Added Discard instance

[Risto-Stevcev]

No description provided.

[paf31]

I don't think errors should be discardable.

[Risto-Stevcev]

liftEff and launchAff get used often enough, and those values are worth ignoring a lot of the time. For liftEff, if it has a type signature like Either Error Unit, I don't see an issue with discarding that

[paf31]

I think encouraging users to discard errors like this would be a bad thing.

[Risto-Stevcev]

They would still be able to discard it using _ <-. Isn't the Discard class and instances (or lack thereof) just avoiding the extra boilerplate? What would it be stopping people from doing that they can't already do?

[paf31]

The point of Discard is to force users to acknowledge that they are discarding something which is non-trivial. Error is non-trivial, so users should be forced to acknowledge it. In fact, it's sort of the perfect example of what Discard is for 😄

[Risto-Stevcev]

The point of Discard is to force users to acknowledge that they are discarding something which is non-trivial

I guess I just don't understand what this is accomplishing. Doesn't this only benefit people who are new to purescript and don't understand that every line in a do block returns a value?

Everyone who's familiar with do notation knows that if they write foo bar in a do block, that it's equivalent to _ <- foo bar, and not having a Discard instance doesn't prevent someone from writing _ <- instead of whatever <-

[paf31]

Doesn't this only benefit people who are new to purescript and don't understand that every line in a do block returns a value?

No, I've had bugs recently which were caused by my forgetting to use a value returned in a do block.

[paf31]

Also, if you think Discard is not accomplishing anything, then we should discuss it on the compiler issue tracker. It seems like Error is an arbitrary choice in that case.

[garyb]

No, I've had bugs recently which were caused by my forgetting to use a value returned in a do block.

Likewise, that's why it was proposed in the first place - GHC warns about it too for similar reasoning.

👎 from to this from me I'm afraid!

[Risto-Stevcev]

Huh, that's interesting. I've never ran into an issue with that, but you guys have way more experience coding in purescript than me. Ok I guess if it's a 👎 I'll go ahead and close this PR

[garyb]

There's an example of the kind of place it can occur in purescript/purescript#1803 (comment) - due to a use of <$> rather than a =<< in that case. It's not super common, but I have done it (or things like it) several times, it's tricky to track down why an effect is not occurring :)