00:00:00 --- log: started haskell/07.09.14 00:01:03 --- quit: coffeemug (Remote closed the connection) 00:01:18 huh you can't make it tell you which parser it errord in can you? 00:01:31 I don't think that would be possible. 00:01:59 you mean if it's stuck in an infinite loop? 00:02:05 @type Text.ParserCombinators.Parsec.() 00:02:10 Couldn't find qualified module. 00:02:11 When it says, Parse error:(line 1, column 10): 00:02:18 yes, you can 00:02:25 parsec has powerful error handling stuff 00:02:30 a little complex, but powerful :) 00:02:56 --- quit: ToRA (Read error: 110 (Connection timed out)) 00:03:12 --- join: dc__ (n=dc_@62-31-55-56.cable.ubr01.camd.blueyonder.co.uk) joined #haskell 00:05:30 @type (Text.ParserCombinators.Parsec.) 00:05:32 forall tok st a. Text.ParserCombinators.Parsec.Prim.GenParser tok st a -> String -> Text.ParserCombinators.Parsec.Prim.GenParser tok st a 00:05:44 --- quit: drice (Read error: 110 (Connection timed out)) 00:06:48 --- join: doserj (n=doserj@dewey.inf.ethz.ch) joined #haskell 00:07:24 can now tokenize :D 00:07:34 next to parse intro prolog AST 00:07:55 --- join: amiddelk (n=arie@melinda.cs.uu.nl) joined #haskell 00:11:14 --- quit: cognominal_ (Client Quit) 00:13:20 --- join: Tac-Tics (n=tactics@71.239.99.94) joined #haskell 00:17:41 --- quit: scodil ("Lost terminal") 00:17:53 --- quit: Pseudonym ("using sirc version 2.211+ssfe") 00:19:17 --- quit: glguy ("leaving") 00:20:07 --- quit: ChilliX () 00:24:31 --- join: njbartlett (n=njbartle@cpc2-barn4-0-0-cust110.brnt.cable.ntl.com) joined #haskell 00:25:18 --- quit: dc__ () 00:28:10 --- join: boegel (n=boegel@konijn.elis.UGent.be) joined #haskell 00:31:03 --- join: etnt (n=tobbe@84-55-124-2.customers.ownit.se) joined #haskell 00:35:31 --- join: iblechbot (n=iblechbo@ppp-62-216-197-89.dynamic.mnet-online.de) joined #haskell 00:36:09 --- quit: sjanssen ("Konversation terminated!") 00:36:22 --- join: Ben` (n=Ben@cpc2-hem12-0-0-cust201.lutn.cable.ntl.com) joined #haskell 00:37:05 --- join: ADEpt (n=ADEpt@jabber.hst.ru) joined #haskell 00:37:23 --- quit: Weremanatee () 00:39:57 --- join: IvdSangen (n=ivdsange@s55914d08.adsl.wanadoo.nl) joined #haskell 00:45:24 --- quit: RogerTaylor () 00:47:29 --- quit: fluctus (Read error: 110 (Connection timed out)) 00:47:29 --- join: triplah_ (n=tripped@ppp59-167-101-56.lns3.bne1.internode.on.net) joined #haskell 00:47:38 --- quit: qwwqe (Read error: 110 (Connection timed out)) 00:50:11 --- quit: ivo (Read error: 110 (Connection timed out)) 00:50:18 great, i have a lazy semantics checker, which does not check those types which are not used in a code later on ... what means there can be illegal definitions and I do not catch them :-) 00:50:29 how can I force a parameter of a function to evaluate? 00:50:59 checktypes te ve e = ... -- i would like that te and ve would be strict 00:51:35 @go bang patterns 00:51:38 http://hackage.haskell.org/trac/haskell-prime/wiki/BangPatterns 00:51:38 Title: BangPatterns - Haskell Prime - Trac 00:52:10 I've never used them, I'm not sure if they're actually in ghc yet 00:52:39 If they aren't, there's always seq, though figuring out how to use it is another matter 00:53:14 --- quit: Plareplane (Read error: 104 (Connection reset by peer)) 00:55:08 Though that won't actually help you make your semantics checker report any more errors unless you use _|_ 00:57:34 Mitar: I would be very surprised if bang patterns are the correct way to solve this problem 00:57:58 hmm 00:58:17 so when i am adding new defined types to type environment i check them 00:58:18 --- join: RogerTaylor (n=rogertay@58.186.58.197) joined #haskell 00:58:20 normally the 'type' of the whole program 00:58:29 implies that all the sub parts are typed 00:58:32 in some sense 00:58:34 but the problem is that if i never read this type from the environment it is never checked 00:59:09 you could see this as a disadvantage of the way you use 'error' instead of a proper representatio of failure 00:59:19 if you used a proper representation of failure you could write: 00:59:23 welltyped :: Program -> Bool 00:59:46 Mitar pasted "types" at http://hpaste.org/2741 01:00:07 yes, i do this :-) 01:00:28 but it finds it too early it is OK 01:00:39 for example, in my language this is not possible 01:00:57 as types have recursive loop 01:01:02 well the point is, if you represent errors explicitly 01:01:06 such as with an Either type 01:01:14 then 'welltyped' automaticlaly forces evaluation 01:01:19 since it checks for 'Right' everywhere 01:01:28 yes 01:01:30 i know .. 