I've been interested in contributing to this project for a while! Is it ok if I take a shot at this issue?
I think maybe checking the length, the first and the last element of every dataset could constitute the tests.

Hi @Magalame , any contribution large or small is very welcome ^^

Hi again!
I was wondering if it wouldn't be more practical to using some form of hashing? Maybe something like hashable?

@Magalame how would you use hashing for this ?

I think I misunderstood the purpose of the issue, I thought it was to check the integrity of the data

Actually I think there is a reasonable way to use hashing for this. We basically want unit tests to make sure that the functions in datasets load the data sets properly.

So a nice way to do that would be, approximately, for each data set:

do
    abas <- getDataset abalone

    head abas `shouldBe` valueOfHeadOfAbas
    last abas `shouldBe` valueOfLastOfAbas
    force (hash abas) `shouldBe` fixedHash

and then the same for getDatavec

It would have the advantage of partly solving #29. And Haskell has a nice Hashable class

ping @ocramz