April 2015 - Beginners

Re: [Haskell-beginners] StateT MyState IO a vs ReaderT (IORef MyState) IO a
by Grzegorz Milka 19 Apr '15

19 Apr '15

ReaderT (IORef MyState) IO a is introducing a global mutable state into your IO actions. This style is discouraged, because you have less guarantees about what can happen with your state. For example IORef is not protected from multi-threaded access. If you try to use this state in multiple threads you may run into bugs. StateT MyState IO a doesn't have that problem. For a typical use-case, that is an action that: reads state, performs action, modifies state, both are similar and neither should be significantly more comfortable. I would recommend to use StateT version unless you find that either it does not fit your model of computation or that it is underperforming. As for the performance it is hard to say. StateT allows the compiler to reason more about your code, so it might optimize better. However that's only one factor. You would need to test it to find out which one is better.

1 0

Re: [Haskell-beginners] GCJ 2015 - Round 1A Problem - Haircut
by Doug McIlroy 18 Apr '15

18 Apr '15

> Can someone tell me how I could have avoided or fixed this? The trouble manifested as stack overflow on solving > https://code.google.com/codejam/contest/4224486/dashboard#s=p1 For efficiency, you'll probably need more cleverness in math than in Haskell. You can predict an approximate start time for the Nth customer's service from the average per-customer service time M, where 1/M = sum 1/M_k. Starting from that estimate, one can skip over almost the entire service simulation. Doug McIlroy

3 3

GCJ 2015 - Round 1A Problem - Haircut
by Sourabh 18 Apr '15

18 Apr '15

So I decided to try the Google Code Jam this year, and decided to code in Haskell. I passed the qualifiers, and today was Round 1A. I managed to get the first problem right (both small and large input), and got the 2nd problem right (for the small input). However, for the large input, it ran out of stack space with the following error: Stack space overflow: current size 33632 bytes Use '+RTS -Ksize -RTS' to increase it. Can someone tell me how I could have avoided or fixed this? It is also very possible that I had an algorithmic issue (needed to optimize my code or have a different algorithm for the large input perhaps?). Here is the problem description: https://code.google.com/codejam/contest/4224486/dashboard#s=p1 And here is my code: https://github.com/cbrghostrider/Hacking/blob/master/codeJam/2015/haircut/h… I made sure to use foldr, which is supposedly more stack efficient I believe. I don't know if the list append (++) was causing issues. If anyone can spot any problems, or provide suggestions, I would really appreciate it! Even though the round is done, I'd like to optimize this to make it run within 8 minutes, as intended. Thanks in advance! -- SSJ

2 1

State Monad
by Thomas Jakway 18 Apr '15

18 Apr '15

I'm having some trouble using the state monad with a random number generator. I'm trying to follow the LYAH example (http://learnyouahaskell.com/for-a-few-monads-more) but apparently the state monad API has changed, so I'm using Control.Monad.Trans.State.Lazy (state monad from the transformers package). Side question: is using transformers a good idea? I've heard some people complain about mtl's performance, others don't seem to care. I'm far too new to be able to judge anything (does it even make sense to compare transformers vs. mtl?) but if one has overtaken the other I'd rather use that. Here's what I currently have: import Control.Monad.Trans.State.Lazy type GeneratorState = State StdGen genThree :: Int -> GeneratorState Int genThree listMax = do --highest index in the list let listMin = 0 :: Int --lowest index in the list generator <- get let (generatedMin, state) = randomR (listMin, listMax) generator return generatedMin Although it typechecks it doesn't seem like an improvement over just using StdGen by itself. What I think I should have: genThree :: Int -> GeneratorState Int genThree listMax = do --highest index in the list let listMin = 0 :: Int --lowest index in the list generatedMin <- state randomR (listMin, listMax) return generatedMin * *I feel like I really botched this--what am I missing? * *

3 4

texture mapping with SDL
by Florian Gillard 18 Apr '15

18 Apr '15

Hi everyone, I am not sure my previous message went trough, if so, sorry for the double post. I am trying to make a basic raycaster (something like the first wolfenstein 3D) using haskell and SDL 1.2 So far I have it working using coloured lines and I would like to know if there is any way to apply transforms to textures loaded in memory using SDL, in order to achieve basic texture mapping on the walls. I looked at the SDL doc but I didn't find anything looking like what I need. The code is there: https://github.com/eniac314/maze-generator/blob/master/raycaster.hs sreenshot here: https://github.com/eniac314/maze-generator/blob/master/raycaster.png (the last part is not done yet but has nothing to do with the raycaster) I would appreciate any suggestion :)

2 2

texture mapping with SDL
by Florian Gillard 17 Apr '15

17 Apr '15

Hi everyone, I am trying to make a basic raycaster (something like the first wolfenstein 3D) using haskell and SDL 1.2 So far I have it working using coloured lines and I would like to know if there is any way to apply transforms to textures loaded in memory using SDL, in order to achieve basic texture mapping on the walls. I looked at the SDL doc but I didn't find anything looking like what I need. The code is there: https://github.com/eniac314/maze-generator/blob/master/raycaster.hs (the last part is not done yet but has nothing to do with the raycaster) I would appreciate any suggestion :)

1 0

Left vs Right
by Shishir Srivastava 17 Apr '15

17 Apr '15

Hi, Can someone please explain the difference in outputs of the following two expressions - -------------- ghci> fmap (replicate 3) (Right "blah") Right ["blah","blah","blah"] ghci> fmap (replicate 3) (Left "foo") Left "foo" --------------- Why does 'Right' return a list of Strings whereas 'Left' returns just a String. Thanks, Shishir

9 12

Point free mystery
by martin 17 Apr '15

17 Apr '15

Helly all for ad-hoc benchmarking I wrote bench x = (timeIt . print) x Hlint suggested Found: bench x = (timeIt . print) x Why not: bench = (timeIt . print) but when I do this, I get No instance for (Show a0) arising from a use of ‘print’ Why is that so?

3 2

Inventing QuickCheck properties
by Dan Serban 16 Apr '15

16 Apr '15

Hi, I'm writing QuickCheck properties for the swipeLeft function in a game of 2048. I could think of 3 very basic properties so far and I'm wondering if you guys can think of any other interesting ones. Here are mine: - Verify That It Makes The Number Of Zeroes Go Up (applies "length . (filter (==0))" to both input and output and then does a <= comparison) - Verify That It Cleanly Separates Zeroes To The Right-Hand Side (applies "all (==0) . snd . break (==0)" to the output) - Verify that It Preserves The SumTotal (assert that the sum of all integers is left unchanged) -Dan

1 0

Are Arrows good for simulation purposes
by martin 16 Apr '15

16 Apr '15

Hello all, little do I know about arrows, but the "stream processor" metaphor suggests, that they can be used to simulate a network of things, services or data. Is this correct? I came across the following thought: When I simulate a billard game with a DES, I would compute collisions of balls with each other and with the banks, creatign a set of events, of which only the earliest will be considered to compute the next state. This is pure DES and does not seem to be a good candidate for arrows. In this case I'd be more interested in composing collision detectors than in stream processing. OTOH, when I move parcels around and process then in various stages, then a "stream processor" would make perfect sense to me. In that case, would I abandon the DES paradigm entirely (including the notion of an event queue) and just model the network and let it run?

2 1