I have a question related to malli, but it's probably just an error on my part but I cannot spot it: I defined macro as this (simplified example):

(defmacro x [y]
  `(println ~(m/schema [:tuple y])))

Syntax error compiling fn* at (/private/var/folders/fm/5mzhclpd7mj0tzjjq796ftc00000gn/T/form-init14721255528686671623.clj:1:1).
Can't embed object in code, maybe print-dup not defined: malli.core$_tuple_schema$reify$reify__4266@2c3b816

(defmacro x [y]
  `(println (m/schema [:tuple ~y])))
=> #'clj-rest-client.core/x
(x int?)
[:tuple int?]

I wanted to compute the schema value at macro expansion time to make it faster, but it doesn't seem to work for some reason

maybe someone else can spot it, I've been looking at this for a while with no idea

@roklenarcic macros should return source code, the first example returns a reified protocol instance. You can run (macroexpand ...) to that

Just curious, "to make it faster"?

I thought that m/schema does some preprocessing

which would make it faster than sticking simply [:tuple x] in there

btw, it seems that [:tuple] doesn;'t work

(m/schema [:tuple]) -> :tuple but (m/schema [:tuple int?]) -> [:tuple int?]

malli has it's own compiler for validation, explain & transform. Reaults of those should be pretty fast.

Creation of schema has some overhead, but if you can store a reference to a transformer, it should be as fast as one can do with Clojure.

thanks for the info

the :tuple thing... some schemas has a constraint that they should have 1+ childs. I guess tuple doesn't and empty tuple is ok...

I had spec1 code that said -> give me a vector of values and a vector of specs and I will generate a s/fdef spec (which uses s/cat), so the code generated a spec like this (s/cat :arg1 arg-spec1 ...) and potentially (s/cat) which took an empty sequence just fine. I replaced this with :tuple from malli and it seems to have a corner case for empty tuple

should it require at least 1 child?

it works the opposite

(m/explain [:tuple] [])
Execution error (NullPointerException) at malli.core/-tuple-schema$reify$reify$fn (core.cljc:447).
null

but spec with empty (s/cat) would work fine

Oh, that's a bug.

yeah I thought so

NPEs are usually not intended

I'm porting my rest client lib from spec to malli now... hopefully I'll have encode/decode soon

I'd like for user to be able to specify the transformation to string I should do before sending any opaque types over the wire

transforms seem to be doing the opposite direction

e.g. if user has LocalDateTime object I'd like for them to specify format in schema that I format it to before I send it...

https://github.com/metosin/malli/commit/c2f6b45bd12751f18b737cb227d17fae7f024663

encode & decode is done (#99), but requires (#98) before will merge that in.

I'll be finally rid of the nightmare that is trying to conform s/merge specs