CODEWITHSUNDEEP
haskellmonadslazy-evaluationevaluationoperator-precedence
Stéphane Laurent
Stéphane Laurent

Reputation: 84519

Evaluation of "something <- stuff" statements

Is it true that statements like something <- stuff are always evaluated in Haskell, even when something is not called in the rest of the code ? (is something <- stuff called an "action" ? - I don't know the technical wording).

If this is true, I have another question.

I have some code starting like this:

computeContour3d voxel voxmax level = do
    voxelmax <- somefunction voxel
    let max' = fromMaybe voxelmax voxmax

That is, if the argument voxmax is not Nothing, then voxelmax is not necessary, since max' = fromJust voxmax in this case. Therefore, if the answer to my first question is "Yes", how could I avoid the evaluation of voxelmax when it is not necessary ?

Upvotes: 0

Views: 89

Answers (2)

user8174234
user8174234

Reputation:

Is it true that statements like something <- stuff are always evaluated in Haskell, even when something is not called in the rest of the code ?

Not in general, no. The IO monad forces such evaluation, but many others do not.

is something <- stuff called an "action" ?

Usually that line would be called a monadic bind. Some monads (e.g. lists) don't really "act" in any meaningful way.

Upvotes: 0

Daniel Wagner
Daniel Wagner

Reputation: 152682

No, monadic binding is no guarantee that anything gets evaluated. There are lazy monads; e.g. the reader monad would not force the result of somefunction voxel unless voxmax was Nothing.

But there's no reason to rely on such behaviors; it is easy to reliably avoid extra computation:

computeContour3d voxel voxmax level = do
    max' <- case voxmax of
        Nothing -> somefunction voxel
        Just max -> return max
    -- use max'

You could consider using maybe, which is often shorter than an explicit case, as in:

    max' <- maybe (somefunction voxel) return voxmax

Upvotes: 5

Related Questions