01:01:30 --- join: Plareplane (n=Plarepla@216-197-152-148.sktn.hsdb.sasknet.sk.ca) joined #haskell 01:01:33 any other way? 01:01:33 and you don't need to mess around with strictness anotations and bang patterns 01:01:47 i will have to mess much more with this ... 01:01:54 (i use error now everywhere) 01:01:59 --- join: Itkovian (n=Itkovian@dragnipur.elis.UGent.be) joined #haskell 01:02:08 and to check now for Either all around ... 01:02:21 and to change to do ... 01:02:34 then you have to write an ugly 'welltyped' which sprinkles 'seq' everywhere 01:02:39 to check for errors 01:03:21 i have to force stricknes only on environment variables 01:03:28 and on just one place 01:03:34 well you could check length t > 0 01:03:40 hmm 01:03:42 since your types are strings 01:03:51 that will force them not to be _|_ 01:04:08 the length of environment? 01:04:23 are bang patterns defined in ghc? 01:04:38 no, the length of the individual type strings 01:04:46 but then you can map that over the whole environment 01:04:54 hmm 01:05:05 what type does your environment have? 01:05:07 Map String String ? 01:05:18 --- quit: njbartlett () 01:05:26 --- join: Saul__ (n=Saul@10-15-ftth.onsnetstudenten.nl) joined #haskell 01:06:36 --- join: QplQyer (n=jeroen@state.ugent.be) joined #haskell 01:06:41 no, [(String,Type)] 01:06:56 where the type is data Type = TyNil | TyUnknown | TyVoid | TyInt | TyString | TyFunc [TypedVar] Type | TyArray Type | TyRecord [TypedVar] | TyVar TypeId 01:06:56 --- join: drrho (n=rho@chello213047112079.11.11.vie.surfer.at) joined #haskell 01:07:02 :-) 01:07:27 maybe i could use a quard for this :-) 01:08:01 typecheck te ve e | forceStrict te && forceStrict ve = ... 01:08:02 :-) 01:09:11 forceStrict = id . ($!) 01:09:11 ? 01:09:21 hm, no 01:09:26 doesn't really work like that 01:09:30 because && is lazy 01:09:46 EVERYTHING is lazy, remember that 01:09:51 :-) 01:10:03 and just using 'seq' on your te and ve will acheive nothing 01:10:12 all that proves is that the list exists 01:10:16 (the list of pairs) 01:10:26 this is why I really don't recommend bang patterns, or seq 01:10:51 I recommend thinking of a function which by executing it obviously requires the types. 01:11:10 (well actually I recommend using a proper error signalling and not abusing error ;P) 01:11:17 even strict things are lazy! Calling them "eager" is a calumny, they're only evaluating because the runtime is breathing down their necks, and they decide it's less hassle to just play along... 01:11:39 lol 01:12:12 you can use deepSeq or rnf if you like 01:12:16 * quicksilver shrugs 01:12:19 I despise them 01:12:22 ;) 01:12:42 ok, what would be the proper error handling? using either? but then I have to either use a monad and rewrite the code into a do notation or use Either and manually pass the error message higher if it comes 01:13:22 you don't have to use do notation to use a monad 01:13:38 > Right 4 >> Right "foo" 01:13:39 Add a type signature 01:13:47 > Right 4 >> Right "foo" :: Either String String 01:13:48 Right "foo" 01:13:55 > Left "oops" >> Right "foo" :: Either String String 01:13:57 Left "oops" 01:14:01 no 'do' in sight :) 01:14:30 --- join: igel (n=igel@brsg-4db2c402.pool.einsundeins.de) joined #haskell 01:15:19 "Right" and "Left" are *terrible" names for "OK" and "Error"... 01:15:21 --- join: int-e (n=noone@td9091a30.pool.terralink.de) joined #haskell 01:15:41 I don't think they're that terrible 01:15:47 but it would be more idiomatic to write 01:15:59 > throwError "oops" >> return "foo" :: Either String String 01:16:00 Right is the right answer. :) 01:16:00 Left "oops" 01:16:17 and Left is the, er, not-right answer? 