Jun 21 17:59:01 * DanielYoung (n=quassel@124-171-216-56.dyn.iinet.net.au) has joined #bfpg Jun 21 17:59:03 Hehe Jun 21 17:59:20 hi all Jun 21 17:59:32 * TheColonial (n=OJ@115.64.213.55) has joined #bfpg Jun 21 17:59:44 Evening all. Jun 21 17:59:53 <`Zerax`> 50 seconds till brain melting Jun 21 18:00:08 s/melting/expanding/ I think you mean Jun 21 18:00:23 * toddity (n=toddity@c211-31-28-249.rochd5.qld.optusnet.com.au) has joined #bfpg Jun 21 18:00:48 we'll give it another 5 for the late starters Jun 21 18:00:56 melting implies expanding ;) Jun 21 18:01:05 TonyMorris: seconds or minutes? :p Jun 21 18:00:34 hours Jun 21 18:00:39 to make you regret waking up Jun 21 18:00:42 DanielYoung: not really... you melt ice, and initially it takes up less volumes... Jun 21 18:00:45 medfly: heh Jun 21 18:00:51 I'm glad I came in early - I had it down as 6:30 in my calendar Jun 21 18:01:09 * ivanm notices that for the most part, the "use real names for your nick" suggestion didn't take off... Jun 21 18:01:22 * TheColonial is now known as OliverReeves[OJ] Jun 21 18:01:29 * OliverReeves[OJ] forgot ;) Jun 21 18:01:34 I didn't know Jun 21 18:01:37 * `Zerax` wasn't told Jun 21 18:01:37 For those on linux I just discovered there is also a haskell interpreter called "hugs" Jun 21 18:01:42 * `Zerax` is now known as OliverTaylor Jun 21 18:01:50 GHCi works fine on Linux Jun 21 18:01:58 aha... Jun 21 18:01:59 please stick with ghci :) Jun 21 18:02:00 I don't use my real name online. Jun 21 18:02:06 didn't realise there was that. Jun 21 18:02:09 and hugs used to work on windows, i haven't used it in a long time tho Jun 21 18:02:11 * Woof is now known as VM- Jun 21 18:02:15 * Apocalisp is now known as RunarOli Jun 21 18:02:31 SteveDalton: don't use it, it's old ;-) Jun 21 18:02:36 Yeah - I am just using ghc as instructed tony - I was just messing about with other stuff while I waited - didn't realise there was ghci... don't need hugs now then Jun 21 18:02:44 ok - apt-get remove... :) Jun 21 18:02:45 ghci seems to have a lot more activity around it Jun 21 18:02:46 SteveDalton: hugs devel is not really active anymore afaik Jun 21 18:02:57 OliverReeves[OJ]: well, some activity > no activity ;-) Jun 21 18:03:08 hmm time for one more donut then ;o) Jun 21 18:03:11 right it is gone... ghci all the way :) Jun 21 18:03:14 ivanm: I'm certianly no authority. Hence "seems" :) Jun 21 18:03:21 SteveDalton: good man :) Jun 21 18:03:36 by the way... my Real World Haskell book arrived! WAHOO! Jun 21 18:03:52 I was referring to some old doco that didn't have ghci - recommended hugs for interpreter. Jun 21 18:03:55 OliverReeves[OJ]: my main problem with RWH is how big the errata list is :( Jun 21 18:04:13 bad news is, the stupid Australia Post dude threw it up on the balcony. Resulting in dog-eared corners.. and it being wet coz of the rain :( Jun 21 18:04:14 * VM- is now known as Vin Jun 21 18:04:15 there's a section where the code evaluated to an error (forgot an import), and they talk about the output of their commands :s Jun 21 18:04:25 OliverReeves[OJ]: :o Jun 21 18:04:25 ivanm: true, but hey, it's better than no book at all. Jun 21 18:04:30 where did you get it from? Jun 21 18:04:38 OliverReeves[OJ]: Where did you get it from. Jun 21 18:04:47 If it's Amazon - tell them - they'll send you a new one. Jun 21 18:04:52 * boscop (n=boscop@e181181202.adsl.alicedsl.de) has joined #bfpg Jun 21 18:04:53 SteveDalton: ivanm: yeah, not impressed. Fishpond. But it came via OzPost. Jun 21 18:04:56 www.booko.com.au Jun 21 18:05:06 OliverReeves[OJ]: I got it from fishpond... no problems :s Jun 21 18:05:10 i'm going to have a go at fishpond to see if they can give me another.. but it's the fault of ozpost. Jun 21 18:05:11 BOSCOP Jun 21 18:05:12 You made it Jun 21 18:05:15 yeah i dont blame fishpond. Jun 21 18:05:15 $55, pre-ordered! \o/ Jun 21 18:05:22 Was gonna kill you Jun 21 18:05:28 i blame the idiot