Minimal repro case:

(def schema [:map-of :string :int])

(malli/coerce schema {"foo" 1 "bar" :baz} malli.transform/strip-extra-keys-transformer) ; => {"foo" 1}

Now I don't know if there's a right or wrong way to think about this, maybe both perspectives are valid but I'd like to verify my understanding of how coercion of map-of is supposed to behave.

I think there's a discussion of this in a ticket

just a sec

I guess it was implemented in this PR but I can't see the discussion I'm remembering

https://github.com/metosin/malli/pull/871

Perhaps it was me&Tommi in a chat somewhere...

this is the commit https://github.com/metosin/malli/commit/0e5b5e3ac56479f75c812c1320ac1f1ebfba8f76

anyway, what you're seeing is the intended behaviour: all entries that don't validate are stripped

Thanks for all the context!

Given that it is the intended behavior, and I think that's a perfectly valid choice, what would be the way to opt-out of it? I would guess a custom transformer, right?

Yeah that's certainly one way

I'm thinking about some way to override it case-by-case but I didn't figure out anything yet

I think there are at least 3 good options: 1. per schema override, e.g. ::mt/strip-extra-keys false property 2. per transformer, like 1 but as a option for mt/strip-extra-keys-transformer 3. both of 1 & 2

would this work ? [:map-of keyword? [:fn var?]]

oh yeah, interesting, let me try that

That works great, idk why I didn't think of it. Thank you!