- Haskell - Haskell.org

RE: Dimensional analysis with fundeps
by Andrew Kennedy 11 Apr '01

11 Apr '01

You shouldn't need rational exponents to take square roots as long as no *ground* type requires them. If polymorphism over units were primitive, then we'd have something like sqrt :: Real (u.u) -> Real u for a fixed numeric type Real that's parameterized over its units. (BTW, it's not possible to define such a function in the language using, say, only standard arithmetic operators and comparison; you have to build it in instead). In your encoding, is the following a valid type? sqrt :: (Num rep, Add kg kg kg', Add m m m', Add s s s') => Dimensioned kg' m' s' rep -> Dimensioned kg m s rep Or have I misunderstood multi-parameter classes with functional dependencies? - Andrew. > -----Original Message----- > From: anatoli [mailto:anatoli@yahoo.com] > Sent: Monday, April 09, 2001 5:37 PM > To: haskell(a)haskell.org > Subject: Dimensional analysis with fundeps > > There is a couple of things :) left to make this usable: > ... > > 2) Make it work with rational (not just integer) exponents, > so one can take square roots and the like. > (Can one do GCD in this style, without resorting to > undecidable and/or overlapping instances?); >

2 1

FINAL CALL: 3rd SCOTTISH FUNCTIONAL PROGRAMMING WORKSHOP
by Greg Michaelson 11 Apr '01

11 Apr '01

FINAL CALL FOR PAPERS 3RD SCOTTISH FUNCTIONAL PROGRAMMING WORKSHOP AUGUST 22nd-24th 2001 UNIVERSITY OF STIRLING We are pleased to announce the 3rd Scottish Functional Programming Workshop, to be held this August at the University of Stirling. All participants are expected to present a paper at the Workshop on some contemporary research topic within the broad field of Functional Programming. Participants will be selected on the basis of a 2 page abstract of their paper. All presented papers will appear in the draft proceedings. After the Workshop, a refereed selection of papers will be published as a book 'Trends in Functional Programming 3' by Intellect. Key Dates: 4th May: Submission of 2 page abstract 18th May: Notification of acceptance 20th July: Registration 10th Aug: Submission of paper for draft proceedings 22nd to 24th Aug: Workshop For further details, please see: http://www.cee.hw.ac.uk/~greg/sfp3/ Sharon Curtis | Greg Michaelson| Workshop Coordinators Phil Trinder |

1 0

RE: Profiling burn-rate
by Julian Seward (Intl Vendor) 10 Apr '01

10 Apr '01

| Is there any way to profile the "burn-rate" of a Haskell program? | | Profiling as I understand it tells you what the "live" information | on the heap is. It doesn't tell you what garbage collector has just | freed. So, if a function were generating tons of intermediate values | which it then immediately consumed, you wouldn't see it on | the profile. | Yet this is a place where rewriting the code or adding | strictness could | help reduce time spent garbage collecting. At least in GHC, and, I suspect, for any system with a good generational GC, the above may not really hold. A good Gen GC should get rid of short-lived junk very cheaply, since that's what it's designed to do. My experience is that a more effective way to reduce GC time is to reduce the residency. In other words it's the area bounded by the space-time curve that counts, IMO. J

3 2

Re: Dimensional analysis with fundeps
by John.Velman＠HSC.com 10 Apr '01

10 Apr '01

1) What is a fundep? 2) This is a very interesting topic, and rather complex. It has come up (dimensions, units, -- not Haskell implementations of them) in some recent work on STEP (ISO 10303). I'm only now trying to come up to some speed on Haskell, and only now trying to recall what I used to know about the algebra of units and dimensions, so I'm not really prepared to comment deeply on this suggestion. However, I'd suggest that someone have a good long look at "An Ontology for Engineering Mathematics" (available at http://www-ksl.stanford.edu/KSL_Abstracts/KSL-94-18.html) (or similar work) before building anything in to standard libraries. John Velman Tom Pledger <Tom.Pledger(a)peace.com>@haskell.org on 04/09/2001 02:48:48 PM Sent by: haskell-admin(a)haskell.org To: anatoli <anatoli(a)yahoo.com>, haskell(a)haskell.org cc: Subject: Dimensional analysis with fundeps I like it! : | 3) Allow arbitrary user-defined "fundamental" dimensions | (for things like dollars or radians) -- this may be | very tricky; | | 4) Allow several unit systems (such as SI and Imperial) | to coexist. Some suggestions/quibbles... If you clearly make the type system deal with dimensions rather than units, there's no problem with plugging in multiple unit systems. You just have to pick a scale for the representation. newtype Dimensioned mass length time rep = Dimensioned rep type Mass rep = Dimensioned One Zero Zero rep kg, lb :: Num a => Mass a kg = dm 1 lb = dm 0.4535924 Angles are dimensionless. (Think of the Taylor series for trig functions.) radian, degree :: Unit radian = dm 1 degree = dm (pi/180) Regards, Tom _______________________________________________ Haskell mailing list Haskell(a)haskell.org http://www.haskell.org/mailman/listinfo/haskell

