February 2009 - Haskell-Cafe

[Colin Paul Adams] Re: [Haskell-cafe] Ready for testing: Unicode support for Handle I/O
by Colin Paul Adams 03 Feb '09

03 Feb '09

>>>>> "John" == John Goerzen <jgoerzen(a)complete.org> writes: >> latin1, utf8, utf16, utf16le, utf16be, utf32, utf32le, utf32be, >> localeEncoding, John> Will there also be something to handle the UTF-16 BOM John> marker? I'm not sure what the best API for that is, since John> it may or may not be present, but it should be considered -- John> and could perhaps help autodetect encoding. Not really. The UTF-16 BOM is only allowed for utf16. Utf16le and utf16be do not have a BOM. What it does is alloww you to auto-detect the endianness if utf16 has been requested as the encoding (if it is not present the big-endian is assumed). Ditto for utf32/utf32be/utf32le. A BOM is thesedays allowed (but deprecated - it should never have been allowed and originally wasn't) for utf8 also. But it has no value other than to introduce ambiguity. :-( -- Colin Adams Preston Lancashire

2 2

Verifying Haskell Programs
by Paulo J. Matos 03 Feb '09

03 Feb '09

Hi all, Much is talked that Haskell, since it is purely functional is easier to be verified. However, most of the research I have seen in software verification (either through model checking or theorem proving) targets C/C++ or subsets of these. What's the state of the art of automatically verifying properties of programs written in Haskell? Cheers, -- Paulo Jorge Matos - pocmatos at gmail.com Webpage: http://www.personal.soton.ac.uk/pocm

7 10

type and data constructors in CT
by Gregg Reynolds 03 Feb '09

03 Feb '09

Hi, I think I've finally figured out what a monad is, but there's one thing I haven't seen addressed in category theory stuff I've found online. That is the relation between type constructors and data constructors. As I understand it, a type constructor Tcon a is basically the object component of a functor T that maps one Haskell type to another. Haskell types are construed as the objects of category "HaskellType". I think that's a pretty straightforward interpretation of the CT definition of functor. But a data constructor Dcon a is an /element/ mapping taking elements (values) of one type to elements of another type. So it too can be construed as a functor, if each type itself is construed as a category. So this gives us two functors, but they operate on different things, and I don't see how to get from one to the other in CT terms. Or rather, they're obviously related, but I don't see how to express that relation formally. If somebody could point me in the right direction I'd be grateful. Might even write a tutorial. Can't have too many of those. Thanks, Gregg

5 13

FP simulators for real-time systems?
by Lee Pike 03 Feb '09

03 Feb '09

Hello, I'm interested to hear if anyone out there has used Haskell (or other functional languages for that matter) to build simulators for real- time systems. I'm somewhat familiar with Timber <http://www.timber-lang.org/> and similar languages for actually constructing real-time systems. However, I'm more interested in documented uses of FP "out-of-the- box" to build simulators and associated test harnesses. Pointers to published papers are particularly appreciated. Thanks! Lee

3 2

Cabal and more than one version
by Jason Dusek 03 Feb '09

03 Feb '09

I'd like to be able to do something like: if (template-haskell < 2.3) cpp-options: -D TH_THE_YOUNGER else cpp-options: -D TH_THE_ELDER I guess this kind of thing is not possible at present? -- _jsn

3 9

Re: 1,000 packages, so let's build a few!
by John Lato 03 Feb '09

03 Feb '09

Duncan Coutts wrote: > > Some are trivial and should be done away with. For example the ones that > just check if a C header / lib is present are unnecessary (and typically > do not work correctly). The next point release of Cabal can do these > checks automatically, eg: > > Configuring foo-1.0... > cabal: Missing dependencies on foreign libraries: > * Missing header file: foo.h > * Missing C libraries: foo, bar, baz > This problem can usually be solved by installing the system > packages that provide these libraries (you may need the "-dev" > versions). If the libraries are already installed but in a > non-standard location then you can use the flags > --extra-include-dirs= and --extra-lib-dirs= to specify where > they are. Thank you! Thank you! Thank you! For those of us who want to write cross-platform (i.e. Windows) bindings to C libraries, this is great news. John Lato

4 4

Re: type metaphysics
by oleg＠okmij.org 03 Feb '09

03 Feb '09

I believe the original notion of type by Russell is most insightful, bridging the semantic notion of type (type as a set of values) and the syntactic notion (type system as a syntactic discipline, a statically decidable restriction on terms). That point is discussed at some length in Sec 3 (pp. 7-8) of http://okmij.org/ftp/Computation/FLOLAC/lecture.pdf The one of the main points of these FLOLAC lecture notes was to justify and illustrate the claim that A type is a (quickly decidable) approximation of a dynamic behavior that can be derived from the form of an expression. (The other claim was the relationship between sharing and polymorphism). Incidentally, Russell has published his paper that introduced types as we in Computer Science know them in 1908. Last year there was 100th anniversary of types. I have been trying to prompt people to have a party and celebrate the occasion -- alas, to no avail.

