00:00:00 --- log: started haskell/02.01.29 00:06:11 * yusri is off to work 00:06:20 tks, ski 00:06:25 ok, bye yusri ! 00:17:34 * Heffalump wonders if ski ever sleeps 00:18:41 hmm, i'm think i should go home and sleep in some minutes, just reading a little about refactoring and adaptive and aspect programming (monads and aspects are related, right ?) 00:19:47 they are? 00:25:01 well, with monads one can take certain things scattered throughout code (such as explicit state passing) and define it in one place, and then use the monad operations in standard monad way. i you discover that you need more things, perhaps exceptions or a name supply or nondeterminism, just change the monad definition and the "primitive" monad operations (add the new ones) and just change the places in the code using the monad that conceptually n 00:25:32 instead of having to change all the code that explicitely does whatever the monad should do 00:25:40 I guess 00:26:03 but aspect-oriented programming tries to be a lot more general than that, I think 00:26:22 but, i don't know very much about aspect-oriented programming. so this is just a feeling i've got of a similarity. 00:27:34 yes, i suppose it's more general. but the point is that monads (or arrows, etc , other combinator libraries) can be used to accomplish certain AOP tasks, you know what i mean ? 00:27:41 yeah 00:27:46 ok 00:30:56 but, now i'm really tired so i'll give up for today and go home to sleep 00:31:11 * ski says bye ! 00:31:35 --- quit: ski ("zzzzz") 00:39:28 hmm, is anyone else using the SOE sources with Hugs98 for windows? 01:26:45 --- join: Yurik (~yrashk@gw.telcos.net.ua) joined #haskell 01:27:45 Yurik: do you use Hugs on Windows? 01:28:02 discobob nope, I use GHC 01:28:38 dang 01:31:53 --- join: jewel (~jleuner@57.66.12.99) joined #haskell 01:32:23 --- quit: discobob (Read error: 104 (Connection reset by peer)) 02:23:09 python regex isn't anymore tedious than perl/ruby's, it's just not syntactically sugared. 02:24:03 (and note that ruby doesn't even have first class functions, and thus no closures) 03:01:21 --- join: shapr (~user@p-c2fbab34.easy.inet.fi) joined #haskell 03:09:39 shapr hi 03:09:50 hi Yurik, what's up? 03:10:31 shapr fine, thanks :-) 03:14:07 have you had any further thoughts on the A/ES server? 03:17:55 E/AS :-) and not server :-))) But I have, of course. We're currently resolving some issue to go ahead :-) 03:17:55 If you interested, you may track our convesation at #eas, on this irc net. 03:17:55 s/issue/issues/ 03:18:20 E/AS? 03:19:07 E/AS stands for Enterprise / Automation System, an opensource project to create framework for enterprise automation needs 03:19:58 erm, right 04:59:45 guys, what is Concurrent Haskell? 05:00:13 Haskell with primitives (in the IO monad) to support threads and signalling between them 05:00:16 [I think] 05:00:44 I thought Concurrent was what we use, and Paralell was PVM support added? 05:01:03 I don't think Concurrent stuff is Haskell '98, though I could be wrong 05:01:27 as in it's not in the Haskell '98 libraries, not that it doesn't satisfy the language spec 05:01:43 but all this is my impressions, not the gospel truth 05:04:01 I see concurrent lib in GHC, does it conforms to Concurrent Haskell now? 05:04:09 I'm sure GHC does 05:04:18 not sure about other Haskell compilers/interpreters, that's all 05:04:18 Heffalump: you mentioned that functions over type classes use dictionaries, right? 05:04:24 jemfinch: yes 05:04:40 Heffalump: how are those dictionaries implemented? I haven't seen any O(1) hashtable implementations in Haskell. 