May 2008 - Haskell-Cafe

Aren't type system extensions fun?
by Andrew Coppin 23 Jul '08

23 Jul '08

Haskell 98 provides a simple and clean type system, which I feel I understand very well. GHC provides a vast zoo of strange and perplexing type system extensions, which I don't understand at all. (Well, some of it is simple enough - e.g., multiparameter type classes. But GADTs? FDs? ATs? Hmm...) Anyway, it seems there is a large set of such type system extensions that involve writing "forall" all over the place. I have by now more or less managed to comprehend the fact that data Thing = forall x. Thing x allows a type variable to appear on the RHS that is *not* present on the LHS, thus "hiding" the type of something inside the structure. And for some reason, they call this "existential quantification" [which I can't spell never mind pronounce]. Today I was reading a potentially interesting paper, and I stumbled across something referred to as a "rank-2 type". Specifically, class Typable x => Term x where gmapT :: (forall y. Term y => y -> y) -> x -> x At this point, I am at a complete loss as to how this is any different from gmapT :: Term y => (y -> y) -> x -> x Can anybody enlighten me? This is probably the first real use I've ever seen of so-called rank-2 types, and I'm curios to know why people think they need to exist. [Obviously when somebody vastly more intelligent than me says something is necessary, they probably know something I don't...] At this point, I don't think I even wanna *know* what the hell a rank-N type is... o_O

23 42

NW Functional Programming Interest Group
by Greg Meredith 19 Jul '08

19 Jul '08

All, Apologies for multiple listings. A small cadre of us are organizing a Northwest Functional Programming Interest Group (hey... NWFPIG, that's kinda funny). Our first official meeting is at the The Seattle Public Library 1000 - 4th Ave. Seattle, WA 98104 from 17:00 - 18:00 on February 20th. On the first meeting's agenda we'll be - giving people who are interested in or actively using FP for work or play a chance to meet - seeking to build up a pipeline of presentations and guest speakers - trying to keep organizational mishigosh to a minimum Hope to see you there. Monadically yours, --greg -- L.G. Meredith Managing Partner Biosimilarity LLC 806 55th St NE Seattle, WA 98105 +1 206.650.3740 http://biosimilarity.blogspot.com

2 7

I/O system brokenness with named pipes
by John Goerzen 06 Jul '08

06 Jul '08

So I have a need to write data to a POSIX named pipe (aka FIFO). Long story involving a command that doesn't have an option to read data from stdin, but can from a named pipe. I have created the named pipe from Haskell no problem. But I can't use writeFile to write data to it. Worse, it returns: *** Exception: /tmp/bakroller.zD0xHj/fifo: openFile: does not exist (No such device or address) which is completely false, as it *does* exist, and I can cat to it as expected. The call should block until everything is read on the remote end. I thought maybe writeFile is being weird, so I tried: openFile "/tmp/bakroller.zD0xHj/fifo" WriteMode Same thing. There is no logical reason I can see for this behavior. In fact, something must be going to *extra* effort to avoid writing to a named pipe, and I can't work out why. Named pipes are a standard, useful part of a system and shouldn't be ignored like this. Interestingly, readFile works fine on a named pipe. What's going on here? Am I going to have to resort to the System.Posix interface just to be able to write to a named pipe? -- John

10 20

appending an element to a list
by Adrian Neumann 12 Jun '08

12 Jun '08

Hello, I was wondering how expensive appending something to a list really is. Say I write I'd say "longList ++ [5]" stays unevaluated until I consumed the whole list and then appending should go in O(1). Similarly when concatenating two lists. Is that true, or am I missing something? Adrian

10 15

hmatrix
by Anatoly Yakovenko 05 Jun '08

05 Jun '08

What is the most efficient way to update a position in a matrix or a vector? I came up with this: updateVector :: Vector Double -> Int -> Double -> Vector Double updateVector vec pos val = vec `add` v2 where v2 = fromList $ (replicate (pos) 0.0) ++ ((val - (vec @> pos)):(replicate ((dim vec)- pos - 1) 0.0)) but this seems pretty inefficient to me. thanks, Anatoly

4 12

Ubuntu and ghc
by Galchin, Vasili 05 Jun '08

05 Jun '08

Hello, https://bugs.launchpad.net/ubuntu/+source/gtk2hs/+bug/229489 .... this is almost identical to my problem. I am just trying to help others on this list who are using Ubuntu Linux to avoid my predicament! Kind regards, Vasili

