July 2009 - Haskell-Cafe

Simple quirk in behavior of `mod`
by Nathan Bloomfield 22 Jul '09

22 Jul '09

Hello haskell-cafe; I'm fiddling with this<http://cdsmith.wordpress.com/2009/07/20/calculating-multiplicative-inverses…>blog post about inverting elements of Z/(p), trying to write the inversion function in pointfree style. This led me to try executing statements like n `mod` 0 which in the ring theoretic sense should be n, at least for integers*. (MathWorld agrees. <http://mathworld.wolfram.com/Congruence.html>) But Hugs gives a division by zero error! I'm more of a recreational haskell user and not too familiar with how the Prelude works. But I dug around a bit and saw this inGHC.Real: ( link<http://www.haskell.org/ghc/docs/latest/html/libraries/base/src/GHC-Real.htm…> ) > a `mod` b > | b == 0 = divZeroError > | a == minBound && b == (-1) = overflowError > | otherwise = a `modInt` b Is there a reason why n `mod` 0 is undefined in Haskell? Maybe this has already been considered for Haskell' and I'm just unaware. I did some digging in the archives and this discussion <http://markmail.org/message/5dmehw4lhu56x4zw#query:haskell%20%22%60mod%60%2…> <http://markmail.org/message/5dmehw4lhu56x4zw> from 2002 is the most relevant one I could find; it is suggested there that n `mod` 0 should be an error. Thanks all- Nathan Bloomfield *- The mod function is defined in the Integral class, and I'm not even sure how to interpret that. It looks kind of like a Euclidean domain.

3 2

Re[2]: [Haskell] TABI 0.1: a typeful tagged cross-language calling convention
by Bulat Ziganshin 21 Jul '09

21 Jul '09

Hello Felipe, Tuesday, July 21, 2009, 7:38:44 PM, you wrote: >> * passing complex datastructures between various languages forth >> and back, providing faster alternative to serialization approach > I've read the code and I don't see how tabi manages to do this. it's not yet implemented :))) my idea is to provide one more basic type, TABI_TABLE=6, that allows to nest TABI_MAPs recursively. also arrays of TABI_* types may be supported > It is serializing as well but it is including the type and name > information, so it seems that at best it would have the same > overhead as just serializing the data without any metadata. my first approach was to use just serialization (it was implemented in Lua sources included). but then i realized that it is monkey work - most time will be spent copying data to buffer, from buffer, searching for delimiters, processing type conversion. i'm pretty sure that C implementation of TABI_MAP will do its work (filling 3-4 words per parameter) an order of magnitude faster than serialization approach. also, serialized data anyway will be need to converted to datastructure like TABI_MAP in order to be used by calls like _str and _int for Haskell code, good optimized serialization library should have comparable speed. moreover, my current approach with malloc() calls probably makes encoding operations much slower if you have any specific needs, we can discuss it. my own needs is mainly combination of first and third clauses of quoted list, so second claim is somewhat artificial :) it is possible but only partially implemented -- Best regards, Bulat mailto:Bulat.Ziganshin@gmail.com

