros look/syntax.
21:14 lewis1711: ah
21:14 Diarmid: Yeah "metaprogramming."
21:15 Diarmid: asumu: That's a relief.   I suppose by having a more complex internal representation you could do better static checking or something, which is always nice.
21:17 lewis1711: speaking of brackets, is there a way to get brackets to match in xrepl or the like? like when you type a closing ), it briefly goes to its corresponding ( then back
21:17 asumu: Diarmid: syntax objects let you correlate your syntax with locations in your original program. They're also how the macro system internally tracks marks for hygiene and such. AFAIK anyway.
21:17 (join) dnolen
21:18 Diarmid: lewis1711: I assume Racket's OO system is inspired by CLOS.  Most are.  But even without it apparently doing similar things with Common Lisp and Scheme records isn't that hard.  I brought it up because it lets you just do I/O imperatively, without any suffering, unlike 'pure' (no such thing, if you have I/O) declarative languages, and do the rest in whatever style pleases you (that you can build on the imperative or functional p
21:18 Diarmid: (that you can build on the imperative or functional parts).
21:19 lewis1711: the CLOS one was called swindle and is now considered deprecated (I think).
21:19 lewis1711: the current one is single dispatch
21:20 Diarmid: It'd be interesting to see how hard it would be to embed something like Lucid in Racket.  I/O is already done with streams, so it seems natural for certain kinds of tasks, like audio/video processing, and maybe certain kinds of networking tasks.
21:21 (quit) dous: Remote host closed the connection
21:21 lewis1711: I remember reading this paper on some Self or smalltalk derivative... which had prototype based double dispatch. it bent my fragile undergraduate mind.
21:22 Diarmid: That would have been Self, yes.  It was, I think, the first 'prototype based' language, like JavaScript (EMCA-Script, or whatever, now).  Basically you don't have a class, instead you 'copy' objects from the few the system starts with.
21:23 Diarmid: I'm not all that thrilled with changing the semantics of stuff at run-time (which may make me bad company :P I know that's popular with Lisp and Ruby folks).  The main thing I was impressed with it was the incredible performance they got out of their VM.
21:25 lewis1711: Diarmid: it was Slate, not Self. Self is single dispatch
21:26 Diarmid: I'm more fascinated by new string mauling techniques (e.g. SNOBOL, Icon, Scsh's SRE's) since I do that a lot, and weird evaluation methods (e.g. Forth, Pure (not sure what the first language of this type was), Lucid, Prolog, Refal...).
21:28 Diarmid: There doesn't seem to be much interest or appreciation of oddball evaluation methods any more.  It can make certain tasks easier to describe in those languages, usually while making others very painful.  I think it's odd that algorithms get so much focus when they only solve one problem, but programming languages which can theoretically solve any problem get much less.
21:30 Diarmid: Formal verification is pretty neat to me, as are object-capability security (mostly at the OS, rather than language level; i.e. installation endowments and powerboxes, not unforgable references), but I'm guessing those won't go anywhere due to lack of cooperation/demand.
21:31 lewis1711: I don't even know what an evaluation method is :(
21:31 Diarmid: Imperative code is like a list of instructions from a manual.  It's sequences with branches and loops.
21:32 Diarmid: A lot of the time functional programming isn't that different.  Lazy evaluation can change that.  In that case stuff only gets executed if the results are actually *used*.
21:33 Diarmid: Beyond that, we get into stuff most people don't bother studying any more.  Like with dataflow-style evaluation you can do something called speculative evaluation, which is where things are calculated as soon as all inputs are (re)defined.  It's essentially the opposite of lazy evaluation.
21:34 Diarmid: As for why you might want to do that, the processor is idle most of the time on modern systems.  If a lengthy computation can be run before the user asks for the results, then the results can often be given instantly.
21:34 lewis1711: haha, that's quite amusing
21:35 Diarmid: And then their's weirder stuff, like Prolog, which is pretty hard to explain...  it's easier to just mess with it.