11 27

library for drawing charts
by Peter Gammie 05 Jun '08

05 Jun '08

Hello, Has anyone got some code for drawing charts? I don't mean graphs of functions, ala http://dockerz.net/twd/HaskellCharts and I don't mean calls to the Google Charts API ala http://community.livejournal.com/evan_tech/241080.html I would like something that can generate PNGs in memory, i.e. not directly to a file. This is close to what I have in mind: http://www.fred.net/brv/chart/ I'd like 2D pie charts, bar charts and something like a google-o-meter. Any pointers? cheers peter

4 5

problems with derive/TH
by Thomas Hartman 04 Jun '08

04 Jun '08

Haskellers/Chris, As part of my quest to better understand STM and eventually build a concurrent haskell server app, I checked out an embryonic haskell mud demo from Chris Smith's repo. It seemed like a great sample app. But I ran into a build issue that seems to have to do with derive/TH. I am assuming this used to work, but something changed, either in TH itself or switching from ghc6.6 to ghc6.8. From the error message (below) I surmise either a) M.Map based values used to automagically work with derive/TH, but no longer do... in which case I need to create this instance manually or b) this should still work, more or less... but I have to import the map instance via an additional module, or something like that. Any advice? Thomas. ***** This is the error I am getting: thartman@thartman-laptop:~/mud>cat _darcs/prefs/defaultrepo http://cdsmith.twu.net/demos/mud thartman@thartman-laptop:~/mud>ghc --version The Glorious Glasgow Haskell Compilation System, version 6.8.2 thartman@thartman-laptop:~/mud>runghc Setup.lhs configure; runghc Setup.lhs build Configuring mud-0.1... Preprocessing executables for mud-0.1... Building mud-0.1... [2 of 5] Compiling World ( World.hs, dist/build/mud/mud-tmp/World.o ) World.hs:42:0: No instances for (Eq (a (M.Map String Player)), Eq (a (M.Map ItemId Item)), Eq (a (M.Map PlayerId Player)), Eq (a (M.Map RoomId Room)), Eq (a RoomId)) arising from the 'deriving' clause of a data type declaration at World.hs:(42,0)-(50,14) Possible fix: add an instance declaration for (Eq (a (M.Map String Player)), Eq (a (M.Map ItemId Item)), Eq (a (M.Map PlayerId Player)), Eq (a (M.Map RoomId Room)), Eq (a RoomId)) When deriving the instance for (Eq (World a))

5 5

[ANN] bloomfilter 1.0 - Fast immutable and mutable Bloom filters
by Bryan O'Sullivan 03 Jun '08

03 Jun '08

I'm pleased to announce the availability of a fast Bloom filter library for Haskell. A Bloom filter is a probabilistic data structure that provides a fast set membership querying capability. It does not give false negatives, but has a tunable false positive rate. (A false positive arises when the filter claims that an element is present, but in fact it is not.) The library is easy to use. As an example, here's a reimplementation of the Unix "spell" command. import Data.BloomFilter.Easy (easyList, elemB) main = do filt <- (easyList 0.01 . words) `fmap` readFile "dictionary.txt" let check word | word `elemB` filt = "" | otherwise = word ++ "\n" interact (concat . map check . lines) It is also carefully tuned for performance. On my laptop, I can sustain a construction or query rate well in excess of a million ByteStrings per second. Source code: http://hackage.haskell.org/cgi-bin/hackage-scripts/package/bloomfilter Latest bits: darcs get http://darcs.serpentine.com/bloomfilter <b

10 12

ANN (2 Libs) -- hvac 0.1b, a lightweight web framework and HStringTemplate 0.3
by Sterling Clover 03 Jun '08