01:16:43 Mitar: to use 'rnf' from Control.Parallel.Strategies you would, I think, have to derive NFData for your type 01:16:53 Maybe Either should be changed to the Picard monad instead of Maybe. 01:17:03 Picard? 01:17:24 data Picard a b = MakeItSo a | BelayThat b 01:17:29 lol 01:18:02 hah 01:18:31 but at least that indicates some sort of Yesness and Noness, rather than 2 alternatives (which is what Either says to me) 01:18:56 Well, Either is also used to be the sum of two types, so it's got kind of a dual role. 01:19:28 It just happens that the monadic interpretation of Either is error handling. 01:20:19 Be glad they didn't use the category theory (I think) names like Coq. :) 01:20:27 you could use it for nonlocal returns of a fixed type. errors are just a special case of that. 01:20:31 data Either a b = Inl a | Inr b 01:20:40 ChilliX++ # "This is GHC after all, it tries to gently nudge you in the right direction, but if you insist, it happily let's you drill arbitrarily large holes in your foot." 01:21:03 actually I think osfameron is right 01:21:08 I think that Either is a bit generic 01:21:20 and it would make sense to have another (isomorphic) type with better constructor names 01:21:27 but I also don't think it's all that important :) 01:21:29 Yeah, it might be better if there were Error in addition to ErrorT. 01:21:51 It is of course easy enough to define your own 01:21:55 and make it an instance of MonadError 01:24:09 Mitar pasted "Either?" at http://hpaste.org/2742 01:24:25 quicksilver, but even Either is not a solution 01:24:47 Mitar: different problem, I think 01:24:53 --- quit: Itkovian () 01:24:55 Mitar: cyclicity you're going to have to check for explicitly 01:25:03 no no 01:25:03 Mitar: that's different from being too lazy about type errors 01:25:08 i check it in the newenv 01:25:26 and the problem is that because typecheck te ve (IntLit _) = Right TyInt never consultes the te and ve 01:25:32 te' and ve' are never evaluated 01:25:33 --- join: Itkovian (n=Itkovian@dragnipur.elis.UGent.be) joined #haskell 01:26:02 Mitar: the newenv needs to be in the either monad in a similar way 01:26:30 Mitar: then you do (te',ve') <- newenv (te,ve) 01:26:43 Mitar: and that forces it to be right and not left 01:28:28 Using recent ghc snapshot: Is there a easy fix for Network/Socket.hsc:1901:9: Not in scope: `openeFd'? 01:28:54 Change it to "openFd"? 01:31:39 dolio: Sorry. That has been my type. It is openFd 01:31:49 But that seems to be no longer present ? 01:31:56 Oh. I'm out of ideas, then. :) 01:32:17 MarcWeber: you may have better luck on the #ghc channel 01:32:28 MarcWeber: but maybe it's just a broken snapshot. Happens sometimes. 01:34:06 --- join: KatieHuber (n=katie@203-97-111-120.cable.telstraclear.net) joined #haskell 01:34:59 --- part: ADEpt left #haskell 01:35:31 --- join: ADEpt (n=ADEpt@jabber.hst.ru) joined #haskell 01:35:38 MarcWeber: does this help? http://www.google.com/search?hl=en&q=openfd+site%3Ahaskell.org 01:35:39 Title: openfd site:haskell.org - Google Search 01:36:21 I had a similar problem. There's a discontinuity between the POSIX lib and the Network lib. 01:37:23 Thanks! 01:38:22 --- join: qwwqe (n=fluctus@unaffiliated/fluctus) joined #haskell 01:38:22 --- join: fluctus (n=fluctus@dhcp-0-50-f2-7c-be-b7.cpe.mountaincable.net) joined #haskell 01:39:56 --- join: roman (n=shkoder@ll-226.216.82.212.sovam.net.ua) joined #haskell 01:41:49 --- quit: fax () 01:43:25 --- quit: igel ("Leaving") 01:43:46 --- quit: Saul__ (Remote closed the connection) 01:44:14 --- quit: Pupeno (Read error: 110 (Connection timed out)) 01:47:54 --- join: mauke_ (n=mauke@d83-189-123-154.cust.tele2.de) joined #haskell 01:48:09 --- join: chris2 (n=chris@p5B16C1E8.dip0.t-ipconnect.de) joined #haskell 01:50:10 --- join: Rebooted (n=opera@88-104-185-170.dynamic.dsl.as9105.com) joined #haskell 01:50:15 --- quit: QplQyer ("*poef* *roekwolk*") 01:51:14 --- join: dv^1 (i=dv@gateway/tor/x-5507f6bf451e3424) joined #haskell 01:51:26 --- part: Rebooted left #haskell 01:52:53 --- join: b_jonas (n=b_jonas@dsl51B619BC.pool.t-online.hu) joined #haskell 01:53:02 --- quit: gip ("Leaving") 01:53:41 --- join: Qerub (n=qerub@kr-lun-120-155-233-83.3.cust.bredband2.com) joined #haskell 01:53:48 --- join: patrick_ (n=patrick@88-96-27-150.dsl.zen.co.uk) joined #haskell 01:54:29 Does Haskell have any nice syntax for Map literals? Something like { "foo" => 2, "bar" => 3 }? 