05:04:42 (or, to be precise, what shapr said: that functions which are polymorphic in a type class constrained variable do) 05:04:59 jemfinch: no idea 05:05:03 read the GHC source :-) 05:05:19 eh? what I said? 05:05:38 does Haskell provide mutable arrays? 05:05:40 shapr: I think he meant ski :) 05:05:50 oh 05:05:52 whew 05:06:05 oops, sorry, I did :-) 05:06:17 jemfinch: GHC does, can't remember if it's in the standard libraries 05:07:36 Heffalump: does the mutation have to be put in a monad? 05:08:16 jemfinch: yep, either the IO monad or (possibly) the ST monad 05:08:22 the ST monad? 05:08:26 State 05:08:30 ah, ok. 05:08:34 * jemfinch doesn't know Haskell, actually. 05:08:51 GHC has a built-in one with some nice implementation characteristics (but if you use it your code won't be Haskell 98) 05:12:27 * jemfinch might have to look into monads in the future. 05:12:37 monads are nifty 05:12:39 Haskell is interesting, but I'm a bit leery of the lazy evaluation. 05:12:53 lazy evaluation rocks! 05:13:08 plus you can always force strict evaluation with seq 05:13:26 shapr: why does it rock? 05:15:13 at the simplest end of life, it lets you used stuff like [1..] 05:15:27 so you can make an infinite sequence that's a recursively built list 05:15:36 and the list is only calculated as you need it 05:17:37 You don't have to worry about structuring your code to support functions that generate data intermittently. 05:18:17 Logan: explain that :) 05:18:58 Well, under lazy evaluation, all you have to do is write a function that generates a list of answers... 05:19:14 That generation will only occur as needed, but you don't have to worry about dealing with that explicitly. 05:19:28 Whereas in a strict language, you have to support that sort of data access explicitly if you want it. 05:19:33 yah, in python terms, *everything* is xreadlines and xmap 05:21:20 In other words, in a strict language, you'd have to explicitly code the overhead of maintaining state, so you know what the _next_ data you should generate would be. 05:22:16 I think it also makes it so that the order of evaluation matters not at all (does order matter in all strict languages, even if they're functional?). 05:30:26 If sideeffects are not possible then the order only affects runtime and termination 05:52:38 btw, guys is there the good way to read/write binary files? (ghc) 05:54:36 readFile and writeFile? 05:55:08 I think that some functions take an option to specify it's a binary file but you'd have to check the report for that 05:57:05 it's in prelude? 05:57:23 Think so 06:05:28 I thought binary files are a known problem 06:07:53 IWW? 06:08:29 --- nick: Logan -> LoganClass 06:08:47 IWW? 06:08:53 oh, in what way 06:08:59 errm. I've seen stuff on the mailing list. ICBW. 06:09:14 I thought I saw something about making it easy to read bits at a time (as opposed to bytes). 06:09:17 inter continental ballistic whale? 06:09:21 I doubt there's a problem with just reading raw bytes. 06:09:21 ICBW? 06:09:33 It could be worse. 06:09:38 I could be wrong. 06:09:48 I can't breathe water. 06:09:50 Later. 06:09:52 :-) 06:10:09 someone should teach clog to do explanations 06:10:37 well, my irc client has some cool tricks that can help 06:11:10 currently, EmacsWiki:EmacsNiftyTricks is a valid url, and I can hit enter on it, and it calls browse-url.el, and drops my browser to the correct page on the wiki 06:11:25 the other possibility is that purl.org/wiki/haskell/ will get approved by the guy I submitted it to 06:11:32 and then we can use it as a valid web address 06:16:46 * Heffalump reads about purls 06:17:03 given how cheap domains are now, using them for http redirects seems unnecessary 06:17:19 cos you can get a domain and be pretty certain you'll be able to keep it for ever 06:25:43 what exactly readFile and writeFile does? 06:28:13 seems that found 06:29:08 "All I/O functions defined here are character oriented" 06:29:10 :-( 06:33:50 Yurik: you might want: http://www.haskell.org/ghc/docs/latest/set/sec-ioexts.html#AEN12992 06:36:16 dennisb thanks. when I can read more about it (examples, for example)? I'm a newbie in Haskell.. 