ClojureScript support for Kaocha.
;; deps.edn
{:deps {lambdaisland/kaocha {...}
lambdaisland/kaocha-cljs {...}}}
Note that you must be using at least Clojure 1.10.
;; tests.edn
#kaocha/v1
{:tests [{:id :unit-cljs
:type :kaocha.type/cljs
;; :test-paths ["test"]
;; :cljs/timeout 10000 ; 10 seconds, the default
;; :cljs/repl-env cljs.repl.node/repl-env ; node is the default
;; :cljs/repl-env cljs.repl.browser/repl-env
}]}
For nodejs, install ws
.
npm i ws
Run your tests
clojure -m kaocha.runner unit-cljs
:kaocha/source-paths
(or :source-paths
when using #kaocha/v1
) :kaocha/test-paths
(or :source-paths
when using #kaocha/v1
) :cljs/timeout
:cljs/timeout
seconds, the test fails. This also causes
subsequent tests are skipped, because we assume the ClojureScript runtime is
no longer responsive.:cljs/repl-env
cljs.repl.node/repl-env
cljs.repl.browser/repl-env
figwheel.repl/repl-env
:cljs/compiler-options
{}
.:cljs/precompile?
cljs.build.api/build
before launching the
REPL. Certain REPL types like Figwheel REPL require an initial build before
the REPL is able to connect. If this is the case you can set this to true
.
Defaults to false
.The :test-paths
do not get automatically added to the classpath (at least not
in a way that makes the sources visible to ClojureScript), so you need to also
have any :test-paths
in your project.clj
/deps.edn
/build.boot
.
This is a discrepancy with regular Kaocha, where you only need to specify the test paths once.
On Linux the cljs.repl.browser/repl-env
requires the browser process to already be started
before running Kaocha (see: https://clojure.atlassian.net/browse/CLJ-2493).
To support running browser tests on CircleCI add an early config step like:
- run:
command: /usr/bin/google-chrome-stable --no-first-run
background: true
Most ClojureScript testing tools work by building a big blob of JavaScript which contains both the compiled tests and a test runner, and then handing that over to a JavaScript runtime.
Kaocha however enforces a specific execution model on all its test types.
[config] --(load)--> [test-plan] --(run)--> [result]
Starting from a test configuration (e.g. tests.edn
) Kaocha will recursively
load
the tests, building up a hierarchical test plan. For instance
clojure.test
will have a test suite containing test namespaces containing test
vars.
Based on the test plan Kaocha recursively invokes run on these "testables", producing a final result.
During these process various "hooks" are invoked (pre-test, post-test, pre-load, post-load), which can be implemented by plugins, and test events (begin-test-var, pass, fail, summary) are generated, which are handled by a reporter to provide realtime progress.
Kaocha's built-in features, plugins and reporters are rely on this model of execution, so any test type must adhere to it. Note that all of this is on the Clojure side. Kaocha's own core, as well as plugins and reporters are all implemented in (JVM-based) Clojure, not in ClojureScript, so even in the case of ClojureScript tests the main coordination still happens from Clojure.
To make this work kaocha-cljs makes use of a ClojureScript PREPL (a programmable
REPL). Given a certain repl environment function (e.g. browser/repl-env
or
node/repl-env
) Kaocha will boot up a ClojureScript environment ready to
evaluate code, and load a websocket client that connects back to Kaocha-cljs, so
we have a channel to send data back from ClojureScript to Kaocha. It will then
send code to the PREPL to load the test namespaces, and to invoke the tests.
Anything written on stderr or stdout will be forwarded to Clojure's out/err streams, and possibly captured by the output capturing plugin.
The test events produced by cljs.test
(pass, fail, error) are sent back over
the websocket, and ultimately handled by whichever Kaocha reporter you are using.
Events received from the PREPL and the websocket are all placed on a queue, which ultimately drives a state machine, which coordinates what needs to happen next, and gathers up the test results.
If you're having issues, first try running with --no-capture-output
. There may
be relevant information that's being hidden.
To see all messages coming in over the PREPL and Websocket you can set
kaocha.type.cljs/*debug*
to true
. You can do this directly from tests.edn
.
#kaocha/v1
{:tests [,,,]
:bindings {kaocha.type.cljs/*debug* true}}
This will also set the goog.log
root logger, and the
kaocha.cljs.websocket-client
logger both to the DEBUG
level. Have a look at
glogi for more information about Google
Closure's logging facilities.
When not using *debug*
you can still set these log levels seperately through
:closure-defines
.
#kaocha/v1
{:tests [{:type :kaocha.type/cljs
:cljs/compiler-options {:closure-defines {kaocha.type.cljs/log-level "ALL"
kaocha.type.cljs/root-log-level "INFO"}}}]}
;; Log levels:
;; OFF SHOUT SEVERE WARNING INFO CONFIG FINE FINER FINEST ALL
Copyright © 2019 Arne Brasseur
Available under the terms of the Eclipse Public License 1.0, see LICENSE.txt
Can you improve this documentation? These fine people already did:
Arne Brasseur, Mark Ingram & Daniel ComptonEdit on GitHub
cljdoc is a website building & hosting documentation for Clojure/Script libraries
× close