01:54:31 --- join: ToRA (n=tora@colorado.doc.ic.ac.uk) joined #haskell 01:55:17 I think just 01:55:20 yourmap x = case x of { "foo" -> 2; "bar" -> 3; } 01:55:28 fromList [("foo", 3), ("bar", 7)] 01:56:27 (To answer very literally, Data.Map gets no special syntactical status.) 01:57:54 Qerub: if you want some special syntax, introduce your own operators (functions + fixity decs) 01:58:14 --- part: scook0_ left #haskell 01:58:14 Examples abound, e.g. HaskellDB 01:58:36 Alright! Thanks for the answer. 01:58:39 "=>" is reserved IIRC, but you could use "==>" 01:58:42 or something 01:58:57 (k,v) ==> m = Map.insert k v m 01:59:22 k ==> v = Map.singleton k v 01:59:24 ("foo",2) ==> ("bar",3) ==> empty 01:59:41 I was thinking: fromList ["foo" ==> 3, "bar" ==> 7] 01:59:51 but taruti's idea is nice too 01:59:52 (&) = Map.union 02:00:03 x ==> y = (x,y) 02:00:04 "foo" ==> 2 & "bar" ==> 3 02:00:18 go to town :-) 02:01:33 qerub: ah sorry I totally misunderstand your question. 02:03:17 --- join: seafoodX (n=sseefrie@alphad261.lnk.telstra.net) joined #haskell 02:04:08 --- join: dibblego (n=nobody@220-245-107-64.static.tpgi.com.au) joined #haskell 02:04:46 --- quit: mauke (Read error: 110 (Connection timed out)) 02:05:02 --- quit: dv^^ (Remote closed the connection) 02:05:17 --- join: tuxplorer (i=tuxplore@nat/yahoo/x-264e8c262e363ad0) joined #haskell 02:05:24 --- quit: patrick__ (Read error: 113 (No route to host)) 02:06:04 is there a function in haskell to strip a string of all mysql non-compatible characters and to avoid SQL injection? 02:06:43 that's certainly not really a question for haskell 02:06:49 it's something you'd hope to find in the SQL libs 02:07:19 which db lib are you losing? 02:07:57 --- nick: mauke_ -> mauke 02:08:19 doestn' look like HSQL does 02:08:56 --- join: mr_tenor (n=nick@c220-239-43-146.rivrw7.nsw.optusnet.com.au) joined #haskell 02:08:59 HDBC has a sane design using parameters 02:09:04 so you don't need 'escape' 02:09:11 or that's what cosmicray claims in the docs 02:10:04 I think HDBC is a bit more shiny than HSQL, overall 02:10:07 although I've used neither 02:10:48 --- join: cognominal_ (n=cognomin@62.210.200.99) joined #haskell 02:12:52 HSQL doesn't have parameters? 02:14:20 --- join: slipstream-- (n=irssi@ppp85-140-74-51.pppoe.mtu-net.ru) joined #haskell 02:14:32 osfameron: maybe it does, I didn't look hard enough 02:14:38 osfameron: it seems to be mostly String based though 02:14:44 @paste 02:14:44 Haskell pastebin: http://hpaste.org/new 02:15:24 either way, I don't think I could use a DB system that didn't support placeholders. It's more or less insane. 02:15:45 certainly insane if you have untrusted data 02:15:49 have you looked at haskellDB? 02:15:53 that's a much more interesting approach 02:16:06 not SQL based at all; has its own ADT representation of queries 02:16:12 nope I haven't. is that the queries-as-haskell-data approach ? 02:16:17 that's right 02:16:35 like DBIx::Class in concept, but arguably it's quite a bit easier to get it looking nice in haskell 02:16:51 with a type-system and custom operators and stuff 02:17:21 true. I can vaguely imagine some goodness coming out of that :-) Will play with it at "some point" 02:17:58 --- join: roconnor (n=roconnor@vhe-540354.sshn.net) joined #haskell 02:18:14 --- quit: enolan ("Leaving") 02:18:27 hi, i've 750M tab-delimited database which has to be read in order to populate sqlite3 database. here is python script (not mine) used to grok 10M file, but it cannot handle 750M. what are the prospects of haskell handle it with hdbc-sqlite, ie. can haskell's laziness do the job? 02:18:31 gour pasted "python script" at http://hpaste.org/2744 02:18:50 --- join: njbartlett (n=njbartle@194.42.125.16) joined #haskell 02:19:23 --- join: hkBst (n=hkBst@gentoo/developer/hkbst) joined #haskell 02:20:19 gour: should be fine. it will work line by line 02:20:33 gour: the weak link will be sqlite 02:20:37 (if there is a problem) 02:20:41 not the file import 02:20:50 quicksilver: why sqlite? 