Limitation on using conformed values limits usefulness of some `s/and` specs about expound HOT 8 CLOSED

Ah, good point, that's not a great experience for someone working through the official docs 😞 . Thanks for pointing that out 😄

Certainly defaulting to normal output is one option I'm considering. I'm not sure if falling back on normal explain should be the default or something you can opt into with a configuration. Before I realized that and had this issue, I was leaning towards "you must opt in" since using specs for coercion was an advanced use case and not recommended for various reasons.

But given that this more common (and, as you point out, included in the official docs), it might make sense to make falling back to normal explain the default. Or it might be possible to print a reasonable error in this case by looking for the the conformed value.

In any case, the workaround for now is to wrap the printer in a try/catch and default to s/explain-printer e.g.

(require '[expound.alpha :as expound])

(defn safe-printer [explain-data]
  (let [exp (expound/custom-printer {})]
    (try
      (exp explain-data)
      (catch Exception e e
             (s/explain-printer explain-data)))))

(set! s/*explain-out* safe-printer)

(s/def ::sorted-pair (s/and (s/cat :x int? :y int?) #(< (-> % :x) (-> % :y))))

;; Usually use expound
(s/explain int? "a")
;;-- Spec failed --------------------
;;
;;  "a"
;;
;;should satisfy
;;
;;  int?
;;
;;-------------------------
;;Detected 1 error

;; If there is an issue, fallback on s/explain-printer
(s/explain ::sorted-pair [0 0])
;; val: {:x 0, :y 0} fails spec: :expound.alpha/sorted-pair predicate: (< (-> % :x) (-> % :y))

from expound.

@bhb I think you can rely on both conform and unform working for a "pure" and implementation since those are standard transformations: https://github.com/clojure/spec.alpha/blob/739c1af56dae621aedf1bb282025a0d676eff713/src/main/clojure/clojure/spec/alpha.clj#L1136

However, it is of course possible that a user just mixed conformers directly with an s/and call; I am guilty of this btw 😅.

Generally I dont create an unform function for those cases but just put identity to "get the job done". You would have a hard time getting back the values unformed.

So I guess that you would need some kind of heuristics to figure out if you are able to unform the values or if you just default to the classical output. Btw, if you default to the classical output, would it be possible to still have a warning printed before it mentioning that you are getting the default output due to the unform part? otherwise the user would be left wondering if expound is actually working or not.

Hope it helps

from expound.

@carocad Thanks very much for trying it out! I appreciate it.

I think there are some cases where the output could be slightly more accurate but those are pretty much nice to have

Yes, you're right: Expound could do more here, but since string parsing is advanced at best, and discouraged at worst, I'm not inclined to add much more here. I may change my mind in the future, but I think this is sufficient for now 😄

from expound.

Sadly, it means expound fails on the very first example of instrumentation in the official spec guide use s/def (the ranged-rand example). I wonder, would it be possible to drop back to the original explain in cases where expound fails? That'd be better than just getting an exception.

from expound.

I wonder if I could look for both the conformed and unconformed value in the original data, then show both. Something like:

[1 2]

when conformed to

{:x 1 :y 2}

should satisfy

#(< (-> % :x) (-> % :y))

I this more about this.

from expound.

@carocad @jmlsf I've got a fix for this issue in 0.7.2-SNAPSHOT. If you have time to try it out, please let me know what you think!

from expound.

@bhb first of all, I have to say "Thanks a lot for putting the effort of figuring this out"

I tried the following cases and they all worked wonderfully.

I think there are some cases where the output could be slightly more accurate but those are pretty much nice to have; here I am referring to the first case 'string parsing', where it is clear from the output that there is a single value not a collection so the text part of the value doesn't really fit. Unfortunately I have stopped using conformer since they were creating too many problems so I cannot test my previous errors anymore 😅

from expound.

Fixed in 0.7.2-SNAPSHOT (to be released soon)

from expound.