Should context-manipulating functions return effectful unit instead of context?

[aztecrex]

The context-manipulating functions all return a context in Eff. However, the original
context is modified by the function. For do notation, this adds the noise of either
void-ing operations or capturing their values to underscores. I understand there is
purescript-free-canvas but that is not yet in package-sets so I hesitate to use it.

In fact, because of the Graphics.Canvas API, I originally (wrongly) assumed that I
had to capture each context and pass it to the next operation. In this API, the types
don't tell the story which makes it harder to learn.

I propose that such operations return an Eff Unit instead of Eff Context2D. That
is a breaking change, of course, so no PR yet.

[paf31]

Yes, I agree we should return Unit especially with the new Discard warnings in the compiler. We should break this in the next major release.

[jacereda]

Given a signature like:

setLineWidth :: forall eff. Number -> Context2D -> Eff (canvas :: CANVAS | eff) Context2D

I'd either:

• Retain current signature and add (note that Context2D is now the first argument):

setLineWidth_ :: forall eff. Context2D -> Number -> Eff (canvas :: CANVAS | eff) Unit

• Break compatibility and use the above as setLineWidth.

[jacereda]

I've been surveying usage of this module on github and didn't find instances where chained operation was used (via >>= I mean). So, I guess the second option wouldn't be too radical.

[hdgarrood]

Resolved by #53.