02:20:50 I have successfully applied haskell to analyzing 2GB ascii files - the end result was both faster and more readable than the equivalent perl script 02:21:06 olsner: good to know. thanks 02:21:13 gour: well I don't know about sqlite's performance characteristics when you instert 750M of data 02:21:20 gour: so I can't vouch for that 02:21:24 but getting it there required a few tricks... for example, going with ByteString from the start is a good idea 02:21:37 but I *do* know that line by line processing in haskell doesn't require memory 02:21:56 I also know that lazy IO is extremely evil and you should just call hGetLine :) 02:22:00 quicksilver: well, i don't need all the 'fields' from CSV database, so the sqlite3 database will be probably much smaller 02:24:20 --- quit: prb () 02:24:37 lol 02:25:06 quicksilver: I use HSQL-MySQL lib of haskell 02:25:07 --- quit: mrsolo (Connection timed out) 02:25:30 these days I rather use ruby instead of perl for text processing unless I need some perl module (like for xml) -- ruby is not that much slower and it's easier to write programs in 02:26:41 --- quit: RogerTaylor () 02:26:50 --- join: swiert (n=wss@sneezy.cs.nott.ac.uk) joined #haskell 02:26:55 I would argue that haskell is faster then ruby and even easier to write programs in :) 02:27:08 --- quit: c9s_ ("leaving") 02:27:10 I think that ruby is mostly a clean perl5 with batteries included. More useful modules in the core 02:27:12 ruby seems to be a poor middle ground: slower AND less powerful :) 02:27:16 b_jonas: I must play with ruby some time... but Moose is quite cute and gives you some of that goodness (but with CPAN) 02:27:31 It's not relaly clean, it just has dirt in different places than perl5 02:27:43 --- quit: slipstream (Read error: 110 (Connection timed out)) 02:27:43 but yes, the core libraries are indeed more useful 02:27:50 i have to prepare that sqlite3 database for use with another gui program which will be in haskell, so ruby & co. are not alternatives ;) 02:27:58 it's not less powerfool I think 02:28:08 the syntax of reference in Perl 5 is awful 02:28:21 well speaking of power in programming languages is always a bit woolly 02:28:28 since in a technical sense, they are all the same power 02:28:31 obviously 02:28:35 as for haskell, I guess that it must be fine but the learning curve is difficult 02:28:44 when I say 'ruby is less powerful than haskell' what I'm really talking about is abstraction 02:28:48 and expressivity of abstractions 02:29:00 oh, you mean less powerful than haskell 02:29:02 I can agree that 02:29:03 yes 02:29:07 I thought you compared to perl 02:29:10 no 02:29:14 like cognominal_ 02:29:16 sorry 02:29:18 the perl comparison was the 'slower' part 02:29:29 I've done measuremens about their speed 02:29:30 I was saying ruby is slower than perl, but less powerful than haskell :) 02:29:43 which makes it a poor middle ground IMO 02:29:46 it is cute though 02:29:53 and I have no objections to it on cuteness grounds 02:30:03 --- join: tommyd (n=tkeller@85.232.10.50) joined #haskell 02:30:22 that sentence could mean that ruby is slower than perl but faster than haskell and less powerful than haskell but more powerful than perl 02:30:46 I mostly use ruby as a perl replacement with ruby threads really 02:30:52 it doesn't cover much new ground after perl 02:31:10 go python 02:31:20 arf 02:31:25 Sometimes I wish I could say 'list.sort_by { rand } 02:31:31 Whoops, hit enter. 02:31:37 anyway, the result I've got is that ruby wasn't that much slower than perl as I've expected it to be 02:31:43 In Haskell, that is. But it's a small price to pay. :) 02:31:54 * roconnor wonders what units power is measured in. 02:32:02 dolio: I have an obfuscation in Perl that does shuffle with sort thru rand 02:32:10 (why, you could just have a specific shuffle function or something instead of sort_by) 02:32:18 haskell is young.. soon enough it will be 2x as fast 02:32:19 i'm sure 02:32:29 young? 