Allow exceptions and logging in render function

[whittle]

The runEma and runEmaWithCli functions expect the Asset renderer passed to be a pure function. It would be helpful to be able to raise exceptions or log messages from the render action.

[srid]

@whittle What sort of errors would your render function be raising? Just trying to better understand the original goal of this solution.

[whittle]

Currently, I’m working on a static wiki-alike generator, and I started from srid/ema-template. srid/ema-template uses Control.Exception.throw to raise exceptions from the rpBlock and rpInline functions, which are part of the render pathway. In my generator, I’d like to avoid throwing IO exceptions from pure code, even if it does muddy the type signatures.

When I originally made this request, I also wanted to be able to raise exceptions from the Pandoc transformations that happen in the render pathway, but I’m currently working on moving those to the modeling thread.

[srid]

I’m currently working on moving those to the modeling thread.

Right ... generally speaking, it is best to do all the impure stuff inside the runEma block where the model is being (incrementally) built. And when it comes to the pure render function it is expected to terminate successfully with the asset value (unless when raising errors that are not to be handled).

But I can see the value of having Ema take a monadic render function, with most users just doing Ema.runEma (runIdentity . render) or something like that.

[srid]

The render function (siteOutput) in master branch supports IO, so this should no longer be an issue, @whittle

I'm yet to release 0.8, but docs are more or less up to date: https://ema.srid.ca/