delivery guy. obviously has the brains of a wet lettuce. Jun 21 18:05:29 Auspost soaked 3 books of mine... Amazon just sent me out new ones... never asked for old ones back so I dried them out on radiator and gave them to my mates :) Jun 21 18:05:31 >ivanm< http://projects.tmorris.net/public/haskell-parsers/artifacts/0.9/chunk-html/index.html Jun 21 18:05:44 SteveDalton: Yeah we really need amazon.com.au :) Jun 21 18:05:48 Vin: how often do you change your nick? Jun 21 18:05:50 ok get started then? Jun 21 18:05:51 Yep. Jun 21 18:05:57 yup! let's go. Jun 21 18:06:00 did it just start? Jun 21 18:06:02 boscop: It was suggested that we use real names here Jun 21 18:06:13 delivery guys here send things in closed plastic containers even though it is unlikely it'll rain Jun 21 18:06:19 boscop: /nick Foo Jun 21 18:06:21 OK, so what we are going to do is write our own library for parsing Jun 21 18:06:42 we will do it incrementally stopping for any questions along the way Jun 21 18:06:56 ivanm can also help out with questions and so can Cale but he doesn't seem to be around at the moment Jun 21 18:07:18 feel free to /msg me (if I'm still around, might be going off for dinner soonish) if you don't want to ask everyone Jun 21 18:07:24 s/everyone/publically/ Jun 21 18:07:33 I have some slides here that I will make available afterward -- which includes the final Haskell source file -- I have also written the parsing library in Scala and Java for reference Jun 21 18:07:45 type 'q' at any time there is a question so we know to stop and await a question Jun 21 18:08:00 you will need a text editor and a ghci command prompt Jun 21 18:08:05 let's go Jun 21 18:08:34 http://video.java.no/2009/tony/docbook/01/MyParser.hs Jun 21 18:08:36 ok Jun 21 18:08:47 enter that into a text file called MyParser.hs Jun 21 18:08:54 then at the ghci prompt type Jun 21 18:08:58 :load MyParser.hs Jun 21 18:09:03 the prompt should then read Jun 21 18:09:07 *MyParser> Jun 21 18:09:46 * toddity has quit () Jun 21 18:10:01 I will assume we are all good in 10 seconds Jun 21 18:10:21 ok, so the next thing we will do is define what a parser is Jun 21 18:10:26 A parser is a function that accepts an input String and either fails or produces a value and the remaining input String. Jun 21 18:10:57 http://video.java.no/2009/tony/docbook/02/MyParser.hs add this to your source file Jun 21 18:11:00 then at ghci type Jun 21 18:11:01 * medfly (n=lio@unaffiliated/medfly) has left #bfpg ("Leaving") Jun 21 18:11:03 :reload Jun 21 18:11:25 examine the type of the 'parse' function that we just defined Jun 21 18:11:30 you can do this by typing Jun 21 18:11:32 :type parse Jun 21 18:11:48 parse accepts a parser and input and either fails or produces a value and the remaining input. Jun 21 18:11:59 * MitchellRiley (n=user@d58-111-182-36.rdl7.qld.optusnet.com.au) has joined #bfpg Jun 21 18:12:27 10 seconds Jun 21 18:12:53 we see that when our Parser fails, it doesn't produce any error message or a position of failure Jun 21 18:12:59 we are doing this to keep the exercise simple Jun 21 18:13:04 * mvr has quit (Client Quit) Jun 21 18:13:05 let's define our first parser Jun 21 18:13:13 http://video.java.no/2009/tony/docbook/03/MyParser.hs Jun 21 18:13:36 this parser consumes no input (by returning the input) and always produces the given value Jun 21 18:13:56 if you are used to OO-like languages, the data Parser declaration is like an interface, and value is like an instance of that interface Jun 21 18:13:58 * __andrew__ (n=andrew@124-171-175-253.dyn.iinet.net.au) has joined #bfpg Jun 21 18:13:59 :q this is with "data" right? Jun 21 18:14:12 SteveDalton, with "data"? I'm not sure what you mean Jun 21 18:14:27 within the {} I meant - sorry Jun 21 18:14:31 I got it. Jun 21 18:14:36 no sorry, we are appending to the file as we go Jun 21 18:14:47 ok. got you Jun 21 18:15:08 http://hpaste.org/fastcgi/hpaste.fcgi/view?id=6101#a6101 