Comments on: Stemming with Haskell reloaded http://blog.tupil.com/stemming-with-haskell-reloaded/ (Get up early, code often) Sat, 28 Aug 2010 04:45:18 +0000 hourly 1 http://wordpress.org/?v=3.0.1 By: Eelco Lempsink http://blog.tupil.com/stemming-with-haskell-reloaded/comment-page-1/#comment-143 Eelco Lempsink Thu, 24 Jul 2008 11:42:31 +0000 http://blog.tupil.com/?p=23#comment-143 ForgeDestiny, that's valuable feedback :) You're right, it's pretty readable on my screen, but we should probably make it a bit brighter (I think the comments are much better readable already) ForgeDestiny, that’s valuable feedback :) You’re right, it’s pretty readable on my screen, but we should probably make it a bit brighter (I think the comments are much better readable already)

]]> By: ForgeDestiny http://blog.tupil.com/stemming-with-haskell-reloaded/comment-page-1/#comment-142 ForgeDestiny Thu, 24 Jul 2008 10:43:45 +0000 http://blog.tupil.com/?p=23#comment-142 The Haskell code, me like. But the light gray text on black background made my eyes itch....and quiver briefly. I didn't read any of it. The Haskell code, me like. But the light gray text on black background made my eyes itch….and quiver briefly. I didn’t read any of it.

]]> By: Holden Karau http://blog.tupil.com/stemming-with-haskell-reloaded/comment-page-1/#comment-125 Holden Karau Mon, 21 Jul 2008 20:33:36 +0000 http://blog.tupil.com/?p=23#comment-125 yays! I was looking for a clean stemming library. Thanks :) yays! I was looking for a clean stemming library. Thanks :)

]]> By: Eelco Lempsink http://blog.tupil.com/stemming-with-haskell-reloaded/comment-page-1/#comment-119 Eelco Lempsink Mon, 21 Jul 2008 06:35:45 +0000 http://blog.tupil.com/?p=23#comment-119 That's a great tip, thanks Twan. I wasn't aware of <code>ForeignPtr</code> so I seem to have replicated its behavior. I think that it will even make <code>stemWords</code> unnecessary, since a simple <code>map stem</code> will be just as efficient. The C library claims to be thread-safe (and I'll take their word for it), so exporting multiple stemmer objects might also work very nicely. Thanks again :) That’s a great tip, thanks Twan. I wasn’t aware of ForeignPtr so I seem to have replicated its behavior.

I think that it will even make stemWords unnecessary, since a simple map stem will be just as efficient.

The C library claims to be thread-safe (and I’ll take their word for it), so exporting multiple stemmer objects might also work very nicely.

Thanks again :)

]]> By: Twan van Laarhoven http://blog.tupil.com/stemming-with-haskell-reloaded/comment-page-1/#comment-115 Twan van Laarhoven Sun, 20 Jul 2008 20:09:52 +0000 http://blog.tupil.com/?p=23#comment-115 Have you considered using a ForeignPtr to hold the stemmer? In that case you no longer need to delete manually, you use a finalizer instead. The interface can then be simply: <pre lang="haskell"> newtype Stemmer = ... stemmer :: Algorithm -> Stemmer stemWith :: Stemmer -> String -> String stemWith = C.stem </pre> The <code>stem</code> and <code>stemWords</code> can then be implemented trivially as <pre lang="haskell"> stem = stemWith . stemmer stemWords = map stemWith . stemmer </pre> If the C library is threadsafe and the stemmer objects are not too large, you could even eliminate the <code>Algorithm</code> type from the interface altogether, and just export multiple <code>Stemmer</code> objects. These will then be intialized when they are first used. Have you considered using a ForeignPtr to hold the stemmer? In that case you no longer need to delete manually, you use a finalizer instead. The interface can then be simply:

newtype Stemmer = ...
stemmer :: Algorithm -> Stemmer
stemWith :: Stemmer -> String -> String
stemWith = C.stem

The stem and stemWords can then be implemented trivially as

stem = stemWith . stemmer
stemWords = map stemWith . stemmer

If the C library is threadsafe and the stemmer objects are not too large, you could even eliminate the Algorithm type from the interface altogether, and just export multiple Stemmer objects. These will then be intialized when they are first used.

]]>