2 1

Dimensional analysis with fundeps
by anatoli 09 Apr '01

09 Apr '01

Hello everybody. This is inspired by some recent threads in comp.lang.functional and comp.object. It seems that fundeps are powerful enough to do compile-time dimensional analysis in Haskell. That is, one can devise types that represent dimensions of quantities, such as (meter / second) or (kilogram * meter / second^2), and have a complete set of type-correct operations on such types. Presented below is my rather crude attempt to devise a proof-of-concept kind of "Dimension" module. ========== Dimension.lhs > module Dimension where Natural numbers expressed as Haskell types. Zero is 0. If type a is Zero or positive, then (Succ a) is positive; if type a is Zero or negative, then (Pred a) is negative. This is very crude and should be replaced with sign-magnitude representation. > data Zero > data Succ n > data Pred n Natural number addition. Both operands are either positive, negative or zero, so we consider 3*3=9 cases. > class Add a b c | a b -> c > instance Add Zero Zero Zero > instance Add Zero (Succ b) (Succ b) > instance Add Zero (Pred b) (Pred b) > instance Add (Succ b) Zero (Succ b) > instance Add (Pred b) Zero (Pred b) > instance Add a b c => Add (Succ a) (Succ b) (Succ (Succ c)) > instance Add a b c => Add (Pred a) (Pred b) (Pred (Pred c)) > instance Add a b c => Add (Succ a) (Pred b) c > instance Add a b c => Add (Pred a) (Succ b) c Subtraction is very similar to addition. > class Sub a b c | a b -> c > instance Sub Zero Zero Zero > instance Sub Zero (Succ b) (Pred b) > instance Sub Zero (Pred b) (Succ b) > instance Sub (Succ b) Zero (Succ b) > instance Sub (Pred b) Zero (Pred b) > instance Sub a b c => Sub (Succ a) (Succ b) c > instance Sub a b c => Sub (Pred a) (Pred b) c > instance Sub a b c => Sub (Succ a) (Pred b) (Succ (Succ c)) > instance Sub a b c => Sub (Pred a) (Succ b) (Pred (Pred c)) A Dimensioned type carries along exponents for three basic dimensions: mass, length, and time. For simplicity we restrict ourselves to SI units. > newtype Dimensioned kg m s rep = Dimensioned rep You can add only identically dimensioned quantities. > plus, minus :: Num rep => > Dimensioned kg m s rep -> > Dimensioned kg m s rep -> > Dimensioned kg m s rep > plus (Dimensioned x) (Dimensioned y) = Dimensioned (x+y) > minus (Dimensioned x) (Dimensioned y) = Dimensioned (x-y) You can multiply any two dimensioned quantities; exponents from the operands add up in the result. > times :: (Num rep, > Add kg kg' kg'', > Add m m' m'', > Add s s' s'') => > Dimensioned kg m s rep -> > Dimensioned kg' m' s' rep -> > Dimensioned kg'' m'' s'' rep > times (Dimensioned x) (Dimensioned y) = Dimensioned (x * y) Similarly for division, except exponents are subtracted. > divby :: (Fractional rep, > Sub kg kg' kg'', > Sub m m' m'', > Sub s s' s'') => > Dimensioned kg m s rep -> > Dimensioned kg' m' s' rep -> > Dimensioned kg'' m'' s'' rep > divby (Dimensioned x) (Dimensioned y) = Dimensioned (x / y) Some handy abbreviations. > type One = Succ Zero > type Kg = Dimensioned One Zero Zero Double > type Meter = Dimensioned Zero One Zero Double > type Second = Dimensioned Zero Zero One Double > type Unit = Dimensioned Zero Zero Zero Double > dm x = Dimensioned x > kg = (dm 1) :: Kg > m = (dm 1) :: Meter > s = (dm 1) :: Second > u = (dm 1) :: Unit ==== End of Dimension.lhs So, now one can do this: (kg `times` m) `plus` (m `times` kg) which has type Dimensioned (Succ Zero) (Succ Zero) Zero Double but not this: kg `plus` (m `times` m) The latter would produce a (rather cryptic, to be fair) type error: ERROR - Constraints are not consistent with functional dependency *** Constraint : Add Zero Zero (Succ Zero) *** And constraint : Add Zero Zero Zero *** For class : Add a b c *** Break dependency : b a -> c There is a couple of things :) left to make this usable: 1) Define obvious classes and instances to make standard operations like (+) and (*) work intuitively. Unfortunately this means hiding standard Num and Fractional classes; 2) Make it work with rational (not just integer) exponents, so one can take square roots and the like. (Can one do GCD in this style, without resorting to undecidable and/or overlapping instances?); 3) Allow arbitrary user-defined "fundamental" dimensions (for things like dollars or radians) -- this may be very tricky; 4) Allow several unit systems (such as SI and Imperial) to coexist. I've done most of these things in C++, and others did too, so I guess there's some demand for such stuff. Even if no one wants it, a nice example of what a powerful type system can do is often handy in religious debates :) -- anatoli t. __________________________________________________ Do You Yahoo!? Get email at your own domain with Yahoo! Mail. http://personal.mail.yahoo.com/