our file should look like this Jun 21 18:15:15 we are :reloading at ghci as we go Jun 21 18:15:25 __andrew__, want to join? Jun 21 18:15:35 <__andrew__> hello Jun 21 18:15:38 <__andrew__> yeah just catching up Jun 21 18:16:05 __andrew__, add this to a source file called MyParser.hs http://hpaste.org/fastcgi/hpaste.fcgi/view?id=6101#a6101 and load it into ghci using :load MyParser.hs Jun 21 18:16:17 * Twey (n=Twey@unaffiliated/twey) has joined #bfpg Jun 21 18:16:28 <__andrew__> ok Jun 21 18:16:45 so we might now experiment with our two functions, value and parse Jun 21 18:16:50 :type parse (value "boo") Jun 21 18:16:58 parse (value 7) "input" Jun 21 18:17:03 parse (value "boo") "input" Jun 21 18:17:14 note that the input remains unaltered and the given value is also returned Jun 21 18:17:25 let us now write another function which is a parser that always fails Jun 21 18:17:32 http://video.java.no/2009/tony/docbook/04/MyParser.hs Jun 21 18:17:47 this parser is not particularly interesting but we will use it later on Jun 21 18:18:19 the underscore denotes that we are ignoring the input, since we are always failing, regardless of the input Jun 21 18:18:42 Let's create a parser that consumes one character and produces that character, unless the input is empty, in which case fail. Jun 21 18:18:55 http://video.java.no/2009/tony/docbook/05/MyParser.hs Jun 21 18:19:10 Experiment with the character parser Jun 21 18:19:15 :type parse character Jun 21 18:19:19 parse character "abc" Jun 21 18:19:24 parse character "" Jun 21 18:19:53 notice that the parser fails when there is no more input, otherwise, one character is produced and the input moves along one character Jun 21 18:20:21 Write a function that accepts two parsers and returns a parser that tries the first parser for success. If that parser fails, try the second parser. Jun 21 18:20:30 we will call this 'the choice parser' Jun 21 18:20:36 http://video.java.no/2009/tony/docbook/06/MyParser.hs Jun 21 18:20:52 there may be some new syntactic surprises there -- questions? Jun 21 18:21:21 experiment with the choice parser Jun 21 18:21:26 the ||| parser will have a default left fixity Jun 21 18:21:34 *MyParser> let p = character ||| value 'x' in parse p "abc" Jun 21 18:22:12 *MyParser> let p = character ||| value 'x' in parse p "" Jun 21 18:22:18 *MyParser> let p = character ||| failed in parse p "abc" Jun 21 18:22:23 *MyParser> let p = character ||| failed in parse p "" Jun 21 18:22:26 questions? Jun 21 18:22:39 :q Where is "{...}" of data Parser a = P { ... } documented? Or what does it mean? Jun 21 18:22:55 AdrianPronk: it creates a record data structure Jun 21 18:22:57 that's called record syntax Jun 21 18:23:17 it is a bit like naming a method of a class in OO Jun 21 18:23:21 it gives you functions to pull out individual functions for free Jun 21 18:23:33 and you can also use them as shortcuts to change part of a data structure Jun 21 18:23:41 observe the type of parse :: Parser a -> String -> Maybe (String, a) Jun 21 18:23:46 they become more useful for larger data structures (e.g. 10 fields) Jun 21 18:23:59 AdrianPronk, ok? Jun 21 18:24:45 Does it define an implemantion (for parse in this case)? Jun 21 18:24:46 It essentially means that you have a P (String -> Maybe (String, a)), but you get a function parse :: Parser -> String -> Maybe (String, a) for free, as well as some magic syntax for pattern-matching and updating individual fields Jun 21 18:24:50 AdrianPronk: no Jun 21 18:24:51 Yes Jun 21 18:24:56 heh Jun 21 18:25:10 it defines what the type parse must take Jun 21 18:25:12 AdrianPronk, no, it doesn't -- each of our functions thereafter define implementations Jun 21 18:25:21 actual implementations are done with P ( \s -> ... ) Jun 21 18:25:22 * Twey tilts his head. Jun 21 18:25:32 AdrianPronk, it is very much like (Java): interface Parser { Pair parse(String s); } Jun 21 18:26:04 