06:37:52 okay, well most files are character oriented, what is the problem? 06:38:17 dennisb I need to work with binary ones 06:40:32 well, in unix it's probably the same otherwise the function openFileEx is what you need instead of openFile that are in the haskell 98 standard 06:42:19 so, I should use openFileEx to open file.. right? and what should I to read/write/seek? 06:42:24 as far as I know it's only using win32 where it makes a difference between binary or textmode on files (in the old days I guess it was more systems) 06:42:43 http://www.haskell.org/onlinelibrary/io.html 06:42:49 (normal IO functions) 06:43:32 hPutChar, hGetChar, hSeek 06:44:52 will hPutChat and hGetChar work with byte-values? 06:48:16 moreover, there is no way to put and get numbers? 06:50:11 (for example) 06:51:52 I dont think so, you have to convert 06:52:56 but it is easy to write a small function that reads 4 bytes and constructs a number or something 06:53:15 but it will be probably very slow, eh? 06:54:35 well, there is functions hGetBuf, but I don't think it worth it 06:54:56 and slow, it is going to be bufferd input anyway, so it's not going to be that slow 06:55:20 yes, I found hGetBufBA 06:55:29 may be it is better than hGetBuf? 06:56:00 I don't know. 06:56:06 is hPutBuf operates on any kind of data? 06:56:12 I mean numeric at least 06:57:03 No, if you want to convert to an Integer in haskell you still need to convert. You could use other types, like 32 bit words or something 06:57:49 Integer in haskell is arbitrary big numbers, it's not going to be a 1 to 1 map against the value in a file anyway 06:58:00 Int32? 06:58:37 and then you have all the issues with little and big endian data, if you ask me, just do a simple solution first and see if it is fast enough 06:59:08 btw, ghc says ghc: can't find module `IOExts'. ghc-pkg -l says that lang package is installed. What might be the problem? 06:59:47 -package lang 06:59:59 thanks :-) 07:00:05 -package plain-brown 07:00:33 what's in that then? :-) 07:00:45 none of your business ;) 07:00:48 * shapr is joking 07:01:06 it at least implies that it holds some kind of vice 07:03:51 :-) 07:04:30 so how should I use hPutBuf to write Int32? 07:05:05 Yurik: I don't know. I would just read 4 characters and put together 07:05:22 he wants to write... 07:05:25 just use show 07:05:29 But i'm sure it can be done in some way with... 07:05:30 aha 07:05:50 but he wants to write it binary? show converts to a ascii-string 07:05:56 oh, sorry 07:06:12 split it up into bytes and use chr, then? 07:06:40 that is what I would do, if it is to slow I would start looking for faster solutions 07:07:49 hmm. 07:07:51 IOExts> :t hPutBuf 07:07:51 forall a. 07:07:51 PrelIOBase.Handle 07:07:51 -> PrelPtr.Ptr a -> PrelBase.Int -> PrelIOBase.IO PrelBase.Int 07:07:55 it is still going to be buffered so you don't send it one byte at a time to disk anyway 07:07:56 hPutBuf fh (300 :: Int32) 0 07:08:07 Main.hs:9: 07:08:08 Couldn't match `Int32' against `PrelPtr.Ptr a' 07:08:09 Expected type: Int32 07:08:10 Inferred type: PrelPtr.Ptr a 07:08:11 In an expression with a type signature: 300 :: Int32 07:08:12 In the second argument of `hPutBuf', namely `(300 :: Int32)' 07:08:15 * dennisb have never used hPutBuf 07:08:24 Where can be my mistake? 