3 2

understanding HXT's insertChildrenAfter
by Jake 20 Jul '09

20 Jul '09

Playing on HXT 8.3.0 via cabal on GHC 6.10.2. While there are plenty of useful examples of extract data from XML, applying the appropriate API to transform and/or merging still baffle me. In this case, it is the insertChildrenAfter, let's try it with the following example: given 2 files --- update.xml <?xml version="1.0" encoding="UTF-8"?> <update> <hot>sky is now orange</hot> <hot>fish is fresh</hot> <hot>violets are blue</hot> </update> --- master.xml <?xml version="1.0" encoding="UTF-8"?> <news> <page1> <hot>sky is blue</hot> <hot>roses are red</hot> </page1> <page2> <story>game shipped on sunday</story> </page2> </news> with the intention of augment the master list with the new items: import Text.XML.HXT.Arrow import qualified Text.XML.HXT.DOM as DOM getMaster = readDocument [(a_validate,v_0)] "master.xml" >>> removeDocWhiteSpace getUpdate = readDocument [(a_validate,v_0)] "update.xml" >>> removeDocWhiteSpace selectUpdateArea = getChildren >>> hasName "news" >>> getChildren >>> hasName "page1" updateContent = insertChildrenAfter (selectUpdateArea) (getUpdate >>> getChildren >>> getChildren) main = do runX ( getMaster >>> updateContent >>> putXmlTree "-" ) instead the following tree is produced: ---XTag "/" | +---XTag "hot" | | | +---XText "sky is red" | +---XTag "hot" | | | +---XText "fish is fresh" | +---XTag "hot" | | | +---XText "violets are blue" | +---XTag "news" | +---XTag "page1" | | | +---XTag "hot" | | | | | +---XText "sky is blue" | | | +---XTag "hot" | | | +---XText "roses are red" | +---XTag "page2" | +---XTag "story" | +---XText "game shipped " Can someone give some examples of how insertChildrenAfter works? How about also some general guidelines when comes to merging/grafting XML documents. Jake

1 0

Pruning the Front Page
by Thomas DuBuisson 20 Jul '09

20 Jul '09

All, I think we should remove unmaintained or refocused implementations from the haskell.org front page. This would mean removing: YHC - unmaintained, hasn't built for me in a while LHC - Not a standalone compiler these days - perhaps should be linked as a GHC subproject? Other than those I'd propose NHC but I'm guessing that would get some pushback (even though it currently doesn't build for me). Before anyone brings up HUGS I'll just say that 1) it is used 2) it builds 3) at least it's been updated more recently than the Haskell langauge. Thomas

3 2

Ambiguous type variable - help!
by phil＠beadling.co.uk 20 Jul '09

20 Jul '09

Hi, I'm trying to work out how to handle a choice at runtime which determines what instance of a State monad should be used. The choice will dictate the internal state of the monad so different implementations are needed for each. I've concocted a very simple example to illustrate this (below) - but it doesn't compile because ghc complains that my type is ambiguous arising from my use of 'fromSeq'. I can kind-of see what the compiler is complaining about, I'm guessing because it is the internals of my type which dictate which state Monad to use and it can't know that? Thinking about it I tried making SeqType an instance of Sequence class, but had no luck here. I understand that haskell is static at compile time, so I'm looking for something like a template solution in C++ (rather than a virtual function style implementation). I see there are libraries out their which do this, but I was wondering in my simple example if there was a way of doing this without writing a load of boilerplate code in main (this would get ugly very quickly if you had loads of choices). If this is impossible does anyone have an example / advice of implementing simple template style code in Haskell? Any help or suggestions would be really appreciated. Many Thanks, Phil. Thus just implements a state Monad which counts up from 1 to 10, using either an Int or a Double depending on user choice. It's pointless of course, but illustrates my point. {-# LANGUAGE TypeSynonymInstances #-} import Control.Monad.State.Strict data SeqType = SeqDouble Double | SeqInt Int class SequenceClass a where nextSeq :: State a Int fromSeq :: SeqType -> a instance SequenceClass Int where nextSeq = State $ \s -> (s,s+1) fromSeq (SeqInt i) = i fromSeq _ = 0 instance SequenceClass Double where nextSeq = State $ \s -> (truncate s,s+1.0) fromSeq (SeqDouble d) = d fromSeq _ = 0.0 chooser :: String -> SeqType chooser inStr | inStr == "Double" = SeqDouble 1.0 | inStr == "Int" = SeqInt 1 | otherwise = SeqInt 1 main :: IO() main = do userInput <- getLine let result = evalState (do replicateM 10 nextSeq) $ fromSeq $ chooser userInput print result

