clojure-spec 2018-06-01 | Slack Archive

Is this the proper way to spec a multiple arity function? with :args (s/or ...)

(defn foo
  ([a]    ...)
  ([a b]) ...)

(s/fdef foo
  :args (s/or :arity-one (s/cat :a int?)
              :arity-two (s/cat :a int? :b int?))
  :ret int?)

Alex Miller (Clojure team)13:06:26

I’d use s/alt there instead of s/or as it’s a regex op, but that’s one option

Alex Miller (Clojure team)13:06:47

Another is to push the optionality into the cat with s/?

bostonaholic13:06:35

s/alt is better, thanks

bostonaholic13:06:21

when I run my suite with lein test and I get the generator seed “Using generator seed: -1186346283”, how do I pass that seed into lein test again?

Alex Miller (Clojure team)13:06:38

I don’t know of any way to do that from that level

Alex Miller (Clojure team)13:06:50

stest/check takes an option map where you can pass in a seed

Alex Miller (Clojure team)13:06:57

you might be able to do it through some test.check dynvar or something, not sure

kapil16:06:13

Is there an idiomatic way to associate a doc-string with a spec?

kapil16:06:23

I am trying to generate documentation from spec for non-clojure developers to read. Something like this

(s/def ::id string? "Id of resource X. Ex. resource_1")
(s/def ::num integer? "Number of results for pagination")

(s/def ::result (s/keys :req-un
                        [::id ::num]))

(gen-doc ::result) =>

{:id "resource_1" ;; "Id of resource X. Ex. resource_1"
 :num 10 ;;"Number of results for pagination"
 }

Alex Miller (Clojure team)16:06:20

no, not yet but we intend to add that

kapil18:06:05

@U064X3EF3 Is there any workaround I can do for now? I tried using with-meta but it doesn't with no success since keywords are not Objects. I could however maintain a registry of documentation.

Alex Miller (Clojure team)18:06:15

there are some external libs that are providing this, but it’s hard to recommend those as what they are doing is likely to break

Alex Miller (Clojure team)18:06:39

you could build an external map of spec keyword to doc string

kapil18:06:54

(defonce doc-registry (atom {}))

(defmacro def-w-doc
  [k spec-form doc]
  `(s/def ~'k ~spec-form)
  (swap! doc-registry assoc k doc))

(def-w-doc ::id string? "Id of resource X. Ex. resource_1")

kapil18:06:20

This is what I am using now. This is least intrusive change I could think. Later on when clojure.spec adds support for documenting specs I can just modify def-w-doc.

Alex Miller (Clojure team)18:06:11

sounds good

noisesmith16:06:50

for the seed, one workflow would be to keep your randomized check, and also do another with a specific seed that creates a known rare corner case

bostonaholic16:06:10

SEED=<seed> lein test is how to do it

noisesmith16:06:49

oh that works? interesting

noisesmith16:06:17

I like to have regression tests for things that have been known to fail in the past though

👍 4

noisesmith16:06:57

(I guess it's better to do that by hard coding the data that had been generated in an example based test)

bostonaholic16:06:09

yup, setting the SEED as an env var works

gfredericks13:06:26

FWIW, I have never heard of this and have no idea what's going on

gfredericks13:06:38

my suspicion is it's not a test.check seed, it's something else

cjsauer18:06:39

I'm curious what others' tastes are regarding namespaced keys in spec definitions? As an example, let's say we're modeling GPS coordinates. Personally, I've been doing something like this:

(s/def :coordinate/latitude double?)
(s/def :coordinate/longitude double?)
(s/def ::coordinate (s/keys :req [:coordinate/latitude
                                  :coordinate/longitude]))

In which lat and long are both "nested" under the :coordinate namespace, and then the coordinate map itself is defined inside of some kind of :my-project.specs top-level namespace (using the ::coordinate sugar). Does this make sense, or are there other conventions that have been established that would be clearer?

seancorfield18:06:03

@cjsauer That's pretty close to my approach. Much depends on how unique you need the names to be and how your code might be used.

seancorfield18:06:04

In a library, you'd really want all the names to be globally unique. In an application, you can use less unique names.

👍 8

Alex Miller (Clojure team)18:06:10

they should be sufficiently unique :)

cjsauer20:06:38

@seancorfield good to hear. This is for an application, so I'll trade some uniqueness for readability 🙂

2018-06-01

Channels