- Haskell-Cafe - Haskell.org

GHC plugin: custom wired-in package, module, and type class
by J. Stutterheim 19 Nov '16

19 Nov '16

Hi all, I'm looking for a way to add my own wired-in package/module/type class for use in a core-to-core plugin. My first challenge is to identify the type class that I want to find. I have the following code: myPkgUnitId :: UnitId myPkgUnitId = fsToUnitId (fsLit "mypackage") mkMyModule :: FastString -> Module mkMyModule m = mkModule myPkgUnitId (mkModuleNameFS m) myModule :: Module myModule = mkMyModule (fsLit "My.Module") myClassName :: Name myClassName = clsQual myModule (fsLit "MyClass") myClassKey myClassKey :: Unique -- How do I get this? Here the missing piece is the myClassKey. Looking at how this is done for, e.g., the Monad class: monadClassKey :: Unique monadClassKey = mkPreludeClassUnique 8 we see that it has a hard-coded unique identifier. Is there a way to do the same for my own type class? Or to somehow find the Unique key dynamically? The second challenge is more or less the same in that I want to identify a member of MyClass: myFunMName :: Name myFunMName = varQual myModule (fsLit "myFun") myFunMClassOpKey myFunMClassOpKey :: Unique -- How do I get this? Again, the identifier is hard-coded for, e.g., ">>=". Any insights will be much appreciated. Best regards, Jurriën

2 1

4.4.2 Fixity Declarations - and what happens when they are missing
by Johannes Waldmann 18 Nov '16

18 Nov '16

Yeah, I know, it's bikeshedding, some 30 years after the fact, but - How on Earth was this allowed to happen? "Any operator lacking a fixity declaration is assumed to be infixl 9" https://www.haskell.org/onlinereport/haskell2010/haskellch4.html#x10-820061 If I don't specify associativity or precedence for an operator, then the parser should simply reject any expression whose meaning would depend on such declarations. So I'd have to write parentheses at each (ambiguous) use site - or write a fixity declaration. Both would serve to document the code. The current situation feels like "the type of identifiers without type declaration is int". Short of repairing the standard - I'd welcome GHC to print a huge warning for each definition and use (!) of "operator without fixity decl". - J. PS: to show that I did at least some research * there's no mention of this in 4.2 "Infix operators" in http://research.microsoft.com/~simonpj/papers/history-of-haskell/history.pdf * Scala also has default rules (6.12.3) "The precedence of an infix operator is determined by the operator’s first character." "Operators ending in a colon ‘ : ’ are right-associative. All other operators are left-associative." and why do I care at all? This: https://github.com/jwaldmann/haskell-obdd/issues/1

1 0

Preventing dereferencing an STRef after taking an action
by Timotej Tomandl 18 Nov '16

18 Nov '16

Hi, is it possible to prevent an STRef passed to one ST action from being passed further on to another ST action? I am trying to prevent these usages: invalid :: STRef s a -> STRef s b -> ST s () invalid x y=action1 x y >>=action2 x y invalid1 :: STRef s a -> STRef s b -> ST s () invalid1 x y=action1 x y >>=action2 y while still being able to do this: valid :: STRef s a -> STRef s b -> ST s () valid x y=action1 x y >>=action2 x where action1 will modify x and y inplace, but in the end only x will hold the useful information. In other words I am trying to make it impossible for any action after action1 to have a same argument as action1's second argument. timo

1 0

Re: [Haskell-cafe] commenting out a line of code
by Bernhard Herzog 18 Nov '16

18 Nov '16

On 15.11.2016, David Turner wrote: > The outer do block should be desugared first, and the `putStrLn "done"` has > the same indent as the `if` so it gets a semicolon, in that block, meaning > it looks like this: > > main = do > { if True > then do > putStrLn "A" > else do > -- putStrLn "B" > ; putStrLn "done" > } That's what one would expect. Except that GHC treats the 'putStrLn "done"' as part of the do-block in the else branch by default. This is due to the NondecreasingIndentation extension which is enabled by default (see https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/bugs.html#c…) With the option -XNoNondecreasingIndentation (or -XHaskell2010) GHC does produce an error message because of the empty do-block. Bernhard

2 1

[ANN] Ginger: Jinja-like HTML templates for Haskell
by Tobias Dammers 17 Nov '16

17 Nov '16

Dear Cafe, I've been working on this for a while now, and figured now is as good a time as any to announce it properly, so here it is. https://ginger.tobiasdammers.nl/ Ginger is a Haskell implementation of Jinja (http://jinja.pocoo.org/) the best-of-breed in the "DOM-agnostic HTML template engines for Python" arena (or at least I think it is). The reason I built this is because upon researching HTML templating solutions for Haskell, I noticed that almost all of them are typesafe compile-time solutions (either EDSLs or something TH/QQ), which is awesome but doesn't fit my use cases. I believe I am not alone here. Code is on github: https://github.com/tdammers/ginger And also on bitbucket: https://bitbucket.org/tdammers/ginger And of course Ginger is available from hackage: http://ginger.tobiasdammers.nl/ Any kind of feedback is very welcome. Thanks for your time! - Tobias -- Tobias Dammers - tdammers(a)gmail.com

