Clojurians Log v2

thanks for looking, really appreciate your input and help 👍

just been reading spec-tools sources, looks like this has already been considered, as there's fail-on-extra-keys-type-decoders in addition to strip-extra-keys-decoders. Although there's a TODO that makes e wary of directly using that: https://github.com/metosin/spec-tools/blob/master/src/spec_tools/transform.cljc#L113 Out of curiosity, what's the rationale why strip variant was chosen as default for c-api? I think it's great to be able to take request data, convert to EDN/clojure and run it through the spec at the repl; I've been assuming that giving a parameter spec as the metadata for :body-params or :parameters would codify the validity of the request/response, the stripping of extra-keys seems to break that invariant? thanks!

erk, my bad (not= fail-on-extra-keys anything-to-do-with-stripping)

fwiw, I've "fixed" this by: https://github.com/mgrbyte/capi-spec-transforms/commit/47736063b20a1c373140cbaea98acdc6d59a1e54 Hopefully that is "idiomatic" (for some value of idiomatic! 🤔)

@mgrbyte The fix looks good.

about the default of dropping keys… there was a discussion about the default long time ago. Can’t recall why that was chosen. I think at least the option to pass-through all values easy to do. Would you like to do a PR where the spec-coercion could be easily modified to do any of the: strip, fail or nothing?

still alpha, so we can re-visit the decision too.

I will look into doing a PR, possibly this evening. think if it's easy to change, that'd be good. I think the default should be "less surprise" - guessing for folks who are using spec that'd mean not dropping keys, but 🤷

the stripping… sadly there is no extra info emitted if the spec fails on extra keys, just ::s/invalid. there is https://dev.clojure.org/jira/browse/CLJ-2115 and https://dev.clojure.org/jira/browse/CLJ-2251, but both 1.5y old without any progress.

thanks for the links there.

So it looks like my custom coercion wasn't good enough/the complete story

Now values are not being transformed 😞

keywords within map spec specifically

I'm guessing I have to (re)implement the entire Coercion protocol?

keyword leaf values?

yeah

the leaf specs need to be "decodable", which means Spec Records fow now. there are versions in soec-tools.spec for most/all predicates

this works when I use :coercion :spec in api

e.g. spec-tools.spec/keyword?

Oh, hmm.

I can switch between the new custom coercion and the default (`:spec`) and observe the transform works in the later, but not the former

restarting the app in between of course

thats weird.

Could you pr the curent status into the repo?

the example repo or my "real" app?

will try and dup in the example repo, less to look at

just pushed https://github.com/mgrbyte/capi-spec-transforms/commit/38739ade1ddec611ef44c1c2ac654787e87e142e

have to get a 🚌 home now, but will be looking again later tonight to see what I can do. thanks again for taking interest! 😍

https://github.com/mgrbyte/capi-spec-transforms/pull/1

@mgrbyte run the code:

default-transformer-stripping-data-issue.handler=> (-> ::variant-a s/spec :keys)
#{:g/cn :p/how :g/bt :g/sp :p/who :p/when :p/why :g/sn}
default-transformer-stripping-data-issue.handler=> (-> ::variant-b s/spec :keys)
#{:g/cn :p/how :g/sp :p/who :p/when :p/why}
default-transformer-stripping-data-issue.handler=>

… with extra-keys strpping:

default-transformer-stripping-data-issue.handler=> (st/conform ::update {:g/sp "s1", :g/cn "cn1", :g/bt "XXXX"} st/strip-extra-keys-transformer)
[:default-transformer-stripping-data-issue.handler/variant-a #:g{:cn "cn1", :sp "s1"}]

so you are right: c-api first strips out keys that are not defined and then checks the keys => the invalid keys are removed and it succeeds.

Normally, this would be good, here - I guess not.

thanks @ikitommi - i merged and tested, still doesn't work for me though 🤔 I'm an idiot, made a typo that stopped it working

fwiw, I've just pushed the final working version to master: https://github.com/mgrbyte/capi-spec-transforms/commit/432ed6b3011d2eda2e193a550e6e8062fe932a80 There was one additional issue, in that the (repeat json-transformer) code I had in options passed to create-coercion wasn't working, but fixed that easily enough.