07:08:28 :) 07:08:33 you need to create a PrelPtr.Ptr 07:08:36 I'm not sure how to do that 07:11:03 me neitherm and for just a small integer it's probably not that much faster (if any) 07:12:01 that much faster than using hPutChr? 07:12:25 I think the idea with these functions is if you need to write ab image or something that you have in memory to a file. 07:12:26 ops 07:13:12 yes, I think hPutChar is probably going to be fast 07:14:01 ISTR seeing PackedString mentioned for efficient IO, but I don't know what functions exist for manipulating them 07:14:46 well, if use hPutChar, how can I convert Int32 to list of chars? 07:15:29 Igloo: there is a lot of functions in http://www.haskell.org/ghc/docs/latest/set/packedstring.html 07:16:25 Yurik: some div and mod perhaps 07:18:26 I would write two small function readInt32 :: IO Int32 and writeInt :: Int32 -> IO () 07:18:33 writeInt32 even 07:19:03 Is it Int32 or Word32? 07:19:21 Int32, probably 07:20:14 well and I forgot the file handles :-) 07:23:21 an alternative to the div and mod is to use the Bits module and do some shifts and ands. 07:31:49 btw, may be the best way will be write this part of code in C? 07:32:26 Yurik: no, I think it is easy to the conversion 07:32:43 should fit on a single line i think 07:32:53 dennisb but in speed may be C will be faster... 07:33:11 maybe 07:33:35 I don't know what you are trying to do so what do I know about that 07:33:58 I plan to develop specialized database (non relational) 07:34:32 specialized to store trees 07:44:13 --- join: Heffalum1 (ganesh@163.1.125.98) joined #haskell 07:44:27 cool, two Heffalumps 07:45:29 Yurik: to do really fast databases you might want to use C for the most lowlevel parts 07:45:48 dennisb yes, I think so, too... 07:45:50 sorry, the old one is on my old IP address :-) 07:45:53 which I can't use any more 07:46:11 dennisb so initially I want (probably) use C for I/O 07:46:25 Yurik: I converted an Word32 to a 4 byte string that can be outputed, and it became: 07:46:28 f :: Word32 -> [Char] 07:46:28 f n = [b 24 n, b 16 n, b 8 n, b 0 n] 07:46:28 where b i n = chr $ fromEnum $ (n `shiftR` i) .&. 255 07:46:56 Or I wrote a function that convert 07:46:59 thanks :-) 07:47:16 but there is overhead, and even if the disks are slow it is still overhead 07:47:47 --- quit: Heffalump (Read error: 60 (Operation timed out)) 07:48:40 is there any good tutorial of writing C libs for Haskell? 07:49:12 Not really, there is examples of how you call C from haskell, that is usually not so hard 07:49:23 calling haskell from C is harder 07:49:42 but we are short on good tutorials 07:50:48 but normal C-functions with simple types as input and output is just a matter of telling the compiler that they exist and what type 07:52:04 * dennisb thinks it is fun, but have to leave for a while 07:52:43 * shapr waves 08:12:45 --- nick: Heffalum1 -> Heffalump 08:33:45 --- join: Igloo_ (~igloo@c93.keble.ox.ac.uk) joined #haskell 08:49:15 --- quit: Igloo (Read error: 110 (Connection timed out)) 08:58:56 --- quit: Yurik ("reboot") 09:08:24 --- join: Yurik (~yrashk@gw.telcos.net.ua) joined #haskell 09:08:37 re 09:09:36 re Yurik! 