Sorry, you'll have to rephrase the question. I took it to mean you were asking about whether the getter function parse had to be defined manually later on, to which the answer is no. Jun 21 18:26:21 Twey: nah, he meant does parse have a default implementation (I think) Jun 21 18:26:22 so P (\s -> ... ) defines parse for this case? Jun 21 18:26:28 AdrianPronk: yes Jun 21 18:26:32 The function parse gets defined automatically; the function that we insert into the field doesn't Jun 21 18:26:43 Is there some kind of positional correspondence? Jun 21 18:26:46 AdrianPronk, precisely Jun 21 18:26:50 well, parse returns the (\s -> ...) function Jun 21 18:27:10 AdrianPronk, the P constructor must take a function of the correct type (or you will receive a type error) Jun 21 18:27:30 AdrianPronk, but if the data type had more than one field, then they must appear in the order that they are defined Jun 21 18:27:31 ok, press on. I'll see how I go... Jun 21 18:27:51 consider: data Person = Per Int String Jun 21 18:27:59 then you must Per anInt aString Jun 21 18:28:26 if a Parser produces values of type 'a' and we have a function a -> b we can get a Parser that produces values of type 'b' Jun 21 18:28:37 Write a function that accepts a parser and a function from its produced value to another value to return a parser that potentially produces that value. We will see later why this function is useful. Jun 21 18:28:45 http://video.java.no/2009/tony/docbook/07/MyParser.hs Jun 21 18:28:51 * OliverReeves[OJ] is now known as OliverReeves[Bac Jun 21 18:29:00 * OliverReeves[Bac is now known as OliverReeves[Brb Jun 21 18:29:16 Experiment with the mapping parser Jun 21 18:29:20 *MyParser> let p = mapParser character toUpper in parse p "abc" Jun 21 18:29:23 note that this is an instance of a "higher-order function", that is we have a function that takes a function as a parameter Jun 21 18:29:29 *MyParser> let p = mapParser character toUpper in parse p "" Jun 21 18:29:58 notice here that when the parser produces a character value, it has converted it to upper-case from the input Jun 21 18:30:45 Write a function that accepts a parser and a function from its produced value to another parser producing values of some type and returns a parser producing values of that same type. Again, we will see later why this function is useful. Jun 21 18:30:53 Parser a -> (a -> Parser b) -> Parser b Jun 21 18:31:12 http://video.java.no/2009/tony/docbook/08/MyParser.hs Jun 21 18:31:38 Experiment with the binding parser Jun 21 18:31:42 *MyParser> let p = bindParser character (\c -> if isUpper c then value c else failed) in parse p "" Jun 21 18:31:47 *MyParser> let p = bindParser character (\c -> if isUpper c then value c else failed) in parse p "abc" Jun 21 18:31:53 *MyParser> let p = bindParser character (\c -> if isUpper c then value c else failed) in parse p "Abc" Jun 21 18:32:24 I hope I haven't lost anyone :) Jun 21 18:33:03 this all makes perfect sense to everyone so far? Jun 21 18:33:06 do you have the complete source file up until this point? Jun 21 18:33:07 does everyone understand TonyMorris, or -- apart from AdrianPronk -- too shy to ask questions? :p Jun 21 18:33:37 <__andrew__> i think it's good so far Jun 21 18:33:38 koala_man, http://hpaste.org/fastcgi/hpaste.fcgi/view?id=6101#a6102 Jun 21 18:33:39 * juhp is following as fast as he can :) Jun 21 18:33:42 * AdrianPr` (n=user@203-214-84-239.dyn.iinet.net.au) has joined #bfpg Jun 21 18:33:42 * AdrianPronk__ (n=AdrianPr@203-214-84-239.dyn.iinet.net.au) has joined #bfpg Jun 21 18:33:52 great, thanks Jun 21 18:33:55 juhp, shall I slow down? Jun 21 18:34:07 TonyMorris: not yet :) Jun 21 18:34:13 we might do a bit more and take a 5 minute break to have a look and maybe form some questions Jun 21 18:34:24 TonyMorris: maybe give a recap of what we've done so far? Jun 21 18:34:32 yes Jun 21 18:34:37 That'd be good Jun 21 18:34:52 yep, just