- Haskell-Cafe - Haskell.org

Can we write a parser with both name resolution and error handling by using Happy?
by Tsuyoshi Ito 30 Apr '12

30 Apr '12

Hello, I have a question about Happy, the parser generator. Section 2.1 of Happy User Guide explains a neat trick to perform the name resolution inside a parser by turning the semantic values of nonterminal symbols into Reader monads. http://www.haskell.org/happy/doc/html/sec-using.html#sec-other-datatypes (The example uses bare functions instead of Reader monads, but this difference is insignificant.) A compilable Calc.y file is available from https://gist.github.com/2554028. The important part of the example is the following: Exp : let var '=' Exp in Exp { \p -> $6 (($2, $4 p) : p) } | Exp1 { $1 } Exp1 : ... Term : ... Factor : int { \p -> $1 } | var { \p -> case lookup $1 p of Nothing -> error "no var" Just i -> i } | '(' Exp ')' { $2 } The example code calls `error` on the failure in the name resolution. Although this is fine as an example, it is not good in an actual parser because that means that I have to run the parser in IO monad to catch a parse error. I would like to handle this error just like a parse error by using a monadic parser as explained in Section 2.5.1. http://www.haskell.org/happy/doc/html/sec-monads.html#sec-exception However, I cannot make this work. I tried the following approaches: 1. Just combine these two techniques. That is, write `%monad { E } { thenE } { returnE }` in .y file, where `E t` is equivalent to `Either String t` as defined in Section 2.5.1, and make the semantic value a Reader monad. This does not achieve what I want: now Happy expects that the semantic expression of the form {% ... } will have type `E (Env -> t)` (or `E (Reader Env t)`), but I want to write an expression of type `Env -> E t` (or `ReaderT Env E t`) because whether name resolusion succeeds or not depends on the environment. See https://gist.github.com/2554041 for a code (which does not compile). 2. Write `%monad { P } { thenP } { returnP }`, where `P t` is equivalent to `Env -> E t` or `ReaderT Env E t` and `thenP` and `returnP` are defined accordingly. Now I cannot write the semantic expression for the production rule `Exp : let var '=' Exp in Exp`, because the semantic values of children ($2, $4, and $6) accessible in the `{ ... }` or `{% ... }` part is no longer Reader monads but plain values, meaning that I cannot evaluate a subexpression ($6) using a different environment from the current one. Moreover, this feels wrong because not every symbol requires the environment to evaluate. See https://gist.github.com/2554044 for a code (which does not run because it contains `undefined`). 3. Maybe I cannot use the `%monad` support for this purpose, so shall I give up using it and just use `Env -> E t` as a type of semantic values? This is cumbersome at best because now in each production rule, I have to handle the case where the semantic values of some of the children are `Nothing` (after giving an environment). But more importantly, this does not work because Happy expects the `parseError` function to have type `forall a. [Token] -> a` (which is understandable), which means that it cannot return anything like `Nothing`, defeating the purpose of the use of a monadic parser for error handling. See https://gist.github.com/2554122 for a code (which does not run). >From this, it does not seem that name resolution can be performed inside a parser if we need any error handling better than calling `error`. Is this really the case? Because the combination of parsing and name resolution is very natural (which is why the example in Section 2.1 is so appealing), it is hard to believe that it cannot be done nicely. Am I missing something? In case it matters, I am using Haskell Platform 2011.4.0.0 on Windows. (Just in case, I am aware that we can split parsing and name resolution, in which case a parser generates an abstract syntax tree (AST) with unresolved names and a name resolver converts it to an AST with resolved names. But is that a recommended way to use Happy? If so, the example in Section 2.1 is misleading at best.) Best regards, Tsuyoshi

1 0

Announce: HaTeX 3.3
by Daniel Díaz Casanueva 29 Apr '12

29 Apr '12

Greetings Haskellers! A new version of HaTeX is ready to use in Hackage: http://hackage.haskell.org/package/HaTeX-3.3 == Description == HaTeX is a heavily maintained LaTeX code parser/generator. Write LaTeX documents with all the advantages you already have in Haskell (recursion, type system, high order functions, ...), create a LaTeX backend for your own program, make analysis of LaTeX code through its Abstract Syntax Tree (AST), find another way to pretty-printing values, ... == Release notes == Please, to read the release notes go to the following blog post: http://deltadiaz.blogspot.com.es/2012/04/hatex-33-release-notes.html Thanks and good luck, Daniel Díaz.

1 0

Cabal try to install a package I already have
by Daniel Díaz Casanueva 29 Apr '12

29 Apr '12

Hi Cafe, today I opened the command prompt and typed "cabal install ftphs". Then I read: > Resolving dependencies... > Configuring network-2.3.0.5... And I started to feel confused. Because I already have installed that package with that version! I continued reading: > cabal: The package has a './configure' script. This requires a Unix > compatibility toolchain such as MinGW+MSYS or Cygwin. Yes, I knew this would happen! (thought I don't like it with a so common-used library) But I can't understand why cabal is trying to install network-2.3.0.5. If I type "ghc-pkg list" I can see precisely network-2.3.0.5 is installed. In fact, 2.3.0.5 is the version that comes with the Haskell Platform I recently installed (2011.4.0.0) as you can see here: http://code.haskell.org/haskell-platform/haskell-platform.cabal. Could someone, please, explain me this? I will be very glad. I was working with ghc-7.4 but I needed the network package so I switched to the Haskell Platform, because install network in Windows is painful. Yes, I read the Mitchell guide (found in http://neilmitchell.blogspot.com.es/2010/12/installing-haskell-network-libr…) but once I installed the library I was having compile time error with "unknown symbol" errors. I felt very odd so I switched this machine to the Haskell Platform. I hope someone could help me. Thanks in advance, Daniel Díaz.