09:09:44 re 09:28:00 --- quit: dennisb (carter.openprojects.net irc.openprojects.net) 09:28:00 --- quit: Yurik (Read error: 104 (Connection reset by peer)) 09:29:27 --- quit: Heffalump ("Changing server") 09:29:28 --- join: Gowilla (ganesh@munchkin.comlab.ox.ac.uk) joined #haskell 09:29:48 --- nick: Gowilla -> Heffalump 09:29:56 * Heffalump reappears with fixed DNS 09:30:08 --- join: dennisb (~dennis@as2-4-3.an.g.bonet.se) joined #haskell 09:49:05 --- nick: shapr -> shae 09:49:38 --- nick: shae -> hmm 09:50:21 --- nick: hmm -> whatthenoo 09:51:02 --- nick: whatthenoo -> ohreally 09:54:14 --- nick: ohreally -> capr 09:57:23 --- quit: Heffalump (carter.openprojects.net irc.openprojects.net) 09:57:23 --- quit: Igloo_ (carter.openprojects.net irc.openprojects.net) 09:58:38 --- join: Yurik (~yrashk@gw.telcos.net.ua) joined #haskell 09:59:06 --- join: Heffalump (ganesh@munchkin.comlab.ox.ac.uk) joined #haskell 09:59:06 --- join: Igloo_ (~igloo@c93.keble.ox.ac.uk) joined #haskell 10:05:45 --- quit: capr ("brb") 10:08:53 --- join: shapr (~user@p-c2fbab34.easy.inet.fi) joined #haskell 10:17:46 --- nick: LoganClass -> Logan 10:17:52 Bleh, forgot to open my window before I left. 10:31:20 --- quit: shapr ("shooting people") 10:57:55 What would be a good, concise way to specify a permutation function in haskell? 10:58:02 I have one, but I think it's more verbose than it needs to be. :P 10:58:30 specify, or implement? 10:59:12 from memory, something like 10:59:20 inserts :: a -> [a] -> [[a]] 10:59:24 inserts a [] = [a] 11:00:01 inserts a bs@(b:bs') = (a:bs):inserts a bs' 11:00:15 perms :: [a] -> [[a]] 11:00:19 perms [] = [[]] 11:00:26 perms (a:as) = insert a (perms as) 11:00:54 I think this works: p [] = [[]]; p (x:xs) = concat [zipWith (\a b -> a ++ [x] ++ b) (inits ys) (tails ys) | ys <- p xs] 11:01:41 it ought to. I think mine is nicer :-p 11:01:44 Heffalump: I think I did it like you. 11:01:55 Min is less verbose, though :-) 11:01:59 true 11:02:03 Yes, Igloo_ answers my question. :P 11:02:09 hmph. 11:02:11 * Heffalump sulks 11:02:16 permute :: [a] -> [[a]] 11:02:17 permute [] = [] 11:02:17 permute [a] = [[a]] 11:02:17 permute (x:xs') = permute' [] x xs' 11:02:17 permute' :: [a] -> a -> [a] -> [[a]] 11:02:19 permute' before x [] = map (x:) $ permute before 11:02:22 permute' before x after@(a:after') 11:02:23 = 11:02:26 (map (x:) $ permute (before ++ after)) ++ permute' (before ++ [x]) a after' 11:02:48 It has a certain elegance, anyway. although (\a b -> a ++ [x] ++ b) is a bit clumsy 11:03:08 I was just curious, noticed in my programming languages textbook today (which has a few pages devoted to haskell) a statement about using list comprehensions to generate permutations (among other things). 11:03:30 And was wondering if there were some elegant, simple method that I wasn't thinking of. 11:03:49 Now that I understand Igloo_'s method, I sorta like it. :P 11:04:51 --- nick: Igloo_ -> Igloo 11:05:03 Except mine gives the permutations in lexicographical order (if the order of the initial list defines the ordering of the elements). 11:09:21 How about a convex hull algorithm? :P 11:25:49 I keep forgetting how precedence goes in haskell... 11:25:59 Is f a + b equivalent to f (a + b) or (f a) + b ? 11:29:37 (f a) + b 11:31:17 That's what I thought was more likely, but I wasn't sure. 11:31:45 Does anything have higher precedence than function application? 11:32:06 no 11:32:10 well, not that I can think of 11:32:27 Parentheses 11:32:30 That should save some parentheses from my code. *grin* 11:32:38 Heh, I don't typically think of parentheses as operators. 11:32:50 they aren't 11:33:02 not according to any sensible definition of operator, anyway 12:03:38 Here's a good question... 12:03:44 I have a list of points. 