21:35 lewis1711: computers have stopped getting faster as quickly though haven't they?
21:35 lewis1711: ohh prolog. I did a bit of that for university. It never quite clicked with me
21:36 Diarmid: Forth is stack based, and not hard to explain, so you write things like "2 2 +", which shoves a 2 and a 2 onto the stack, which can be viewed as 'functions' that return themselves, and then the + is a function that pops 2 arguments and pushes the result.  That's yet another kind of evaluation...
21:36 lewis1711: Forth has been on the list to learn for a while
21:36 lewis1711: that or Factor
21:36 Diarmid: Pure works by solving equations, algebra-style.  Refal chops strings into trees and transforms them...  All weird evaluation methods, but handy for certain things.
21:37 Diarmid: Yes, Moore's law broke down somewhere around the early 'zeros' (2000 and thereabouts) for normal processors.  It has not broken down for video cards, which have actually exceeded it, if memory serves.
21:39 lewis1711: I wonder if that's a good or bad thing for computing
21:39 lewis1711: maybe good, because it will force efficiency
21:39 lewis1711: but bad, because it delays abstraction
21:39 Diarmid: That said, I don't think it's that big a deal.  We have far more processing power than we *need* for even really good looking 3-D games.  The only reason things are slow are due to implementations being inefficient.  There's no way a chat client like the one I'm using needs 200 to 300MB of RAM (what mine uses after a day or two).
21:40 Diarmid: I had a IRC client on my Pentium 1, and that system only had 16MB. :P  The client was fancier too.
21:40 lewis1711: yeah but what would you do, write everything in highly optimised C?
21:41 Diarmid: I think there's a lot of alternatives.  First, we know only about 3% of the code 'matters' performance wise (I can point you at the paper if you care).  Just fixing the algorithms there will drastically improve things.
21:41 lewis1711: also don't most operating systems allocate far more ram then they need to, because ideal ram is pointless?
21:41 lewis1711: *idle
21:42 Diarmid: I think there's a misunderstanding about how high level programming languages effect code generation.  You hear a lot of stuff like "because X and Y is implementation defined, and you don't specify how everything is done, it can be optimized much more heavily!", or better yet "there are no slow programming languages, only slow implementations!"
21:44 Diarmid: Well, sorry, no, there are slow programming languages.  They're inherently slow because of things the compiler /just can't know/, so it can't perform a optimization for you.  This is why (pure; no cheating using C/assembly modules) Python will always get spanked by Fortran.  There's aliasing.  You can redefine *everything* at run time.
21:46 Diarmid: As for OS's, I guess 'that depends'.  It depends on the OS and architecture. There are things to think about like whether the processor will tolerate things that aren't aligned by some amount (as opposed to crash the kernel or app outright).  There's questions of page size, because the finest grain you can allocate memory to a process is at the page level, at least on most hardware with memory protection boundaries.
21:46 Diarmid: The OS isn't to blame for most bloat though.
21:47 Diarmid: You can't even blame garbage collection.  Early Lisps often did useful things in 64 to 256kB of RAM.  You can still dredge up copies of Micro-Lisp and VLISP 8 and whatnot for things like the Commodore 64.  Also, OCamel often ties with C and C++ in benchmarks, despite garbage collection.
21:49 Diarmid: Though the love of lazy garbage collection probably isn't doing anybody any performance (grinds the disk and trashes the cash) and memory efficiency (wait to free the memory whenever you feel like it) favors.  OCaml's 'trick' is trying to eliminate GC where it can do so safely.
21:50 Diarmid: Anyway, what we could do is start writing the 3% our code 'that matters' efficiency wise in some restricted language (e.g. no aliasing) with heavy annotations to help the optimizer, then let it spew out the assembly, and link it to our comfy high level languages, and maybe tweak the garbage collectors some.
21:54 lewis1711: I always thought the C + high level approach was good. though I dunno, people like the CLR/JVM approach
21:56 Diarmid: There's some very good things about the CLR (and to a lesser extent, the JVM).  The CLR has been made very friendly to 'weird' languages.  It makes it possible to use everybody else's stuff from your pet language, and people using common things like C# to use yours (if your implementation was designed with this in mind, and it should be).  This never existed before.