2 1

Parallel term reduction
by John D. Ramsdell 02 Feb '09

02 Feb '09

I have a reduction system in which a rule takes a term and returns a set of terms. The reduction system creates a tree that originates at a starting value called the root. For most problems, the reduction system terminates, but a step count limit protects from non-termination. Rule application is expensive, so it is essential that a rule is never applied to the same problem twice. This check makes my program sequential, in that parallel annotations don't improve performance on SMPs. There isn't even an obvious place to add them in my program, at least not to me. How do people write parallel reduction systems that avoid redundant rule application? John > module Main (main) where > import System.Time (ClockTime(..), getClockTime > data Eq a => Item a > = Item { item :: a, > parent :: Maybe (Item a) } > instance Eq a => Eq (Item a) where > x == y = item x == item y The reduction system takes a rule, a step count, and an initial value, and computes a tree of reductions. The order of the items in the returned list is irrelevant, because the tree is assembled as a post processing step. > reduce :: (Eq a, Monad m) => (a -> [a]) -> Int -> a -> m [Item a] > reduce rule limit root = > step rule limit [top] [top] > where > top = Item { item = root, parent = Nothing } step rule limit seen todo, where seen in the items already seen, and todo is the items on the queue. > step :: (Eq a, Monad m) => (a -> [a]) -> Int -> > [Item a] -> [Item a] -> m [Item a] > step _ limit _ _ > | limit <= 0 = fail "Step limit exceeded" > step _ _ seen [] = return seen > step rule limit seen (next : rest) = > loop seen rest children > where > children = map child (rule (item next)) > child i = Item { item = i, parent = Just next } > loop seen rest [] = > step rule (limit - 1) seen rest > loop seen rest (kid : kids) = > if elem kid seen then > loop seen rest kids > else > loop (kid : seen) (rest ++ [kid]) kids A silly rule > rule :: Int -> [Int] > rule n = filter (>= 0) [n - 1, n - 2, n - 3] > secDiff :: ClockTime -> ClockTime -> Float > secDiff (TOD secs1 psecs1) (TOD secs2 psecs2) > = fromInteger (psecs2 - psecs1) / 1e12 + fromInteger (secs2 - secs1) > main :: IO () > main = > do > t0 <- getClockTime > ns <- reduce rule 20000 5000 > t1 <- getClockTime > putStrLn $ "length: " ++ show (length ns) > putStrLn $ "time: " ++ show (secDiff t0 t1) ++ " seconds" The makefile ------------ PROG = reduce GHCFLAGS = -Wall -fno-warn-name-shadowing -O %: %.lhs ghc $(GHCFLAGS) -o $@ $< all: $(PROG) clean: -rm *.o *.hi $(PROG)

2 4

why typeRepArgs (typeOf "hello") is [Char] ?
by minh thu 02 Feb '09

02 Feb '09

Hello, With Data.Typeable : *Graph> typeRepArgs (typeOf 1) [] *Graph> typeRepArgs (typeOf 'a') [] *Graph> typeRepArgs (typeOf True) [] *Graph> typeRepArgs (typeOf "hello") [Char] I don't understand why the latter is not []. Could someone explain it ? Thank you, Thu

4 6

ANN: HDBC v2.0 now available
by John Goerzen 02 Feb '09

02 Feb '09

Hi everyone, I'm pleased to announce that HDBC v2.0 is now available. Simultaneously, HDBC-sqlite3, HDBC-postgresql, and HDBC-odbc v2.0 have also been uploaded. All may be found from Hackage, or at software.complete.org. A guide to new features and migration can be found at: http://software.complete.org/software/wiki/hdbc/NewIn20 A partial summary of it follows: New Features: * Pervasive Unicode support now part of the HDBC API and implemented by all backend drivers. Based on utf8-string and conversions between UTF-8 ByteStrings and Haskell Strings. * Full support for GHC 6.10 is new. Full support for GHC 6.8 is maintained. * Native support for the new exception handling in GHC 6.10, with CPP conditional compilation to continue supporting the old-style exceptions for users of GHC 6.8 or Hugs. * HDBC now has direct support for marshalling virtually all of the Data.Time.* types to and from SQL database date/time types. toSql and fromSql/safeFromSql now support conversions between Data.Time.* types and SqlValue. Conversions between System.Time.* types were present from the beginning and will remain. Automatic conversions are, of course, contingent on support in database engines and HDBC database backend drivers and may not be possible with all databases. * Major expansion of SqlValue to be able to carry date/time data to/from databases in terms of Data.Time.* types. + The old SqlTimeDiff and SqlEpochTime types continue to be present, but are deprecated and will be removed in a future. toSql no longer converts anything to them. The System.Time.* types are converted to the new Data.Time-based SqlValues. It should be exceptionally rare that any user code would use these types directly, but they are being retained in this version out of an abundance of caution. * Replace fromSql with safeFromSql, which returns Left on impossible conversions instead of raising an exception. + Wrote a new fromSql that behaves just like the old one, but is implemented in terms of safeFromSql. It converts the safeFromSql errors to exceptions. + The new structure made it possible to greatly enhance many error messages throughout the conversion process instances. + Removed SqlType typeclass. Re-implemented SqlValue conversions in terms of the generic infrastructure in the convertible package. Simplified code significantly due to this. * Lots of additional docs. * New skeleton test suite for HDBC itself (to be expanded upon later). Richer Date and Time Support HDBC 1.x had only two ways of getting dates and times between Haskell and a database: an epoch (an Integer representing the number of seconds since midnight Jan 1 1970 UTC) and an Integer representing a TimeDiff (a count of elapsed time). While this approach could represent a precise instant in time, it was not always the most helpful. Most notably, you might sometimes want to load date information without time, time without date, or record what timezone the time was measured within. HDBC can now convert all of these things with much greater precision. Thanks to Phil Wise for adding the first Unicode support (to HDBC-postgresql originally) and to Alson Kemp for adding the first advanced date/time marshalling support (also to HDBC-postgresql originally). The experience of adding these features to a single backend driver first made clear a good way to add them to HDBC API as a whole. -- John

8 26