trying to keep up :) Jun 21 18:34:52 OK, so far we have defined what a parser is with a data declaration Jun 21 18:35:03 then we have written some parser values and functions over parsers Jun 21 18:35:09 note that Parser is just a wrapper around a parsing function Jun 21 18:35:24 we have written a parser that takes a value and input and returns that value and input unchanged Jun 21 18:35:33 we have written a parser that always fails regardless of the input Jun 21 18:35:47 Hi Tony - you lost me a while back... I am going to do some more reading tomorrow and come back to the transcript and have another go... just go on without me... thanks anyway - appreciate you making all this effort. :) Jun 21 18:35:48 we have written a parser that consumes a single character, unless there isn't one, in which case, it fails Jun 21 18:36:10 we have written a parser that takes two parsers, tries the first one, if it fails, tries the second (it might fail too) Jun 21 18:36:26 we have written a function to map a function across the values produced by a parser Jun 21 18:36:28 :q Jun 21 18:36:38 we have written a function that binds a function across a parser Jun 21 18:36:41 don't quite get this syntax: v@(Just _) -> v Jun 21 18:36:49 toddity, that is called as-pattern syntax Jun 21 18:37:02 so 'v' is the name of the entire expression(Just _) Jun 21 18:37:19 i.e. it is the value that is being matched on (p1 s) Jun 21 18:37:28 and the underscore denotes that we don't care about the value Jun 21 18:37:41 all we care about is that (p1 s) produced a successful result (Just) Jun 21 18:37:56 in which case, return that entire result (Just theResult) Jun 21 18:37:59 which is named 'v' **** BEGIN LOGGING AT Sun Jun 21 18:45:12 2009 Jun 21 18:45:12 * Now talking on #bfpg Jun 21 18:45:12 * Topic for #bfpg is: Brisbane Functional Programming Group http://www.meetup.com/Brisbane-Functional-Programming-Group-BFG/ Next meeting: http://www.meetup.com/Brisbane-Functional-Programming-Group-BFG/calendar/10652263/ | Special thanks to scalabin, the Norwegian Scala group for hosting presentation material Jun 21 18:45:12 * Topic for #bfpg set by dibblego at Sun Jun 21 10:53:12 2009 Jun 21 18:45:13 * Received a CTCP VERSION from freenode-connect Jun 21 18:45:17 * dobblego sets mode -e dobblego Jun 21 18:45:17 * You are now known as TonyMorris Jun 21 18:45:18 -NickServ- This nickname is registered. Please choose a different nickname, or identify via /msg NickServ identify . Jun 21 18:45:21 ivanm: Operation timed out xD Jun 21 18:45:24 \o/ Jun 21 18:45:28 >NickServ< identify **** Jun 21 18:45:29 TonyMorris is back! Jun 21 18:45:29 -NickServ- You are now identified for TonyMorris. Jun 21 18:45:29 * services. sets mode +e TonyMorris Jun 21 18:45:30 how efficient is this sort of parsing compared to LR parsers? Jun 21 18:45:40 koala_man: don't know what LR parsers are, sorry Jun 21 18:45:53 and I'm not sure in which direction TonyMorris is going to take this (backtracking parsers, etc.) Jun 21 18:46:12 it can be very efficient in a lazy language Jun 21 18:46:14 It is an LR parser, so far Jun 21 18:46:19 * JustinNShield (n=miguel_b@ppp118-208-121-143.lns4.bne4.internode.on.net) has joined #bfpg Jun 21 18:46:27 toddity, that is called as-pattern syntax Jun 21 18:46:27 so 'v' is the name of the entire expression(Just _) Jun 21 18:46:27 i.e. it is the value that is being matched on (p1 s) Jun 21 18:46:27 and the underscore denotes that we don't care about the value Jun 21 18:46:33 all we care about is that (p1 s) produced a successful result (Just) Jun 21 18:46:33 in which case, return that entire result (Just theResult) Jun 21 18:46:33 which is named 'v' Jun 21 18:46:40 Python interface unloaded Jun 21 18:46:40 Tcl interface unloaded **** ENDING LOGGING AT Sun Jun 21 18:46:40 2009 Jun 21 18:47:10 moving on then? Jun 21 18:47:12 It's just a different way of structuring it Jun 21 18:47:22 many thanks, all