June 2014 - Haskell-Cafe

ANN: loop - fast for loops
by Niklas Hambüchen 20 Jul '14

20 Jul '14

Announcing the first version of the loop package: http://hackage.haskell.org/package/loop This is for iterating over numeric or otherwise iteratable things in a fast fashion. Looks like C, runs like C. It's trivial, but very robust performance-wise and more independent of GHC smartness or non-smartness in regards to optimising alternatives like `forM_ [1..n]`. It comes with a lot of benchmarks. Contributions are very welcome.

2 2

Is there a name for this property: x + x - x == x
by Omari Norman 20 Jul '14

20 Jul '14

It's not quite idempotence, because more than one function is involved. It's a common property and I figure I can write a higher order function to build QuickCheck tests for it. I was just wondering if it has a name.

7 7

New Functional Programming Job Opportunities
by Functional Jobs 04 Jul '14

04 Jul '14

Here are some functional programming job opportunities that were posted recently: Developer / DevOps Engineer at Klarna http://functionaljobs.com/jobs/8721-developer-devops-engineer-at-klarna Cheers, Sean Murphy FunctionalJobs.com

5 8

Declarative database migrations
by Timur Amirov 03 Jul '14

03 Jul '14

Hello! Looking for tools to use discovered that I miss something like Rails database migrations in haskell. Is there a package for it (describe the whole schema and/or particular migrations)? -- Timur Amirov Berlin, Germany

5 6

handling non-string command line arguments
by Johan Larson 02 Jul '14

02 Jul '14

I've been looking at choices for parsing command line arguments, including getOpt. The examples I can find focus on string arguments, whereas I am interested in numbers. In the application at hand, it is particularly important to issue clear error messages when arguments don't parse as numbers or are out of range. I could use getOpt as a first pass with string arguments and then turn the strings into validated numbers in a second pass, but that's a bit awkward. Alternately I could use the options records with Options -> IO Options functions. But both of these solutions treat type mismatches differently from other options errors. Has anyone found a cleaner solution? -- Johan Larson -- Toronto, Canada

6 6

Haskell Wiki user account creation
by Haitham Gad 01 Jul '14

01 Jul '14

Hi, I read the following note at the log-in page of haskell wiki: NOTE: Automatic wiki account creation has been disabled. If you would like an account please email "nominolo" (at the email service from Google) or on the haskell-cafe mailing list. Does anyone know the exact email address I should be sending to (is it nominolo(a)gmail.com)? Thanks, Haitham

4 6

Need help - my haskell code is over 50 times slower than equivalent perl implementation
by C K Kashyap 01 Jul '14

01 Jul '14

Dear cafe, I have created a reproducible app here - https://github.com/ckkashyap/haskell-perf-repro Essentially I am trying to open a number of files and printing out their sized by reading them in and computing it's length. I have the equivalent perl program also there. I'd appreciate it very much if someone could take a look at my Haskell implementation and point out what I am doing wrong. Since it's over 50 times slower than the perl code, I assume I am doing something obviously incorrect. Regards, Kashyap

10 23

Making new Unboxed Algebraic Types? (A Repa Problem)
by Hans Georg Schaathun 30 Jun '14

30 Jun '14

Hi, could someone offer som advice on handling composite types with the Repa library? I can see a /long/ way around my problem, but I wonder what is the /best/ way ... The problem is that I am trying to parallellise a probabilistic algorithm using Repa. To handle pseudo-randomness in parallel, I couple each data element with a generator state, in a tuple. The generator state is of type TFGen (tf-random package) or StdGen¹ (random package), neither of which is an Unbox instance. Hence, I have been using boxed repa arrays. The code does not execute in parallel, and if I understand correctly this is because of the Boxed type. There is a workaround², but that requires an NFData instance. To make a NFData instance and work through the code to support it should be as simple as it is tedious, but it would still leave the other drawbacks of a boxed type. (In particular, I do sequential fold because the Repa fold function did not support boxed types.) I could of course rewrite the published libraries to use tuples instead of algebraic types, and benefit from the pre-made Unbox instances, but sounds messy and error-prone (as well as boring). If I have to I have to. Is there a quick and robust way to make Unbox instances from composite datatypes? Or any other advice on how to approach the matter? Both random ideas, references, and concrete solutions are welcome. [The other element of the tuple is an vector of unboxed constituent type. I am not sure how that's going to work, but I'll cross that bridge when I get there. For the time being it is the generator that stops type checking.] TIA ¹ Yes, I know that StdGen is statistically unsound. ² http://stackoverflow.com/questions/16097418/parallel-repa-code-doesnt-creat… -- :-- Hans Georg

4 4

Monad laws
by Rafael Almeida 30 Jun '14

30 Jun '14

Hello, You guys have been great taking my questions. Thank you. Now I have another one :) What can go wrong if I make an Monad instance but don't follow Monad rules (identity and associativity)? Sure it would be misleading for someone using the non-conforming class. They may make code that assume those laws, although they don't hold. However, could it make the compiler generate bad code or some standard functions to behave badly or even break? []'s Rafael

10 19

Re: [Haskell-cafe] Making new Unboxed Algebraic Types? (A Repa Problem)
by Hans Georg Schaathun 30 Jun '14

30 Jun '14

On Tue, Jul 01, 2014 at 01:43:32AM +0400, Dmitry Dzhus wrote: > This implies that the amount of generator states used in your parallel > computation is equal to data array size N. This is not the case in > general (unless the computation is split into N processes running in > parallel by the scheduler). On the contrary, commonly you would even > want to minimize the amount of generator states used (for resource > efficiency reasons). Fair point. The large number of states may be a disadvantage. > Generator states may be implicitly passed into sequential > subcomputations (scheduled by the parallel planner). This of course > calls for some hacking on Repa harness that runs the computations in > parallel (computeP). Such a solution would probably not be determinstic, and it would definitely make the output depend on the run-time environment (such as the number of threads). I am interested in the particular problem of pseudo-random, /deterministic/, parallel programs. I think the large number of generator states is necessary to achieve deterministic, parallel behaviour. If someone knows otherwise, I would be curious to learn. > Some time ago I was also struggling find a way to parallelize a > probabilistic algorithm (Monte-Carlo molecular simulation) and > eventually I settled with even simpler solution based on strategies > instead of Repa: > https://github.com/dzhus/dsmc/blob/master/src/Control/Parallel/Stochastic.hs. > It can be easily ported to any stateful mapping computation that needs > parallelization (including your PRNGs of choice). See the whole package > for usage examples. Thanks a lot for the pointer. I shall have a look. In fact I am curious to learn about strategies as well. (Contrary to low-level hacking of Repa features which I am not at all curious about :-) -- :-- Hans Georg

1 0