2 2

Anglohaskell - wifi signups
by Philippa Cowderoy 20 Jul '09

20 Jul '09

Wifi signups are Anglohaskell are now on the wiki - please add your details by the 31st of July if you want a wifi account at MS Research for the Friday. Alternatively, reply to this email with your full name, institution, country of residence and email address. The Anglohaskell wiki page can be found at http://www.haskell.org/haskellwiki/AngloHaskell/2009 There's still room for more talks if anyone's got something they fancy giving a go - everything from cutting edge research to "what I did with Haskell on my holiday" is welcome, as is everyone from GHC implentors to newbies frantically writing Hello World on their laptop in the back row. -- flippa(a)flippac.org

1 0

FFI problem - possibility of CStringLen overflow?
by Robin Green 20 Jul '09

20 Jul '09

While rewriting cautious-file to use ByteStrings and FFI just now, I came across this potential problem: Why does the FFI specification define CStringLen as (Ptr CChar, Int)? As the FFI specification itself notes, Haskell 98 implementors are allowed to have a quite small range for Int (only up to 2^29-1) - so you could in principle have a very large String which cannot be represented as a CStringLen, or a very large CString which cannot be represented as a CStringLen. They would take up wads of memory, so this has probably never happened yet, but it's in principle possible, no? If so, what are the implications of this for code correctness? -- Robin

3 2

ANNOUNCE: graphviz-2999.1.0.1
by Ivan Lazar Miljenovic 20 Jul '09

20 Jul '09

This is a bug-fix release to fix the various Attribute-related problems in the previous release (2999.0.0.0) spotted mainly by Zsolt Dollenstein. Please disregard version 2999.1.0.0; it had a small bug. Similarly, people should avoid using 2999.0.0.0 unless they don't use any Either-based Attributes. The problem (which took me a while to realise) was that whilst I was using Either to denote that an attribute could take one of two different types of values (e.g. Bool and String) and the parsing for that worked. However, since I was using Show (I know it's bad, it was on my TODO, remember?) to generate the Dot code, then whenever one of these values was converted then either Left or Right would appear, and GraphViz doesn't seem to like that for some reason... I also took the opportunity to define custom types for more Attributes (as they only accept certain String values). Furthermore, GraphViz states that attributes that can accept a boolean value are set to True when the attribute is listed on its own; the library now does this even for those Attributes that use a custom type that combines Bool with something else. One other small change is that the Color Attribute now takes in a list of Colors; if you only want one, use a singleton list (it was either that or a custom datatype that takes either a single Color or a list of Colors...). Note that 2999.1.0.0 accidentally only had singleton Color values for the Color Attribute rather than a list, hence the quick version bump. Nothing else has changed in the library apart from the Attributes. Regarding my request in the previous email about the possible new layout of the DotGraph datatype: How many people actually build DotGraphs by hand or pull them apart by hand? If there were getNodes :: DotGraph -> [Node] and getEdges :: DotGraph -> [Edge] functions available, would you accept an all-in-one data type that just contains a generic list of statements and thus follows upstream more closely? Even leaving aside parsing, this would allow for greater flexibility for creating wild and wacky graphs or for a monadic interface or something for building DotGraphs (which someone has asked me for, but which at the moment I have no idea how to do...). -- Ivan Lazar Miljenovic Ivan.Miljenovic(a)gmail.com IvanMiljenovic.wordpress.com

