Hi All, I have been trying to write a spec recently and I can't quite describe what I want... I have a map where two sets of keys depend on each other like this:

{:a "string"
   :b :value-b
   :c :value-c}

{:a "string"
  :b :value-not-b
  :c :value-not-c}

( try three backticks to format that without emojis)

thank you @manutter51

so I either have to first version of the second one.

I assume I need something with an or and an and but how exactly?

I’m fairly shaky on spec still, but I’m thinking maybe you could take advantage of namespaced vs. unnamespaced keys, let me see if I can type up what I’m thinking

thank you... I don't get it either yet I'm 'fraid

(s/def :any/a int?)
(s/def :is-value/b #{1 2 3})
(s/def :is-not-value/b #{4 5 6})
(s/def :is-value/c #{1 2 3})
(s/def :is-not-value/c #{4 5 6})
(s/def :is-or-not/my-map
  (s/or :is (s/keys :req-un [:any/a :is-value/b :is-value/c])
        :is-not (s/keys :req-un [:any/a :is-not-value/b :is-not-value/c])))

I’ve got a suspicion that won’t work, because maps are spec’ed by key, not by value, but a perverse part of my brain is saying maybe it would, because the keys are defined to take two mutually exclusive sets.

let me give that a try @manutter51

I just did a few generates on that and that looks like it does work!

Generates seem like they’d be more likely to succeed, but I’m not as confident about conforms with “bad” data.

maybe, though? 🤞

well... I assume if it generates the correct data, then the other way round should work as well.

but let me try...

yes, validating works as well as expected.

Thank you again @manutter51!!!

That’s pretty wild, I’ll have to remember that one myself.

a good trick to use the namespaces for that... I wouldn't have thought of that.

I suspect it is something that happens more often.

I think you could also combine your s/keys spec using s/and with another predicate, and do any "custom" assertions in the extra predicate