good. Jun 21 18:47:50 so next we'll define a function similar to bindParser, except you don't pass it a function -- instead pass it another parser which is always returned Jun 21 18:47:59 This will be useful for example, when we wish for a parser to consume white-space but not do anything with that white-space. Jun 21 18:48:07 http://video.java.no/2009/tony/docbook/09/MyParser.hs Jun 21 18:48:41 Write a function that accepts a list of parsers and produces a parser of lists by calling bindParser and mapParser as you go along the list. Jun 21 18:48:51 http://video.java.no/2009/tony/docbook/10/MyParser.hs Jun 21 18:50:16 *MyParser> let p = sequenceParser [character, value 'b', character] in parse p "" Jun 21 18:50:21 *MyParser> let p = sequenceParser [character, value 'b', character] in parse p "a" Jun 21 18:50:25 *MyParser> let p = sequenceParser [character, value 'b', character] in parse p "ab" Jun 21 18:50:30 *MyParser> let p = sequenceParser [character, value 'b', character] in parse p "abc" Jun 21 18:50:43 we see here that we pass a list of parsers and get back a parser that produces a list Jun 21 18:50:52 * AdrianPronk has quit (Connection timed out) Jun 21 18:50:57 Write a function that accepts an integer (n) and a parser and returns a parser that binds 'n' times. We can do this by using the library function replicate. Jun 21 18:50:58 * AdrianPronk_ has quit (Read error: 110 (Connection timed out)) Jun 21 18:51:08 * AdrianPronk__ is now known as AdrianPronk Jun 21 18:51:15 http://video.java.no/2009/tony/docbook/11/MyParser.hs Jun 21 18:51:30 just so you can all see what replicate does: Jun 21 18:51:38 > replicate 10 "hi" Jun 21 18:51:40 ["hi","hi","hi","hi","hi","hi","hi","hi","hi","hi"] Jun 21 18:52:01 yes, so replicate n always produces a list with the length n Jun 21 18:52:18 so our returned parser from thisMany will always produce a list of the given length Jun 21 18:52:37 this might be useful, for example, we want to parse exactly 5 characters Jun 21 18:52:44 then: thisMany 5 character Jun 21 18:52:49 how to use >>> ? Jun 21 18:53:34 you mean, an example use? Jun 21 18:53:51 juhp: parser1 >>> parser2 Jun 21 18:53:52 parse (character >>> value "foo") "abc" Jun 21 18:54:08 TonyMorris: yes Jun 21 18:54:11 (which is the same as writing (>>>) parser1 parser2) Jun 21 18:54:20 parse (thisMany 5 character) "123456" Jun 21 18:54:33 Operators in Haskell are just normal functions that can be written in between their first two arguments. Jun 21 18:54:43 thanks Jun 21 18:54:54 *MyParser> parse (character >>> character) "abc" Jun 21 18:54:54 Just ("c",'b') Jun 21 18:55:02 notice that the 'a' is dropped (ignored) Jun 21 18:55:03 -> koala_man Jun 21 18:55:04 The magic is triggered by the identifier consisting entirely of non-alphabetical characters Jun 21 18:55:32 ah Jun 21 18:55:36 * dibblego has quit (Read error: 110 (Connection timed out)) Jun 21 18:55:39 which aren't (), etc. Jun 21 18:55:41 ok, a few more mundane details to go, until we get into the good bit :) Jun 21 18:55:44 * Disconnected (Invalid argument). **** ENDING LOGGING AT Sun Jun 21 18:55:44 2009 Jun 21 18:55:45 Aye. Jun 21 18:55:52 (there are some reserved symbols) Jun 21 18:56:04 we are now going to write two functions, which depend on each other Jun 21 18:56:07 * dibblego (n=nobody@220-245-107-64.static.tpgi.com.au) has joined #bfpg **** BEGIN LOGGING AT Sun Jun 21 18:56:09 2009 Jun 21 18:56:20 One takes a parser and applies itself zero or many times (list) and the other takes a parser and applies itself one or many times (many1). These functions are useful for such things as parsing white-space, where we may take a parser of a single white-space character and obtain a parser of e.g. one or many white-space characters. Jun 21 18:56:29 http://video.java.no/2009/tony/docbook/12/MyParser.hs Jun 21 18:56:32 * AdrianPr` is now known as AdrianPronk_ Jun 21 18:56:35 co-recursive