2 3

Embed Haskell
by Rosario Borda 29 Apr '12

29 Apr '12

Hi All, Is there a simple method for embed haskell in other languages, like latex? An example in quasiquotation notation: \documentclass{article} \title{Hello World} \begin{document} \maketitle Hello world! and reversed: [haskell: putStrLn reverse "Hello, World!"] \end{document} Many thanks, :) Rosario

3 2

Automatic discovery of tests
by Graham Berks 28 Apr '12

28 Apr '12

Hi, trying to learn haskell and the associated build env. I come from a java/scala world so trying to understand the how to invoke tests. Have a cabal file that invokes a main function to execute a test. But wondering if there is any automatic discovery of tests as per junit. I have seen test-framework-th but to me that implies putting a testGroupGenerator into each module. Am i correct about that and if so is there an example of calling multiple modules tests ? Or is there a better way just to invoke specific functions prefixed with case_ prop_ etc in the entire src/test directory of the cabal build ? Thanks

2 1

Update of haskell cafe on google groups
by Graham Berks 28 Apr '12

28 Apr '12

Has this stopped working ? Seems a bit out of date ? Thanks

1 0

Re: [Haskell-cafe] [Haskell] ANNOUNCE: planar-graph-1.0
by Ivan Lazar Miljenovic 28 Apr '12

28 Apr '12

On 28 April 2012 11:34, Christian Höner zu Siederdissen <choener(a)tbi.univie.ac.at> wrote: > Hi, > > one minor correction: >> This probably won't be of many use to people > for bioinformatics I'd disagree ;-) I'll take a look at your library. > > (for example, RNA /secondary/ structure forms a planar graph) I stand corrected! -- Ivan Lazar Miljenovic Ivan.Miljenovic(a)gmail.com http://IvanMiljenovic.wordpress.com

1 0

ANNOUNCE: planar-graph-1.0
by Ivan Lazar Miljenovic 28 Apr '12

28 Apr '12

I uploaded this [1] yesterday, posted the blog article [2] about it... but forgot to send a message to the lists! [1]: http://hackage.haskell.org/package/planar-graph [2]: http://ivanmiljenovic.wordpress.com/2012/04/27/announcing-planar-graph/ planar-graph is an implementation of, strangely enough, planar graphs (that is, a graph that contains an embedding on a surface, can be drawn with no edge crossings and has a specific ordering of edges). It handles graphs on planes and spheres, but I'm not sure about other surfaces (and there seems to be little demand for such). This probably won't be of many use to people, but as I described in the blog post, I've been using this as a test bed for graph library design (specifically usage of abstract node/edge identifiers, using half-edges and the serialisation/encoding setup). -- Ivan Lazar Miljenovic Ivan.Miljenovic(a)gmail.com http://IvanMiljenovic.wordpress.com

3 3

Re: [Haskell-cafe] [Haskell] ANN: Portackage - a hackage portal
by Andrew Seniuk 27 Apr '12

27 Apr '12

Hi Derek, Thanks for your comments. Yes, I am probably going to focus on making it work as a web service and stay away from platform-specific mobile apps. Not been thinking mobile much at all, since the quantities of data are taxing even to my laptop, and also I'm inclined to believe people are mostly at a laptop or desktop when coding Haskell. However, I'm planning to try server-side sorting and filtering, and compare performance. No DB is used on the server yet, but it would be simple enough I suppose -- all this is quite new territory for me. Also my shared hosting resource limits may become a problem if this gets a lot of use. On the other hand, was planning to use persistence (cookies i guess?) to save the data on the client, so that patches based on Hackage RSS could be applied incrementally. Also, I think the sorting could be faster by sending per-column lists of indices. But maybe it is just the browser rendering which is taking all the time? I'll take this opportunity to announce a few improvements: - shift-click a row to see the Hackage page (bare click does nothing) - per-column filtering widgets Presently working on getting the "last updated" column, this information is not in a .cabal file but I'm hoping it is in the TAR metadata... The link again: http://www.fremissant.net/portackage A.

1 0

Uploading a new hsc2hs
by Thomas DuBuisson 27 Apr '12

27 Apr '12

Warning: I, not the maintainer of hsc2hs, will be uploading a trivial fix for hsc2hs to hackage (new build deps). Even after public attempts to contact anyone in charge of hsc2hs (last January) there still has been no word. Speak now or forever hold your peace. Cheers, Thomas P.S. I still think c2hs is the right philosophy even if it's too verbose.

3 3