3 2

Re: Dimensional analysis with fundeps
by Ashley Yakeley 09 Apr '01

09 Apr '01

At 2001-04-09 09:36, anatoli wrote: >4) Allow several unit systems (such as SI and Imperial) > to coexist. That's easy. Just hide the 'Dimensioned' constructor so no-one can see what system you're using. -- type One = Succ Zero type Mass rep = Dimensioned One Zero Zero rep type Length rep = Dimensioned Zero One Zero rep type Time rep = Dimensioned Zero Zero One rep type Unit rep = Dimensioned Zero Zero Zero rep dm x = Dimensioned x kilogramme = (dm 1) :: (Num rep) => Mass rep metre = (dm 1) :: (Num rep) => Length rep second = (dm 1) :: (Num rep) => Time rep unit = (dm 1) :: (Num rep) => Unit rep toUnit :: rep -> Dimensioned Zero Zero Zero rep toUnit = dm fromUnit :: Dimensioned Zero Zero Zero rep -> rep fromUnit (dm x) = x -- ...which allows another module to do this... -- inch = times (toUnit 0.0254) metre someLength = plus (times (toUnit 5.1) metre) (times (toUnit 3.7) inch) someLengthInches = fromUnit (divby someLength inch) -- -- Ashley Yakeley, Seattle WA

1 0

Re: Dimensional analysis with fundeps
by Ashley Yakeley 09 Apr '01

09 Apr '01

At 2001-04-09 09:36, anatoli wrote: >It seems that fundeps are powerful enough to do >compile-time dimensional analysis in Haskell. I'm very impressed and slightly frightened. And there was me complaining that since you can't parameterise types in Haskell by integer (or other values) the way you can with C++ templates, you can't do strongly-typed dimensions in Haskell... -- Ashley Yakeley, Seattle WA

1 0

Fast Reading from File
by Saswat Anand 09 Apr '01

09 Apr '01

I have a need to read large lists of numbers from a file and operate on them. I tried the following: read' =do s <- readFile "temp" print $ sum $ map (fst.head.readDec)(lines s) But this is quite slow even for 1000 numbers. Are there any other ways to do it faster? Will it help if a C function does the reading and Haskell calls that. I have to use Hugs. Thanks, Saswat

1 0

IO recursion
by mapple apple 09 Apr '01

09 Apr '01

Hi everyone, I'm not sure this is the place to ask, coz I hvn't join your mailing list. But anyway .. Here's the case .. does anybody know 'bout IO recursion? ie. We've a file and we want to put a numberline 1: 'first line' 2: 'second line' until n, where n is the last line I know how to convert a file into a huge string using 'readFile' and I'm using the 'lines function' in the prelude. That means i have a huge string like "\nHi mate\nHow r u?" using the helper function 'lines' i can do that by saying : main :: IO () main = do putStr "Filename plz: " file <- getLine content <- readFile file putStr ("1: " ++ read (show (head (lines content)))) and so on that's not recursive function and will not work for all cases. So please .. if anyone knows how to do those things, let me know. one more thing, does anybody know how to join the mailing list?? mapple apple mapple_83(a)hotmail.com _________________________________________________________________________ Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.

3 2

A GUI toolkit looking for a friend
by luc 08 Apr '01

08 Apr '01

<disclaimer> im very rookie in Haskell </disclaimer> i d like to have some GUI for an application (widget and caselike drawing), for either windows an unix. if i summarized todays options : - franttk and fudget are not either widget and (unix and windows ) enabled. -tcl haskell works on both win and unix, and has widgets, but is a tcl wrapper, and as such, is not "functiunal" enough. 1) am i understanding the situtation correctly ? 2) is none of the high level ones are "The right way to do it", do we know why ? any kind of positive criticism on that ? comparison ? 3) how complex would it be to take, say the fudget approach, and build it on top of tcl haskell ? any obvious incompatibilities ? 4) or extend frantk with widgets ? on top I would add: none of these are up to the level of fancy we expect of today's gui. tcl is nice, but a bit primitive looking , isnt it ? (no flame intended) 5)anybody ever considered other toolkit, like -Qt ( www.trolltech.com , true multplatform, now OSS, base of KDE), - or the young PARAGUI (based on SDL, the OPengl gaming multiplatform (high performance) http://www.bms-austria.com/projects/paragui/screenshots.html ) ? 6) how complex, given a C++ interface, to "automate" a wrapper around that ? can Hdirect, do that , or others ? (note. I meant C++, not C. )

2 1