12:04:03 I'd like to do something sorta like a fold that just sums up the length of the distances between each adjacent pair of points in the list. 12:04:18 Any concise way of specifying this? 12:10:56 * Logan wonders what is wrong with this: 12:10:57 dist :: Point -> Point -> Double 12:10:57 dist = sqrt . dist2 12:11:54 how about sum (zipWith dist points (tl points)) 12:12:37 tl? 12:13:10 umm, tail, sorry 12:14:00 Seems like it might work. I didn't think of it as a zipping problem at all. 12:14:05 How about my function composition problem there? 12:14:18 dist2 :: Point -> Point -> Double 12:14:23 Why can't I define dist as sqrt . dist2 ? 12:15:00 wrong number of arguments 12:15:24 . is for functions that take one argument 12:15:48 Didn't realize the number of arguments mattered at all. 12:15:55 look at the definition of . 12:15:58 it has to matter 12:16:07 otherwise the effect of . would be ambiguous 12:16:20 I suppose it could be. 12:17:34 not could, would 12:24:45 I suck at coding geometric stuff. :P 12:59:43 My convex hull algorithm finally works! 13:00:00 That's probably the most interesting algorithm I've implemented in haskell so far. :P 13:00:14 cool :-) 13:00:48 And I think it's reasonably efficient too. 14:16:13 anybody alive? 14:16:37 Yep. 14:17:00 How can I map using function that return another than list element type? 14:17:15 function is hPutChar 14:17:27 You want to apply hPutChar to every element in a list? 14:17:43 yep 14:17:44 That's a good question, because I was wondering that myself. :P 14:17:56 or I should use hPutStr? 14:18:11 That would solve the problem but not answer the more general question. :P 14:18:38 yeah, it should solve, but the problem is interesting for me :-) 14:19:09 I've been running into it several times, processing elements of lists (that aren't just Char lists) for output. 14:22:24 sequence_ it afterwards 14:25:24 does list comprehension function should return elements of the type exact to source list element's type? 14:26:16 No, it returns the type of the left of the | 14:26:50 f :: Word32 -> [Char] 14:27:01 hPutStr fh map(f)[1..32323] 14:27:13 gives me: 14:27:14 Couldn't match `IO ()' against `t -> t1' 14:27:14 Probable cause: `hPutStr' is applied to too many arguments in the call 14:27:16 In a 'do' expression pattern binding: 14:28:25 I changes hPutStr string to 14:28:34 hPutStr fh (map(f)[1..32323]) 14:28:40 but anyway it gives an error 14:28:50 Couldn't match `Char' against `[Char]' 14:40:23 * Yurik is out to smoke 14:45:35 * Yurik is back 14:49:17 hPutStr fh map(f)[1..32323] does not do what you want. 14:49:25 You want hPutStr fh (map(f)[1..32323]) 14:49:35 Or hPutStr fh $ map(f)[1..32323] for short. 14:49:54 hPutStr fh (map(f)[1..32323]) doesn't work, too 14:50:03 What error does it give? 14:50:19 Oh. 14:50:27 The result of the map is [[Char]] 14:50:31 Couldn't match `Char' against `[Char]' 14:50:37 What do you want it to do? 14:51:18 I want to apply f on every element of [1..32323], f returns [Char], and pass it to hPutStr 14:52:00 You'll have to fold the [[Char]] into a [Char], or map hPutStr onto the [[Char]] 14:52:21 how can I fold [[Char]] to [Char]? 14:55:41 One way is: foldl (++) "" list 14:55:54 That just concatenates all the elements of list into one big list. 14:56:07 Might not be the fastest thing in the world, though. 14:56:10 As does concat 14:56:27 Oh, didn't realize there was a concat function. 14:56:55 it works but very slow 14:58:09 * xbill foldl (++) [] is just concat no? 14:58:53 We've done that :-) 14:59:00 I'm not sure if map (\x -> hPutStr fh $ f x) [1..32323] would work. 14:59:12 Seems likve I've tried something like that before and failed. :P 14:59:21 Especially since it doesn't really make sense. 