Integrant elides Midje metaconstants from config

[weavejester]

Integrant elides Midje metaconstants from config

[weavejester]

Sorry, I don't understand what you're asking.

You've created a failing test, and Midje tells you that it's failed. Is there some additional information in the failure message that you expect to see?

[WhittlesJr]

My bad, I should have been more specific. My expectation was that the test should pass as written, but instead the metaconstant is turned into an empty map. I'm trying to look around to see where this would happen.

I noticed that the config's ::origin metadata does retain the metaconstant as given:

{:integrant.core/build {:some/component {:config-key {}}}
 :integrant.core/origin {:some/component {:config-key ..value..}}}

[WhittlesJr]

er, sorry, I see your confusion now better. The fact that my init-key method doesn't actually return anything useful is seemingly incidental to the problem of metaconstants disappearing. I'll try to come up with a better example

[weavejester]

Your init-key method for :some/component returns nil, so running init will produce:

{:some/component nil}

This seems in line with the failure message you're getting.

[WhittlesJr]

Here, this is more what I'm getting at:

(defmethod ig/init-key :some/component
  [_ config]
  (:config-key config))

(fact
  (ig/init {:some/component {:config-key ..value..}})

  => {:some/component ..value..})

Produces:

Error Metaconstants (..value..) can't be compared for equality with {}.

Whereas the following passes:

(fact
  (ig/init {:some/component {:config-key "value"}})

  => {:some/component "value"})

[weavejester]

Isn't this a problem with Midje? Integrant doesn't produce anything unusual; the output of init is an immutable map with some metadata attached to provide halt ordering.

[WhittlesJr]

I'm really not sure. In practice, my init-key and halt-key! multimethods all only have a single call to an external function that handles all the config destructuring and logic. So since Midje works just fine with those regular functions, my test coverage probably probably doesn't need to include my Integrant multimethods.

I just thought it was strange that the metaconstants would behave this way when passed through ig/init.
Stranger yet, even if you put the defmethod inside of a fact macro, this still happens.

Seems like a limitation that could be avoided somehow, but I'm not sure if I should be asking you or @marick

[weavejester]

I don't think this is an issue with Integrant. init is a function that takes in a map and returns a map with metadata - there's nothing particularly unusual about it, and a test library should be able to handle that. I'd suggest opening this issue on the Midje project.

[WhittlesJr]

Alright, I'll do that. Thank you!

[WhittlesJr]

Well, it seems the issue is not easily fixed (if it's even possible to fix it). I still don't understand why this is happening, but I'll just give up and work around it.