02:32:30 cognominal_: worse, I have a benchmark meditation in perl that compares shuffles 02:32:40 okay. ghc is relatively new 02:32:50 compared to other compilers 02:32:58 b_jonas: Yeah, but shuffling requires a monad for random numbers in Haskell, so it feels comparatively heavyweight. 02:33:07 --- join: Sizur (n=eugene@89.132.150.41) joined #haskell 02:34:10 sub shuffle { map { @_=sort { int (3*rand)-1 } @_ } 0..5 } 02:34:42 Yikes. :) 02:35:34 --- join: twanvl_ (n=twanvl@ip5451f037.direct-adsl.nl) joined #haskell 02:35:53 So, 'int (3*rand)' is something in {0, 1, 2} ? 02:35:57 --- join: zx]treads (n=root@83.212.84.69) joined #haskell 02:36:08 yup 02:36:16 Lovely. :) 02:36:16 b_jonas: url? 02:37:02 You can't do that with every kind of sort function. Some don't like very much rand as a sort criteria 02:37:44 we use a fisher-yates shuffle function in perl 02:38:49 --- join: b_jonas_ (n=b_jonas@dsl51B6190D.pool.t-online.hu) joined #haskell 02:40:40 --- join: LeCamarade (n=revence@217.113.73.39) joined #haskell 02:41:37 dolio: http://www.perlmonks.com/?node_id=533396 is my benchmark of shuffles in perl, it has sory_by rand variants as well 02:41:47 Title: Benchmarking perfect shuffles 02:42:01 cognomial: yeah, that one is a famously wrong shuffle in perl 02:42:20 another one is sort { rand <=> rand } 02:42:53 apart from not giving uniform probability of outputs, these can also segfault older perls that assume the compare function is transitive 02:43:07 at least so I've heared, I've never actually seen the segfault 02:43:26 dolio: could you show me your obfu with sort though? I think it could be interesting 02:43:38 Eh? 02:43:39 It could complement something like my perlmonks.com/?node=fast+japh 02:43:51 'dolio: I have an obfuscation in Perl that does shuffle with sort thru rand' 02:43:55 that one 02:44:05 b_jonas, I see that you like pentominos. Have you seen my obfu that won the tpj contest in its category : solving pentominoes with regexen? 02:44:07 That was cognominal_ 02:44:17 Presumably, the one he posted. 02:44:57 oh sorry 02:45:07 'sub shuffle { map { @_=sort { int (3*rand)-1 } @_ } 0..5 }' 02:45:20 heu it's wrong 02:45:54 --- join: obsethryl (n=lab@unaffiliated/obsethryl) joined #haskell 02:46:19 I am very proud of the tpj writeup http://www.foo.be/docs/tpj/issues/vol3_3/tpj0303-0015.html 02:46:21 Title: The 3rd Annual Obfuscated Perl Contest Victors - The Perl Journal, Fall 1999 02:46:39 First Place. Stephane Payrard's Polyominos - fitting problem solver is very beautiful - eclipsed only by the entertaining reading available in the SOLUTION file. Perhaps it's fitting that such a powerful obfuscatory statement comes from the land that bred semiotics, deconstructionism and Jean-Paul Sartre. Or maybe Stephane is just demented. 02:46:39 I was just pining for the simplicity of 'list.sort_by { rand }' from Ruby, which is commonly used to shuffle (since people probably don't care whether it's perfect or not). 02:46:48 Um ... how do I turn a, say, Char to Integer? 02:46:59 What func? 02:47:08 read maybe? 02:47:13 ord? 02:47:32 That'll give you an Int that you can turn into an Integer easily enough. 02:47:34 fromEnum ? 02:47:49 Oh. How could I forget ord????????? 02:47:58 > (fromIntegral . fromEnum) 'b' 02:47:59 98 02:48:00 fromEnum would be direct, yeah. 02:48:08 :t fromEnum 02:48:10 forall a. (Enum a) => a -> Int 02:48:16 Oh, no it wouldn't. 02:48:36 is that the polymyno puzzle quine one? 02:48:43 I think ord and chr were added to make BASIC programmers feel at home with Haskell. 02:48:54 aaim pasted "Here's another one" at http://hpaste.org/2745 02:49:07 --- quit: njbartlett () 02:49:10 dolio: that one is perfect 02:49:19 :o) 02:49:46 int-e: pascal programmers 02:49:53 basic has ASC and CHR$ not "ord" 02:50:06 b_jonas: I dont know the polyomino quine but I don't read perlmonks much anymore 02:50:15 it's not on perlmonks I think 02:50:25 --- join: SamB_XP (n=Sam@216-15-104-38.c3-0.upd-ubr7.trpr-upd.pa.cable.rcn.com) joined #haskell 02:50:26 b_jonas_: but it's funnier with basic ;) 02:50:27 I think it might not even be in perl 02:50:38 aaim annotated "Look, it does what I want, and it's easy... but isn't this built-in somewhere?" with "(no title)" at