21:57 (quit) wtetzner: Remote host closed the connection
21:57 Diarmid: The JVM doesn't seem to be very friendly to weird languages, which is why I guess things like Clojure are missing some typical Lisp features, but you can get kinda close.
21:58 Diarmid: And BTW, both can call native code, so there's nothing saying you can't compile to CLR or JVM and do what I said.
21:58 Diarmid: You're stuck with the garbage collector behaving less than ideally though.
21:59 lewis1711: oh they *can* call native code
21:59 lewis1711: sure
21:59 lewis1711: but in practice it's so much easier in racket or ruby or lua
22:00 Diarmid: Well, none of those things use the CLR or JVM (though I think there might be JRuby).
22:00 Diarmid: But all of those can also call native code, so can still potentially perform really fast. :P
22:01 Diarmid: Ruby may suffer from some of the issues Python has.  I know monkey patching is highly encourage there, and that does pretty much doom your performance.
22:01 lewis1711: huh? that's not what I said
22:01 Diarmid: But I guess you could write anything memory wasting or slow in the native code.
22:02 lewis1711: I said it's much easier to call native code in those 3 languages, than it is in a JVM or CLR language
22:03 (quit) Nisstyre: Remote host closed the connection
22:03 Diarmid: Oh, I dunno about that.  What I saw in Java and C# didn't look all that awful.
22:03 Diarmid: Not that I've tried to do it mind you. :P
22:04 lewis1711: haha. racket isn't too bad
22:04 lewis1711: the languages I've successfully done this with are racket, lua, ruby, gambit scheme (completely trivial as it compiles to C). I failed with java and python
22:05 Diarmid: I've basically given up on 'the world' though.  I'm satisfied with writing inefficient but pretty stuff (design wise) for myself.  Stuff I've done for other people has mostly been code that will probably have to be changed completely in 6 to 12 months, since it's all glue anyway.
22:07 (join) Nisstyre
22:09 Diarmid: Well, if you need a Scheme that's great for speed, I seem to remember Stalin was pretty impressive. It also compiles to C.
22:09 Diarmid: Python is known to be... difficult.
22:09 lewis1711: meh racket is plenty fast enough
22:09 Diarmid: Stackless Python is supposed to be easier to link to native code than CPython.
22:10 Diarmid: Python also has some restrictions that can be frustrating for embedding, namely that it's necessarily single-threaded.  The closest you can get to real multithreading is processes and pipes, which arguably isn't so bad, but doesn't sit well with a lot of people..
22:13 (quit) rostayob: Ping timeout: 244 seconds
22:20 Diarmid: lewis1711: On the 'idle ram' comment, I like how Linux does things.  Any RAM that isn't allocated to a process is used for disk caching.  If you have a bunch to spare, this can be great.  You generally don't want to set your drives up to have a long weight before flushing (since you could lose data if the power goes out or something), but it's really great for read-caching.
22:22 (quit) vkz: Quit: vkz
22:31 (join) dous
22:32 (join) asdfhjkl
22:40 (quit) masm: Quit: Leaving.
22:46 (join) yoklov
22:57 (join) vkz
23:06 (quit) derrida: Ping timeout: 260 seconds
23:11 (join) derrida
23:18 (join) mithos28
23:26 (join) neilv
23:30 (part) lewis1711
23:40 (join) MarkSoric
23:40 neilv: editing a few planet packages at once, having most of the files named main.rkt or info.rkt is not the greatest
23:41 (part) MarkSoric
23:47 (join) mye_
23:48 (quit) dnolen: Ping timeout: 252 seconds
23:55 neilv: this is how response/html-template expands when you have dynamic parts and you don't care about ordering, flushing, or size. http://paste.lisp.org/display/128388
23:56 neilv: when the template is all static, it will always do this: http://paste.lisp.org/display/128380