spec/conform: + vs * differ sometimes, not sure why

[Cole Frederick]

Hi to anyone reading,

I'm trying to understand a difference in behavior when using clojure.spec/conform on two different regex ops: + and *.

They often conform the same when there is at least one item, but not always.

One small case I've found is:

(s/def ::evens-or-odds (s/alt :evens (s/+ even?)

                              :odds  (s/+ odd?)))

(defn conform-test [nums]

  (let [star (s/conform (s/* ::evens-or-odds) nums)

        plus (s/conform (s/+ ::evens-or-odds) nums)]

    [(= star plus)

     star
     plus]))

(conform-test [1 2 3]) =>

[true

 [[:odds [1]] [:evens [2]] [:odds [3]]]

 [[:odds [1]] [:evens [2]] [:odds [3]]]]

(conform-test [1 2 2]) =>

[false

 [[:odds [1]] [:evens  [2 2]]]

 [[:odds [1]] [[:evens [2 2]]]]]

(conform-test [1 2 2 3]) =>

[false

 [[:odds [1]] [:evens  [2 2]] [:odds [3]]]

 [[:odds [1]] [[:evens [2 2]] [:odds [3]]]]]

They conform the same for the first, but on the second and third, the + regex wraps all but the first group in a vector. I believe it has something to do with nested regex ops, but I haven't figured out the pattern.

Same conform value:

[1 2]

[1 2 3]

[1 1 2 3]

Different conform values:

[1 2 2]

[1 2 2 3]

[1 2 3 3]

I appreciate your time and any insight or ideas about why these sets of cases conform differently.

Thanks,

Cole

[Alex Miller]

This is a bug, similar to https://dev.clojure.org/jira/browse/CLJ-2105 and http://dev.clojure.org/jira/browse/CLJ-2003. I've spent a little time on it but have not figured out the exact problem. If you'd like to file it and reference those and/or drop it as a comment on one of those, would be happy to have more cases to verify when it's fixed.

[Cole Frederick]

Awesome. Can do!

Thank you.

[Peter Hull]

On Sunday, July 30, 2017 at 6:55:04 AM UTC-6, Alex Miller wrote:

This is a bug, similar to https://dev.clojure.org/jira/browse/CLJ-2105 and http://dev.clojure.org/jira/browse/CLJ-2003. I've spent a little time on it but have not figured out the exact problem. If you'd like to file it and reference those and/or drop it as a comment on one of those, would be happy to have more cases to verify when it's fixed.

I notice that Tyler Tallman has posted a patch for CLJ-2003 recently (https://dev.clojure.org/jira/browse/CLJ-2003). Would be worth a look to see if this solves your problem and also CLJ-2105.

Pete