14:59:31 You'd need sequence_ $ on the front as I said earlier 14:59:36 Yurik: how do you run it? hugs? 14:59:46 * Logan should look up that "sequence_" think. 14:59:48 thing 14:59:56 dennisb ghc 15:00:04 good 15:00:10 Ah, it's in the tutorial. 15:00:28 Heh, I've been doing everything the sequence_ way all along (just not generalized like sequence_ is). 15:00:30 Silly me. 15:00:43 so is there any way to speed up? 15:01:05 --- join: nodie (~nodie@213-99-66-207.uc.nombres.ttd.es) joined #haskell 15:01:19 hi 15:01:35 it is painfully slow now :-( 15:01:43 Which way are you doing it? 15:02:08 hPutStr fh (foldl (++)""(map(f)[1..10000])) 15:02:18 Yeah, that way is doomed to be slow. 15:02:25 Do what Igloo said... 15:02:32 sequence_ $ map (\x -> hPutStr fh $ f x) [1..32323] 15:03:05 oh, it is much more faster! :-) 15:03:29 foldr (++) "" ought to be much faster 15:03:41 Being O(n) rhather than O(n^2) 15:04:10 * Igloo checks - yup, concat uses foldr 15:04:20 It turns out sequence_ $ map ((hPutStr fh) . f) [1..32323] is equivalent. 15:04:33 I'm still folding my brian around the concept of folding. 15:04:40 easy 15:04:49 It's simple, but I'm not used to thinking about it. 15:04:53 me is testing on ~12Mb of data output 15:05:50 real1m5.625s 15:06:13 * Yurik is thinking, is it fast or slow to write 12mb of data in ~1min 15:06:32 Depends on your OS and hardware. :P 15:06:50 On my computer, I can write probably 50 times that much data in a minute. 15:06:52 Linux 2.4 running on Celeron-533 15:07:08 I know a few days ago I witnessed C code generating 400 MB of data and saving to a file in 30 seconds. 15:08:51 What is your definition of f? I'd like to try benchmarking it myself. :P 15:09:35 f :: Word32 -> [Char] f n = [b 24 n, b 16 n, b 8 n, b 0 n] where b i n = chr $ fromEnum $ (n `shiftR` i) .&. 255 15:10:01 Are you using optimisation OOI? 15:10:32 Hmm, where is shiftR defined? 15:10:40 Bits 15:11:19 OOI? 15:11:50 Out Of Interest 15:12:00 Word32? 15:12:06 Yes, Word32 15:12:20 It is not my function, it is dennisb function :-) 15:12:25 Where is Word32 defined? 15:12:29 Word 15:13:59 So is it possible to optimize f? 15:14:27 I mean with -O2 15:15:14 * Yurik is timing with -O2 15:15:18 Yeah, code without -O is pretty slow. 15:15:25 OK 15:15:36 Bits 15:15:38 is 2 a maximum level of an optimization in ghc/ 15:15:40 is 2 a maximum level of an optimization in ghc? 15:16:01 the same result 15:16:42 real 0m4.815s 15:16:58 Logan and what is the size of generated file? 15:17:06 12582912 15:17:23 Linux, XP 1700+ 15:17:31 1700+ :-) 15:17:34 None of this is actually goign to the hard drive, I bet (just being buffered). 15:18:23 what kind of buffering have you used? 15:19:30 Just referring to the OS's filesystem. 15:19:56 I'm sure 12 MB is smaller than the buffer allocated by the kernel for that device. 15:21:07 so, no kind of buffering you used? 15:21:15 I just used the code you gave me. :P 15:21:20 --- quit: nodie ("[BX] Been around the world and found that only stupid people are breeding.") 15:21:25 heh :-) 15:21:28 Though I didn't use hPutStr, just regular putStr and redirection to a file. 15:21:56 anyway, how should I set up buffering with hSetBuffering? 15:22:34 It was actually a little slower with hPutStr. 15:22:36 I don't know. 15:22:55 No instance for (Num (Maybe Int)) 15:23:08 while I'm trying to ... $ BlockBuffering 4096096 15:23:41 You have to use Just. 15:23:48 BlockBuffering $ Just 4096096 15:24:06 oh, ok 15:24:15 I'm trying with Nothing now 15:24:15 That's the first instance of optional parameters in haskell I've seen. Didn't know you could do that with Maybe. 15:24:30