http://hpaste.org/2718#a1 02:50:47 --- quit: b_jonas (Read error: 110 (Connection timed out)) 02:51:00 10 REM DON'T LAUGH AT BASIC, OK? 02:51:00 int-e: well, perl has lots of ugly syntax originating from other languages (including sh, tchs, sed, awk) 02:51:03 b_jonas_: Is it? It's been a while since I've used Ruby. 02:51:36 LeCamarade: I'd never. I'm too young to die. 02:51:43 dolio: yes, because sort_by uses the block to calculate keys (weights) for each element and sorts "by" comparing those weights 02:51:55 the element with lower weight gets to be the first 02:52:00 :o) 02:52:24 --- join: ejt (n=ejt@85-211-14-250.dyn.gotadsl.co.uk) joined #haskell 02:52:26 the ruby method "sort" is the one that accepts a comparision function like perl's sort or c++'s sort or c's qsort 02:53:44 LeCamarade: I don't. I laugh at visual basic and esp openofficeorg basic all the stupid new oo basics, but I like the non-oo ones like old zx spectrum basic, commodore basic, gwbasic, qbasic, word2basic, corelbasic etc 02:54:07 after all, basic has one of the funniest quines in existence 02:54:09 10 LIST 02:54:14 eeek, spectrum basic with its lack of functions etc. 02:54:24 As I recall Oleg has an argument that attaching random tags to elements and sorting doesn't result in a perfect shuffle... 02:54:33 I liked bbc basic, that was a real eye opener about the relative power of programming languages... 02:54:36 functions 02:54:39 who needs functions 02:54:45 --- join: c9s (n=c9s@61-223-103-147.dynamic.hinet.net) joined #haskell 02:54:46 gosub is enough 02:55:02 dolio: it depends what kind of tags you attach 02:55:30 the Cormen book discusses that algorithm (also F-Y shuffle) 02:55:53 --- join: RogerTaylor (n=rogertay@123.20.194.118) joined #haskell 02:56:09 if there are no equal tags and they are independent with the same distrubution then it's perfect 02:56:20 (provided they are ordered completely) 02:59:22 --- join: MyCatSchemes (n=rb6822@sevkil.cs.bris.ac.uk) joined #haskell 02:59:25 I wrote buublesort in basic once 03:00:03 b_jonas_: we don't care about the O(n^2), but how *dare* you use such an evil algorithm? :) 03:01:20 --- quit: dibblego ("Leaving") 03:01:59 --- join: gip (n=gpi@SAGW-PRIMARY.ARC.COM) joined #haskell 03:02:09 MyCatSchemes: well, that was ages ago 03:02:15 I didn't know about heapsort at that time 03:02:30 and quicksort or the like is difficult to write in basic 03:02:38 --- join: JaffaCak1 (i=simonmar@nat/microsoft/x-3edf73c69ca2bf7f) joined #haskell 03:02:59 > let bubble (a:b:xs) | a > b = b:bubble (a:xs) | otherwise = a:bubble (b:xs); bubble xs = xs; sort xs = last . zipWith (flip const) xs . iterate bubble $ xs in sort [3,1,2,4,561,0,3,4] 03:03:01 [0,1,2,3,3,4,4,561] 03:03:07 otoh it takes very little code 03:03:35 looks perfectly reasonable to me, what's evil about that? ;) 03:03:39 and if I'd have memoized the result (I only used to to sort font names for a dialog box) then code size would definitely be the primary concern 03:04:10 Oh wait, I'm stupid 03:04:17 I wrote quicksort in basic another time as well 03:04:25 and in that case, speed was a concern 03:04:31 now it was bad in that case 03:04:33 but I was young 03:05:05 --- nick: b_jonas_ -> b_jonas 03:05:42 --- quit: MarcWeber (Remote closed the connection) 03:05:57 (anonymous) annotated "Look, it does what I want, and it's easy... but isn't this built-in somewhere?" with "(no title)" at http://hpaste.org/2718#a2 03:06:55 int-e: zipWith const to trim one list down to the size of another is an interesting indiom 03:09:02 :) 03:09:50 oh. but it means that the function doesn't work for empty lists. 03:10:07 that wouldn't have happened with the more obvious (!! length xs) 03:10:39 --- join: ivanm_ (n=Ivan@60-242-0-245.static.tpgi.com.au) joined #haskell 03:12:07 why wouldn't it work with an empty list? 03:12:20 > last [] 03:12:32 Exception: Prelude.last: empty list 03:13:20 but quicksilver said zipWith const 03:13:22 "the function" was the sort function above. The zipWith works just fine. 03:13:42 oh, I see 03:13:44 the bubble 03:14:58 bubblesort! 03:15:04 the king of sort routines! 