functions! \o/ Jun 21 18:57:07 we want brains to expand, right? :) Jun 21 18:57:14 definitely! Jun 21 18:57:15 always Jun 21 18:57:16 Probably better to stick with calling them ‘mutually recursive’ to avoid confusion :-P Jun 21 18:57:31 Twey: awwww.... that so no comanads, etc.? :p Jun 21 18:57:36 s/that// Jun 21 18:57:43 Hehe Jun 21 18:57:51 *MyParser> let p = many1 character in parse p "" Jun 21 18:58:00 notice that this parse fails Jun 21 18:58:05 *MyParser> let p = list character in parse p "" Jun 21 18:58:12 this one succeeds but has exhausted input Jun 21 18:58:18 *MyParser> let p = many1 character in parse p "abc" Jun 21 18:58:23 *MyParser> let p = list character in parse p "abc" Jun 21 18:59:01 everyone on track? Jun 21 18:59:11 I think so Jun 21 18:59:16 Clear as mud! Jun 21 18:59:24 http://hpaste.org/fastcgi/hpaste.fcgi/view?id=6104#a6104 here is where we are up to Jun 21 18:59:39 Write a parser that consumes a character (or fails if there isn't one) and ensures that character meets a given predicate. We might also use this parser to write another parser (is) that parses a specific character. Jun 21 18:59:50 http://video.java.no/2009/tony/docbook/13/MyParser.hs Jun 21 19:00:02 Experiment with these parsers Jun 21 19:00:05 list -> parse as manay as you can; many1 -> parse at least 1, and then as many as you can Jun 21 19:00:23 let p = satisfy isUpper in parse p "" Jun 21 19:00:26 * OliverReeves[Brb is now known as OliverReeves Jun 21 19:00:31 let p = satisfy isUpper in parse p "abc" Jun 21 19:00:52 these two fail since there is not a first character that matches the predicate (isUpper) Jun 21 19:00:57 *MyParser> let p = satisfy isUpper in parse p "Abc" Jun 21 19:00:59 note that in is, (== c) is known as a _section_, as we've filled in one side of the operator == Jun 21 19:01:00 this one succeeds Jun 21 19:01:21 *MyParser> let p = is 'a' in parse p "" Jun 21 19:01:33 fails since no character Jun 21 19:01:34 *MyParser> let p = is 'a' in parse p "abc" Jun 21 19:01:41 succeeds and consumes 'a' Jun 21 19:01:46 *MyParser> let p = is 'a' in parse p "xbc" Jun 21 19:01:54 fails since not 'a' to consume Jun 21 19:02:34 Let us now write a bunch of parsers that use satisfy and apply the predicate. Jun 21 19:02:40 http://video.java.no/2009/tony/docbook/14/MyParser.hs\ Jun 21 19:02:41 http://video.java.no/2009/tony/docbook/14/MyParser.hs Jun 21 19:02:58 * MatCallan (n=irchon@114.73.243.80) has joined #bfpg Jun 21 19:03:08 I will give 3 or 4 minutes to consume at this point Jun 21 19:03:42 note that isDigit does '0'..'9' Jun 21 19:03:50 :q Can you create a section for multi-parameter alphabetic functions so that the arbitray parameters are pre-specified? Jun 21 19:03:52 there are other functions that do other types of "numeric" characters Jun 21 19:04:05 AdrianPronk: Yes and no Jun 21 19:04:12 e.g. isNumber also works on roman numerals Jun 21 19:04:35 AdrianPronk: You'd use instead a technique called ‘currying’ Jun 21 19:04:55 ok, I'll read about currying offline Jun 21 19:05:07 * Twey nods. Jun 21 19:05:10 ivanm: Can you give an example of using isNumber with a roman numeral? Jun 21 19:05:19 It may be a bit in-depth and/or confusing to get into here. Jun 21 19:05:29 Vin: no ;-) Jun 21 19:05:38 I'm reading off the documentation ;-) Jun 21 19:06:03 and I have no idea how it reads Roman numerals in :s Jun 21 19:06:11 Experiment with some of these parsers Jun 21 19:06:14 parse upper "" Jun 21 19:06:16 this fails Jun 21 19:06:18 Okay. isNumber 'v' doesn't work. But nvm Jun 21 19:06:23 parse upper "a" Jun 21 19:06:26 this also fails Jun 21 19:06:31 parse upper "A" Jun 21 19:06:34 Where's this isNumber function, sorry? Jun 21 19:06:37 this succeeds and consume input Jun 21 19:06:42 >lambdabot< @type isNumber Jun 21 19:06:49 @type isNumber Jun 21 19:06:51 Char -> Bool Jun 21 19:06:52 Vin: it uses unicode codpoints, I think Jun 21 19:06:54 Twey: Data.Char Jun 21 