Re: [Haskell-cafe] How Albus Dumbledore would sell Haskell

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 ajb@spamcop.net wrote:

G'day all.

Quoting Isaac Dupree :

...

Okay, looking at that code: The comments before the type definitions are mostly good... now it looks like I'm going into critique mode :)

BTW, for the record, I didn't try too hard with this. It is meant to be illustrative of what you can do with Haskell and not too much time to spare.

I know, (but if you got it from somewhere, the code could really be improved where it came from? >:) or did I misunderstand?)

I didn't haddock-ise the comments because Diff isn't a library. The comments are meant to be more commentary (this is a tutorial, remember!) than developer documentation.

Admittedly. Haddock standardizes what bit of the code a comment is meant to refer to, though, so I like using its syntax anyway. (And all code might be part of a library someday... Haskell lets you think of almost all your code as libraries! it is an excellent way to think about modularity that is far too cumbersome in many other languages!)

...

I suppose Match is the same (inclusive), but I can't tell. And personally I don't know what the significance/meaning of a "found match" is.

OK, this needs explanation. A Match is a match _between_ the two files being diffed. So (a,b) means that line a in file 1 matches line b in file 2. I'll note that, thanks.

Ah, I see now.

...

Comparing "data Range = Range Line Line" and "type Match = (Line,Line)", they are isomorphic, but declared differently (which is fine IMO)

Yup. They only have to be different enough to cause a type error if you accidentally try to mix them.

Tuples are always a bit risky though (every tuple matches every other tuple you introduce, as well as libraries using them) -- but undoubtedly convenient syntax (and certainly worth showing off to those whose languages don't contain native tuples)

...

The only IO that displayDiff does is putStrLn. It should probably return a string instead [...]

Possibly, or even use a Writer monad. I habitually use putStrLn, though, because I regularly program in multiple languages, and this way I don't have to remember how Haskell handles line termination on different platforms.

Hmm, reasonable. (I hope haskell just uses '\n' in-program everywhere..) You could return a list of lines instead (and mapM_ putStrLn the result). Is there any way to make IO be an instance of MonadWriter? Not in a reasonable way, I think, given 'listen' etc... possibly a different class interface would be more useful, but we're certainly not into trying that here! Isaac -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFGKOjaHgcxvIWYTTURAj/tAKCF1NoAR/tsCk/2F90qrJqotEa+GACfbYzs uWYo+EFqlpZTnsdS1YTndl8= =IR0K -----END PGP SIGNATURE-----