Override a context field with nothing

[kirelagin]

(Unfortunately, I did not find an easy way to play around with contexts in a repl, so instead of showing the code, I’ll have to describe the issue using words.)

After reading the documentation for field, I was under an impression that using empty or noResult will allow me to “unset” a field in a context, but apparently it does nothing at all.

Suppose, I want to create a context that will be the same as defaultContext, but without any title field (i.e. $if(title)$ in template should not get triggered). So, naturally, I tried field "title" (const empty) <> defaultContext, but it did not work as expected – I still got a title.

This has another, in my opinion, very unexpected implication: boolField "title" (something) <> defaultContext overrides the title field with True only if the value is True, but if it is False, the value is not overridden at all and remains a (True-valued) string.

[Minoru]

Hi, sorry for such a late reply!

I don't see a way to short-circuit <> i.e. make it abort immediately without looking at the rest of the contexts. I mean, we could add a new CompilerErrors to indicate that the result is empty, or we could wrap all results in Maybe, but both approaches look ugly to me (although perhaps I didn't think deep enough). What I'd suggest instead is this (untested) wrapper:

withoutField :: String -> Context a -> Context a
withoutField key (Context inner) = Context $ \k a i ->
    if k == key
        then noResult $ "Field " ++ k ++ " is nullified"
        else inner k a i

You would then do this:

defaultContext `withoutField` "title"

How does that sound to you?

So, naturally, I tried field "title" (const empty) <> defaultContext, but it did not work as expected – I still got a title.

I guess you have figured out by now that empty or noResult just make the <> operator (from Semigroup) go to the next context. And that's also why boolField is the way it is. Do you have a suggestion (or a PR? :) on how we could make the doc clearer on this?

[Minoru]

Hey @kirelagin, did you get a chance to look at the above suggestions?

[Minoru]

@kirelagin, ping.

[kirelagin]

Apologies for the delay. The workaround wrapper sounds pretty good to me – I can’t remember now how exactly the issue manifested for me, but, I suppose, it would work.

As for the docs, I am not exactly sure what the best way to address this would be. Maybe a note/warning could be added to the haddock of the Context newtype, If there was some sort of a Hakyll Cookbook, I would definitely add your withoutField snippet to it.

[Minoru]

Maybe a note/warning could be added to the haddock of the Context newtype,

Dunno. The doc already says that <> "Tries to find a key in the left context, or when that fails in the right context.". If you have a specific wording in mind, please send a pull request.

Hakyll doesn't have a cookbook, and after thinking about it a bit, I don't think I'd add this snippet to one of the existing modules either. So I'll just let it sit here for now, and if a lot of people stumble into this, I can reconsider.