Duplicated entries in config when using both #kaocha/v1 and #include

Hey friends!

I was a little confused about how to best write a tests.edn config that uses #include while also relying on the #kaocha/v1 transformer.

Given tests.edn:

#kaocha/v1
 {:capture-output? true
  :color? true
  :plugins [:capture-output :filter :randomize]
  :reporter [kaocha.report/dots]
  :tests [{:id :all
           :ns-patterns ["-test$"]
           :source-paths ["src"]
           :test-paths ["test"]}]}

The docs recommend writing a tests-ci.edn like:

#kaocha/v1
 #meta-merge
  [{:color? false
    :kaocha.plugin.junit-xml/target-file "junit.xml"
    :plugins ^:append [:kaocha.plugin/junit-xml]}
   #include "tests.edn"]

which then prints a config with:

{:kaocha/tests
 [{:kaocha.testable/desc "all (clojure.test)",
   :kaocha.testable/type :kaocha.type/clojure.test,
   :kaocha.testable/id :all,
   :kaocha/ns-patterns ["-test$"],
   :kaocha/source-paths ["src"],
   :kaocha/test-paths ["test"],
   :kaocha.filter/skip-meta [:kaocha/skip]}],
 :kaocha.plugin.junit-xml/target-file "junit.xml",
 :kaocha/fail-fast? false,
 :kaocha/color? true,
 :kaocha/cli-options {:config-file "tests-ci.edn", :print-config true},
 :kaocha.plugin.randomize/seed 997097312,
 :kaocha.plugin.randomize/randomize? true,
 :kaocha/plugins
 [:kaocha.plugin/randomize
  :kaocha.plugin/filter
  :kaocha.plugin/capture-output
  :kaocha.plugin/randomize
  :kaocha.plugin/filter
  :kaocha.plugin/capture-output
  :capture-output
  :filter
  :randomize],
 :kaocha.plugin.capture-output/capture-output? true,
 :kaocha/reporter [kaocha.report/dots],
 :capture-output? true}

Note that the :kaocha/plugins list has duplicates and is missing the junit plugin.

After much trial and error, this tests-ci.edn works (leave off #kaocha/v1 and qualify all of the keywords):

#meta-merge
 [#include "tests.edn"
  {:kaocha/color? false
   :kaocha.plugin.junit-xml/target-file "junit.xml"
   :kaocha/plugins ^:append [:kaocha.plugin/junit-xml]}]

which produces this config:

{:kaocha/tests
 [{:kaocha.testable/desc "all (clojure.test)",
   :kaocha.testable/type :kaocha.type/clojure.test,
   :kaocha.testable/id :all,
   :kaocha/ns-patterns ["-test$"],
   :kaocha/source-paths ["src"],
   :kaocha/test-paths ["test"],
   :kaocha.filter/skip-meta [:kaocha/skip]}],
 :kaocha.plugin.junit-xml/target-file "junit.xml",
 :kaocha/fail-fast? false,
 :kaocha/color? false,
 :kaocha/cli-options {:config-file "tests-ci.edn", :print-config true},
 :kaocha.plugin.randomize/seed 1405326847,
 :kaocha.plugin.randomize/randomize? true,
 :kaocha/plugins
 [:kaocha.plugin/randomize
  :kaocha.plugin/filter
  :kaocha.plugin/capture-output
  :capture-output
  :filter
  :randomize
  :kaocha.plugin/junit-xml],
 :kaocha.plugin.capture-output/capture-output? true,
 :kaocha/reporter [kaocha.report/dots],
 :capture-output? true}

I don't know exactly how exactly the meta-merge works or when/how the conversion of plain keywords to qualified keywords works, but it seems that there's something funky happening and the docs should probably be updated.

Thanks so much!

Reproduce the configuration using the issue's tests.edn and tests-ci.edn examples, then trace how #kaocha/v1, #include, and #meta-merge are handled. Done means the combined configuration does not duplicate plugins and includes the JUnit plugin, or the documentation clearly explains the supported form.