03 Jun '08

1) hvac 0.1b: transactional, declarative framework for lightweight web applications. 2) HStringTemplate 0.3 1) hvac 0.1b hvac (short for http view and controller) has been my project for the last little while, and is finally in a fairly usable state, so I'm opening up the repo (darcs get http://community.haskell.org/~sclv/ hvac/) for folks to play with and to get some feedback. While not quite yet ready for hackage, the package as provided should be fully cabal installable. Documentation is available at http:// community.haskell.org/~sclv/hvac/html_docs/hvac/ The aim of hvac is to provide an environment that makes the creation of lightweight fastcgi based web applications as simple as possible, with an emphasis on concise, declarative style code, correct concurrent transactional logic, and transparency in adding caching combinators. There are two included example programs, naturally neither of which is feature complete. They share a common login module of about 50 lines of code, excluding imports, templates, and database schema. The first program is a classic, greenspun-style message board with basic login functionality. It totals roughly 40 lines and tends to use just under 4mb of resident memory on my system. The second is a wiki based on Pandoc and the PandocWiki code. The code totals roughly 30 lines (rendering borrowed from PandocWiki aside) and uses about 5mb of memory. hvac processes all requests in the STM monad, with some bells attached to properly interleave STM with session, database and filesystem operations such that they all conceptually occur together in a single transaction per request. Currently it is only fully tested with sqlite, but it should operate, modulo a few tweaks, with any database accessible via HDBC. hvac is particularly designed to use the HStringTemplate library as an output layer, in a simple declarative fashion. As the StringTemplate grammar is explicitly sub-turing, this ensures a clean separation of program logic from presentation, while providing a nonetheless fairly powerful language to express typical display tasks. The included cache combinators, still experimental, should allow a simple and fine-grained control over the level of caching of various disk-bound operations. Phantom types are used to ensure that no functions that modify state may be cached. To give a flavor of hvac code, the following is the complete (twenty lines!) source of the wiki controller (due to sql statements, some lines are rather long): wikiController tmpl = h |/ "login" *> login_plug tmpl <|> (h |/ "wiki" |\\ \pageName -> h |// "POST" *> withValidation [ ("contents", return) ] (\ [contents] -> do pageId <- selectVal "id from pages where name=?" [toSql pageName] maybe (addErrors [("Login","must be logged in.")] >> continue) (\user -> case fromSql pageId of Just (_::Int) -> execStat "insert into page_hist (pageId,contents,author) values(?,?,?)" [pageId, toSql contents, toSql . userName $ user] Nothing -> do execStat "insert into pages (name,locked) values(?,?)" [toSql pageName, toSql (0::Int)] pid <- selectVal "max(id) from pages" [] execStat "insert into page_hist (pageId,contents,author) values(?,?,?)" [pid, toSql contents, toSql . userName $ user]) =<< getSes continue) <|> do pageId <- selectVal "id from pages where name=?" [toSql pageName] (join $ renderf (tmpl "showPage") ("pageName", pageName) <$> "pageContents" |= selectRow "* from page_hist where pageId=? order by time desc limit 1" [pageId] )) <|> (redirect . ( ++ "/wiki/Index") =<< scriptName) Future directions for work on hvac include: Stress testing for correctness of transactional logic and benchmarks. Exploration of various efficiency tweaks. Unit tests. Further development of the cache combinator API. Improvement of the example apps and addition of a few others (a blog maybe). Expansion of the library of validator functions. Exploration of transferring hvac to the standard FastCGI bindings (currently it uses a custom modified version to work properly with STM). Improvement of the database layer, particularly with regards to common paging functions. Creation of a set of simple combinators for generating CRUD (create read update delete) pages. Creation of a minimal set of standard templates (maybe). 2) HStringTemplate 0.3.1 This release of HStringTemplate (up now at Hackage) fixes a number of bugs pointed out to me by its small but growing user base (thanks, cinema, elliottt!) ranging from the minor (a particular two-level iteration pattern wasn't working properly) to the truly irritating (poor handling of file groups). It's still unfortunately skimpy on the docs, outside of the haddocks and the main StringTemplate grammar documentation at http://www.stringtemplate.org (although the examples from hvac should also prove helpful). However, it does have a set of very nice and handy new features for development. * renderf, a function similar in spirit to printf, that takes an arbitrary number of heterogeneous (String, value) tuples as arguments. This should cut down considerably on long setAttribute chains. Additionally, with custom instances (not, I'll grant, trivial to write) it can be used to declaratively chain together strings of attribute retrieval functions in arbitrary monads, as in the above code example from hvac. * dumpAttribs, a function/template that prints out the tree of the entire attribute environment a template is operating in -- extremely handy for development. * nullGroup, also for use in development, a simple way to display more information about templates that can't be found. Error messages in usafeVolatileDirectoryGroup have also been significantly improved. * getStringTemplate', a version of getStringTemplate guaranteed not to be inlined. While the optimizer will still sometimes rearrange code such that a volatile group is not updated properly, this at least helps remedy the situation (I think). * Some minor changes: For grammar reasons, dots have been removed from template names -- however, underscores and slashes are now available. Additionally, there's a much improved logic for which aspects of a local environment are overridden and preserved when a template is called from another. For both of these libraries, patches, comments, bug reports, requests, and of course contributions more than welcome! Regards, Sterl.

3 3