Testing was Re: main modules in GHC, deriving differences between GHC and Hugs

[Matthew Donadio]

(redirected to haskell-cafe)

Hal Daume III wrote:
> Yes, but there's a problem with this solution.  Namely, if Foo.hs takes a
> long time to compile, then you can't leverage having already created Foo.o
> and Foo.hi when making the Main.

Yeah, it's not perfect, but I think we just have different methodologies
for testing.

Typically, I do most of my development with ghci or hugs.  Each module
will have a single variable that represents the tests

	tests :: [Bool]
	tests = [ test1, test2, test3 ... ]

and then I define a variable called

	test :: Bool
	test = and tests

so I can just load a module and either evaluate tests or test to check
things out.  This generalizes to importing and testing several modules
at once (as long as I take care of name conflists).

This only works for simple modules, though.  For more complicated ones,
I have a "pretty" tester.  For example, my Haskell FFT library is
collection of mutually recursive modules.  I have a specialzed test
function for this that tests a range of lengths.  It looks someone like

	main = testfft n1 n2

	testfft n1 n2 = sequence $ map test1fft [n1..n2]

	test1fft n = do putStr $ show n ++ ":\t"
		        putStr $ if ok then "OK\n" else "ERROR\n"
	     where ok = and [ test1 n, test2 n, test3 n ]

This way I can compile the code, and run the executable as

	ffttest 2 2048 | grep ERROR

and I am confident that I get full coverage of the algorithm.

I am always to hear about other methods of automated testing.

-- 
Matthew Donadio (m.p.donadio@ieee.org)