Make schema?, into-schema? open to extensions

[vemv]

Hi there,

in Malli 0.8.3 it was easier to implement custom Schema types.

In 0.8.4, protocol extension effectively became a closed thing, since custom implementations would be omitted by schema? / into-schema?, and therefore also omitted by dependent logic.

I'd be much thankful if it was kept open, which would allow me to continue experimenting with Var support (which so far has been a success and a justified venture).

In general, other users might have other interesting ideas to implement, so leaving the protocol open doesn't seem too crazy.

I'm aware satisfies? has a higher cost than instance?, however satisfies? is far from being implemented naively AFAIK.

Cheers - V

[ikitommi]

Sadly satisfies? is really slow, 20-30x slower than the Instance check and don't want to take the penalty. See #513 for more details on the perf stuff.

But, I think we can have both: first doing the fast instance check and fallback to satisfies?. Could you change the code to do that?

[ikitommi]

fast & extendable:

(defn schema?
  "Checks if x is a Schema instance"
  [x] (or (#?(:clj instance?, :cljs implements?) malli.core.Schema x)) (satisfies? Schema x))

[vemv]

Seems pretty neat! Thanks.

However this only results for what we can call the "green path" e.g. x was a malli.core.Schema.

If x belonged an unrelated type e.g. 42, then the cost of both or branches would be inflicted, for nothing.

One possible fix would be to reshuffle the couple conds that use schema?:

malli/src/malli/core.cljc

Lines 1974 to 1976 in 68bb521

	(schema? ?schema) ?schema
	(into-schema? ?schema) (-into-schema ?schema nil nil options)
	(vector? ?schema) (let [v #?(:clj ^IPersistentVector ?schema, :cljs ?schema)

By placing the schema? check after the vector? check, one surely most code paths won't ever hit the satisifes?.

[vemv]

Another option would be to make the extra check dependent on a compile-time condition:

(defmacro maybe-slow-schema-check [x]
  (if (System/getProperty "malli.maybe-slow-schema-check")
    `(satisfies? Schema ~x)
    false))

(defn schema?
  "Checks if x is a Schema instance"
  [x] (or (#?(:clj instance?, :cljs implements?) malli.core.Schema x)) (maybe-slow-schema-check x))

This would work with my use case since I only use protocol extensions within a specific dev-time task - not as a general assumption in my production codebase.

Of course it's not as general of a solution but it might be a worthwhile alternative.

[ikitommi]

Thanks for the suggestions.

There is malli.perf.creation-perf-test which has utilities for testing the perf gains / losses.

This code:

(defn validator []
  (m/validator
   [:map
    [:x boolean?]
    [:y {:optional true} int?]
    [:z [:map
         [:x boolean?]
         [:y {:optional true} int?]]]]))

currently calls the schema? and into-schema? with the args that don't match the predicate:

"schema?" [:map [:x #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]] [:y {:optional true} #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]] [:z [:map [:x #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]] [:y {:optional true} #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]]]]]
"into-schema?" [:map [:x #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]] [:y {:optional true} #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]] [:z [:map [:x #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]] [:y {:optional true} #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]]]]]
"into-schema?" :map
"schema?" #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]
"into-schema?" #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]
"schema?" #IntoSchema{:type boolean?}
"schema?" #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]
"into-schema?" #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]
"schema?" #IntoSchema{:type int?}
"schema?" [:map [:x #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]] [:y {:optional true} #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]]]
"into-schema?" [:map [:x #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]] [:y {:optional true} #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]]]
"into-schema?" :map
"schema?" #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]
"into-schema?" #object[clojure.core$boolean_QMARK_ 0x53667cbe "clojure.core$boolean_QMARK_@53667cbe"]
"schema?" #IntoSchema{:type boolean?}
"schema?" #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]
"into-schema?" #object[clojure.core$int_QMARK_ 0x21ca93f9 "clojure.core$int_QMARK_@21ca93f9"]
"schema?" #IntoSchema{:type int?}

perf:

(p/bench (validator))

• currently: 1.3µs
• just satisfies?: 90µs
• or: 90µs
• or + vector? check first: 50µs

to make it fast and extendable, we would have implement Schema and IntoSchema for the default case and use some tagging methods to actually tell if it satisfies the protocol or not. Sieppari uses that. Malli had an elegant hack for caching the satisifies? call results, hinted by Alex. Could be re-introduced. Or some other way to write extensions for classes.

lot of work that could be easily fixed in clojure.core, go vote the issue up :)

[vemv]

go vote the issue up

Thumbed!

Hope some sort of fix can make it into Malli in the meantime. Feel free to close this PR if you think the final approach would look too different (I'm not familiar with the tagging technique at all).

For now I'll use some monkey patching but I'd sure love something that felt more dependable.