1 0

Typeclass default implementation in subclasses
by Andy Gimblett 20 Jul '09

20 Jul '09

Hi all, This email is literate Haskell. I have a question about default implementations of typeclasses. > {-# LANGUAGE TypeSynonymInstances #-} > module Thing where > import Text.PrettyPrint.HughesPJ Let say I want to pretty-print some values, enclosed in double quotes. The natural thing to do (within the HughesPJ pretty-printing framework, anyway - and that's where I am in this problem's wider context) is: > ppQuote :: Show a => a -> Doc > ppQuote = doubleQuotes . text . show Now, this works nicely for (say) Int: > x :: Int > x = 1 *Thing> ppQuote x "1" But less nicely for String and Char, because their Show instances already insert double/single quotes respectively: > y :: String > y = "hello" > z :: Char > z = 'a' *Thing> ppQuote y ""hello"" *Thing> ppQuote z "'a'" I don't want this. I'd like them to be "hello" and "a" respectively. So I thought I'd create a typeclass, whose default implementation is as above... > class (Show a) => Quotable a where > quote :: a -> Doc > quote = ppQuote ... but with specialised instances for String and Char (the former seems to need the TypeSynonymInstances extension?): > instance Quotable String where > quote = text . show -- don't need the doubleQuotes call for String > instance Quotable Char where > quote c = quote [c] -- just lift it to String Unfortunately, while this works great for String and Char... *Thing> quote y "hello" *Thing> quote z "a" ... the "default implementation" mechanism doesn't work as I'd expect/hope: *Thing> quote x <interactive>:1:0: No instance for (Quotable Int) arising from a use of `quote' at <interactive>:1:0-6 Possible fix: add an instance declaration for (Quotable Int) In the expression: quote x In the definition of `it': it = quote x What I would _like_ would be for the compiler to say "OK, the Quotable class depends on the Show class, and Int is an instance of Show so Int is also an instance of Quotable, having the default implementation (since there isn't a specialised one for it)" - but clearly it doesn't. Please can someone tell me why this doesn't happen, and if there is a way of making it happen? Also, if there's a more sensible way of attacking this whole problem, I'd be curious to hear it. I should perhaps add that this isn't a huge problem for me, because my instances will in practice tend to be String and Char anyway, and one can of course add Quotable instances for anything else easily enough - but I'm curious now I've come this far. :-) Many thanks! -Andy

2 1

SpecConstr difficulties
by Reiner Pope 20 Jul '09

20 Jul '09

Hi everyone, I've been having some trouble getting SpecConstr to work as I want it to. The following example (see end of email) came up in some code I was playing around with. The loops g1 and g2 both compute the same thing: the maximum element of the "list" (which has been fused away) of numbers from 1 to 10. Since 'maximum' is a foldl1 not a foldl, I use a strict Maybe type as an accumulator. The Maybe gets filled after the first element is seen, so the Maybe is a Just for almost the entire running of the loop. It would be good to have this recognised by SpecConstr, to create an optimised loop for the Just case. This does indeed happen for g1, but not for g2. My difficulty is that I am only able to produce code similar to g2, i.e. where the pattern matching is in a separate function, 'expose', because the 'expose' function is implemented in a type-class, like in stream-fusion. Is there some way to keep the SpecConstr while leaving the 'expose' as a separate function? Here is the code: > {-# LANGUAGE BangPatterns #-} > > module Test(ans1,ans2) where > > import Prelude hiding(Maybe(..)) > > data Maybe a = Just !a | Nothing > > Nothing `mapp` Just b = Just b > Just a `mapp` Just b = Just (max a b) > > ans1 = g1 Nothing (0::Int) > > g1 m !n = case m of > Nothing -> if n > 10 then m else g1 (m `mapp` Just n) (n+1) > Just x -> if n > 10 then m else g1 (m `mapp` Just n) (n+1) > > ans2 = g2 Nothing (0::Int) > > g2 m !n = expose m (if n > 10 then m else g2 (m `mapp` Just n) (n+1)) > expose Nothing b = b > expose (Just a) b = a `seq` b On a similar note, when I was having difficulties with this problem, I started to wonder if it would be possible to come up with a more direct way to tell GHC, "do SpecConstr on this variable". From reading the source code of the stream-fusion package, it seems that the current way of doing this is with 'expose' functions like I wrote below. Could we instead have a {-# SPECCONSTR #-} pragma, to be used on function arguments, like: foo {-# SPECCONSTR #-} x y {-# SPECCONSTR #-} z = ... This pragma should say to the GHC something like "ignore conditions H2, H5 and H6 of the SpecConstr paper, for this function and this argument". Cheers, Reiner

1 0