03:15:09 --- join: Igloo_ (n=igloo@chaos.earth.li) joined #haskell 03:19:16 --- join: bvd (n=user@n027182.science.ru.nl) joined #haskell 03:19:39 bubblesort works locally, is therefore friendly to the cache and it is nicely parallisable. What else do you want? 03:20:27 --- quit: JaffaCake (Read error: 110 (Connection timed out)) 03:20:32 well, mergesort has those advantages as well 03:21:03 b_jonas: mergesort isn't truly local 03:21:06 oh, is a parallelised bubblesort efficient? 03:21:07 but I somehow like heapsort better, even though I know it's not good for external sort 03:21:14 b_jonas: it works back and forth between two copies 03:21:28 --- join: njbartlett (n=njbartle@cpc2-barn4-0-0-cust110.brnt.cable.ntl.com) joined #haskell 03:21:28 Mergesort requires twice as much RAM as there is data, you can't do it in-place. 03:21:34 b_jonas: and it maintains at all times two 'cursors' into the source list 03:21:44 b_jonas: so it's hitting three active memory locations (two read, one write) at a time 03:21:46 --- join: pedro_dgv (n=pedro@42.Red-217-125-2.staticIP.rima-tde.net) joined #haskell 03:21:50 which can be pretty widely distributed 03:21:56 quicksilver: it's not local if you have one tape, sure, but it's still cache-local or can be done externally with a few tapes 03:22:04 --- quit: boegel (Read error: 110 (Connection timed out)) 03:22:08 are you programming one-tape turing-machines or what? 03:22:21 no, we're programming on modern CPUs 03:22:30 but we're sorting lists so big they don't fit in the cache 03:22:34 at least, I assume that's the point :) 03:22:34 yep 03:22:46 but the cache can do two locations pretty fine 03:22:57 Modern CPUs have a megabyte of cache apiece. How often do you sort lists bigger than, say, 128x2^10 items? 03:22:57 (it's two not three but that's a minor point) 03:23:31 b_jonas: don't forget to count the stack for all those nifty recursive calls. 03:23:53 MyCatSchemes: recursive calls? you're thinking of qsort. mergesort doesn't have those. 03:24:22 non-inplace is true 03:24:25 you're right 03:24:46 I like heapsort which is inplace, but it's obviously really bad for localty 03:24:54 the traditional encoding of mergesort is recursive 03:25:17 of course it's possible to unwrap it with an extra variable for 'current sublist size' 03:25:46 MyCatSchemes: if you're not sorting something bigger than the cache, then you probably don't care about performance 03:25:50 MyCatSchemes: it will be fast anyway :) 03:27:00 --- quit: Igloo (Read error: 110 (Connection timed out)) 03:29:21 --- join: ptx (n=ptxmac@0x5550c853.adsl.cybercity.dk) joined #haskell 03:30:27 quicksilver: not with bubblesort :) 03:30:34 quicksilver: nor with bogosort. 03:31:35 I believe bubblesort is actually the sort of choice when sorting nearly-sorted data ? 03:33:22 osfameron: only if it's already almost-sorted to some ridiculous extent. And mergesort also tends towards O(n) in nearly-sorted data, too. Methinks insertion sort does as well, and that's always faster than bubble except under really contrived circumstances anyway. 03:35:24 yeah, I can imagine that 03:37:21 MyCatSchemes: merge sort generally has a really bad constant 03:37:45 MyCatSchemes: it's a very SLOW O(n log n), and it's also a SLOW O(n) on nearly-sorted data 03:37:49 but on the other hand it's stable 03:37:56 which is important for lots of applications 03:38:06 --- quit: dolio (Read error: 104 (Connection reset by peer)) 03:38:11 --- quit: njbartlett (Read error: 104 (Connection reset by peer)) 03:38:11 --- quit: petekaz (Connection timed out) 03:38:12 --- join: dolio (n=dolio@nr5-216-196-210-207.fuse.net) joined #haskell 03:38:23 insertion is fastest on nearly-sorted, isn't it? 03:38:49 and if you squint a little bit, you can see insertion on nearly-sorted data as a kind of "obvious" improvement to bubble-sort 03:39:00 "skip out of the bubble if this bit's already in the right place" 03:40:24 --- join: matt__r (n=matt__r@CPE-121-209-187-24.nsw.bigpond.net.au) joined #haskell 03:42:16 --- nick: ivanm_ -> ivan_m 03:43:04 --- nick: twanvl_ -> twanvl 03:45:57 --- quit: goalieca ("(sleepy time)") 03:48:08 --- nick: ivan_m -> ivanm 03:50:02