19:06:57 Ah Jun 21 19:07:00 @index isNumber Jun 21 19:07:00 ivanm: Oh, okay Jun 21 19:07:01 bzzt Jun 21 19:07:06 @src isNumber Jun 21 19:07:06 Source not found. Have you considered trying to match wits with a rutabaga? Jun 21 19:07:07 *codepoints Jun 21 19:07:23 Twey: it does some weird parsing thingy using enums (I'm reading the source) Jun 21 19:07:29 @hoogle isNumber Jun 21 19:07:29 Data.Char isNumber :: Char -> Bool Jun 21 19:07:32 Doesn't seem to do Roman numerals. Jun 21 19:07:36 let p = list digit in parse p "79abc" Jun 21 19:07:50 observe that this parser reads off and produces the first two digits Jun 21 19:08:06 So far we have constructed a small library of parsers. Some of these parsers are constructed from other parsers. This notion of constructing units of work from other, smaller units of work is the essence of functional programming. Jun 21 19:08:19 Does it scale? Jun 21 19:08:25 Can we write even more higher-level parsers with our library by gluing parsers? Jun 21 19:08:37 Suppose we have a data structure to represent a person. The person data structure has these attributes: Jun 21 19:08:41 Age: positive integer Jun 21 19:08:45 First Name: non-empty string that starts with a capital letter Jun 21 19:08:50 Surname: string that starts with a capital letter and is followed by 5 or more lower-case letters Jun 21 19:08:54 Gender: character that must be 'm' or 'f' Jun 21 19:08:57 I think I missed something, where is "digit"? Jun 21 19:09:00 Phone: string of digits, dots or hyphens but must start with a digit and end with a hash (#) Jun 21 19:09:06 Vin, http://video.java.no/2009/tony/docbook/14/MyParser.hs Jun 21 19:09:20 Oops, didn't reload Jun 21 19:09:22 Right Jun 21 19:09:31 http://video.java.no/2009/tony/docbook/15/MyParser.hs Jun 21 19:09:37 here is our Person data type declaration Jun 21 19:09:57 "deriving Show" is a bit like saying (Java) "generate an automate toString" Jun 21 19:10:15 deriving is a keyword Jun 21 19:10:23 the Show class (which has the show function) is generally used to print out the "code" of that datatype Jun 21 19:10:31 you normally don't use it for pretty-printing Jun 21 19:10:39 the reverse is the read function in the Read class Jun 21 19:10:45 (you can derive Read as well) Jun 21 19:10:57 notice that the constructor and data type have the same name (Person), though they are separate things Jun 21 19:11:19 so we pass Person 5 arguments to produce a value of type Person Jun 21 19:11:40 anyone want to take a punt at how the parser for parsing the Person's age might look? Jun 21 19:11:51 (using our library) Jun 21 19:11:53 natural Jun 21 19:12:02 * TonyMorris hands koala_man a cookie Jun 21 19:12:10 http://video.java.no/2009/tony/docbook/16/MyParser.hs Jun 21 19:12:33 how about parsing the first name? Jun 21 19:12:42 non-empty string that starts with a capital letter Jun 21 19:13:17 http://video.java.no/2009/tony/docbook/17/MyParser.hs Jun 21 19:13:31 we are "gluing" parsers together using bindParser and mapParser Jun 21 19:13:50 These two functions are turning out to be quite helpful. Jun 21 19:13:55 surname parser? Jun 21 19:14:03 string that starts with a capital letter and is followed by 5 or more lower-case letters Jun 21 19:14:10 why 5? Jun 21 19:14:15 arbitrary Jun 21 19:14:20 ok ;) Jun 21 19:14:22 just to confuse us and show off the parsing functions ;-) Jun 21 19:14:22 because they are the rules for this particular Person Jun 21 19:14:38 (obviously, Mr Ng isn't going to be pleased with us if we use these rules for the ATO or something...) Jun 21 19:14:48 my surname is a capital followed by 4 lower-case letters :( Jun 21 19:14:56 Heh Jun 21 19:14:59 koala_man: I'm sorry, but you're not a Person then Jun 21 19:15:00 ;-) Jun 21 19:15:11 for surname, we might use upper, (thisMany 5 lower) and (list lower) and glue them all together Jun 21 19:15:14 koala_man: you'll have to change your name :) Jun 21 19:15:20