Re: [Haskell-cafe] ANN: pandoc 1.0.0.1

On Sat, Sep 13, 2008 at 09:31:50PM -0500, brian wrote:

On Sat, Sep 13, 2008 at 8:29 PM, John MacFarlane wrote:

...

+ Code is -Wall clean

Thanks, I appreciate it. I wish more people paid attention to this.

Well. I often pay attention to it. That doesn't mean I always heed it ;-) Like all -Wall stuff, sometimes warnings are great and sometimes warnings are noise. There are two warnings that I am prone to ignore: 1) No type definition for top-level declarations Whether or not I ignore this depends on what sort of code I'm working on. If I'm using Haskell as, essentially, a script language, or writing a quick and small Parsec parser, adding a bunch of type declarations can serve to make the code less readable and certainly more difficult to update and maintain. That said, for larger projects or computation algorithms, I'd usually add declarations. When I'm leaving off the declarations, my code sometimes winds up looking like something vaguely resembling Python. Sometimes time declarations just get in the way. Cue type inference. One of my favorite things about Haskell: it can be as compact as Python (or moreso), yet detect type errors at compile time. 2) Variable x defined but not used I most often ignore this when it occurs in a function definition. Sometimes I may have a function that could be written foo _ (x, _) _ = bar (x + 5) But for clarity's sake on what all the unused args are, which really helps for future maintainability, I'll usually use a descriptive -- but unused -- variable name. Some of my libraries, on the other hand, are periodically validated against -Wall -Werror. -- John