3 3

CFP: Dutch Functional Programming Day 2017
by J. Stutterheim 17 Nov '16

17 Nov '16

Call for Participation: Dear Friends of Functional Programming, The next Dutch FP Day will be held on Friday, January 6 2017 at the Radboud University in Nijmegen. You are invited to some or all of: participate, give a presentation, join the dinner. For further information, please see the website. https://wiki.clean.cs.ru.nl/NL-FP_dag_2017 Feel free to forward this invitation to colleagues and students who might be interested. Best regards on behalf of Markus Klinik

1 0

commenting out a line of code
by Johannes Waldmann 15 Nov '16

15 Nov '16

Dear Cafe - Sometimes, a comment can change the meaning of a program. No? Well, but "commenting out" a line of code certainly does. So, can you guess what this program will output? main = do if True then do putStrLn "A" else do -- putStrLn "B" putStrLn "done" I nominate this for inclusion in a hypothetical Haskell equivalent of http://www.javapuzzlers.com/ - J.

6 5

.haddock files
by Patrick Pelletier 15 Nov '16

15 Nov '16

Where can I find documentation about .haddock files? What is in them? Is it possible to generate a .haddock file from source code and then later produce HTML documentation from the .haddock file? Or is that not what it's for? The documentation for Haddock doesn't even seem to mention .haddock files, unless I'm missing something. I tried subscribing to the Haddock mailing list so I could post this question there, but at http://projects.haskell.org/cgi-bin/mailman/subscribe/haddock I got the message: "We're sorry, we hit a bug! Mailman experienced a very low level failure and could not even generate a useful traceback for you. Please report this to the Mailman administrator at this site." Anybody know what's going on with that? Thanks, --Patrick

2 1

bizarre behaviour
by Lawrence Wickert 15 Nov '16

15 Nov '16

Hi, Ok, I am just learning Haskell. So, every day I open a text editor, a pdf file of a Haskell book and a terminal to run ghci. I resize the windows and go on my merry way. Yesterday I decided to install this script https://raw.githubusercontent.com/hotice/webupd8/master/session with the object to save my session windows and restore them the next day. Worked fine although I had to authorize some keyring popup box before it would work. Now I opened Google Chrome and an iHeart radio link. At this point the following appeared at the ghci prompt: Prelude> Vector smash protection is enabled. ghci continues to run fine. If I close and then re-open the iHeart link the message reappears. I am using Ububtu 16.04LTS and I don't use drugs or alcohol. Any ideas as to what is going on? Larry

2 1

Re: [Haskell-cafe] Named arguments to functions?
by Sylvain Henry 14 Nov '16

14 Nov '16

Hi Saurabh, Implicit parameters already look like named arguments: > :set -XImplicitParams > let f = ?x + ?y > let { ?y = 5; ?x = 10 } in f 15 > let g = f where { ?x = 5; ?y = 10 } > g 15 Perhaps we could extend TypeApplications to support the following more natural syntax: > f @(?y = 5) @(?x = 10) which would be syntaxic sugar for the code above. Sylvain On 13/11/2016 14:38, Saurabh Nanda wrote: > Thank you for your reply. > > If you are interested in something like this, you may want to > start a discussion of the design on one of the other lists, or > perhaps write up a proposal for GHC. > > > While I would like the general area of records getting better in > Haskell, I don't think I have the necessary technical chops to lead a > GHC proposal around this. Do you know any proposal/extension in GHC > that is already trying to solve: (a) records, and/or (b) named arguments? > > The proposals currently under discussion are on the `github` URL. > > > Possible to update the prime.haskell.org <http://prime.haskell.org> > website to reflect the current state of affairs? > > -- Saurabh. > > > On 10/11/2016 22:59, Iavor Diatchki wrote: >> Hello, >> >> I would say that this is a too complex of a change to discuss for >> Haskell Prime without implementation or prior experience. If you >> are interested in something like this, you may want to start a >> discussion of the design on one of the other lists, or perhaps write >> up a proposal for GHC. >> >> The proposals currently under discussion are on the `github` URL. >> >> Cheers, >> -Iavor >> >> >> >> >> >> On Thu, Nov 10, 2016 at 2:16 AM, Saurabh Nanda >> <saurabhnanda(a)gmail.com <mailto:saurabhnanda@gmail.com>> wrote: >> >> Hi Everyone, >> >> I was going through >> https://prime.haskell.org/wiki/Process#Proposals >> <https://prime.haskell.org/wiki/Process#Proposals> and noticed >> the following: >> >> > Ideally the language change should be implemented already, so >> that experience gained with the implementation can inform the >> discussion >> >> Is this a hard requirement? I would like to see the ability to >> pass named arguments to functions somehow (most probably through >> anonymous records). I'm unaware of any existing language >> extension that solves this problem. Does this render the >> discussion out-of-bounds from a Haskell Prime perspective? Has >> this already been proposed and considered? >> >> Also, going by what is given on the Wiki, I'm confused about >> where to check the current status of proposals that are being >> considered. Which is the primary link: >> >> * >> https://prime.haskell.org/query?status=new&status=assigned&status=reopened&… >> <https://prime.haskell.org/query?status=new&status=assigned&status=reopened&…> >> * Or, https://github.com/haskell/rfcs >> <https://github.com/haskell/rfcs> >> >> -- Saurabh. >> >> _______________________________________________ >> Haskell-prime mailing list >> Haskell-prime(a)haskell.org <mailto:Haskell-prime@haskell.org> >> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-prime >> <http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-prime> >> >>

1 0