1:22 p.m.
On 1/16/2019 12:58 PM, Guillaume Vincent wrote:
> Coverage is what matters.
I like to tests features and add regression tests for each bugs.
Coverage is not a good metric, you can call your function with no
assertion, you will have 100% code coverage.
But I think we are agree that every function in the adapter (init(),
login() refreshToken(), etc) should be tested.
Yeah, I'm not talking about code
coverage or path coverage. I'm talking
about coverage in a broader sense, as in, "the test suite covers
everything the system is designed to do".
> should be written in TypeScript.
Totally agree, as I said I created the experience in a couple of hours.
But yes JS adapter should be written in typescript
I see 2 strategies to implement tests and upgrade the JS adapter:
1/ write a new lib in Typescript near the old one, add some tests for
every function and test new and old implementation together with
automated tests.
2/ split actual code (each method in a file for example) set up a
blunder to build the js adapter and then add unit and functional tests
one by one for every function. Then migrate code to Typescript
I can try a POC on the first one if you want
Somebody did a POC as part of this
project:
https://github.com/ebondu/angular2-keycloak
The idea was that we would end up with both a javascript adapter and a
TypeScript adapter. I was opposed to that at the time. It's easy
enough to just wrap the javascript adapter and use the type definition
file for TypeScript.
But there would be advantages to */replacing /*the javascript adapter
with TypeScript. One is that we would no longer need a type def file
that gets out of sync with the javascript adapter. And you would get
all the usual TypeScript goodness as we further develop the adapter. It
would result in better, more maintainable code.
The main technical hurdle would be that the interface would need to be
backward compatible to support users who currently have code calling the
javascript adapter directly. We would need to continue supporting code
that calls the adapter with plain javascript.
Stian, can you think of technical reasons why we should not do this? I
know there are probably many non-technical reasons.
On Wed, Jan 16, 2019 at 5:43 PM Stan Silvert <ssilvert(a)redhat.com
<mailto:ssilvert@redhat.com>> wrote:
On 1/16/2019 9:53 AM, Guillaume Vincent wrote:
> Hello Keycloak dev list,
>
> in a previous post I raised the problem that the JavaScript
adapter did not
> have JavaScript tests.
>
> In a couple of hours I created a simple example for Keycloak
with unit and
> functional tests https://github.com/guillaumevincent/keycloak-lite
>
> You can see tests in this file
>
https://github.com/guillaumevincent/keycloak-lite/blob/master/test.js
>
> I also created a blog post on IMO How to test JavaScript code:
> https://guillaumevincent.com/2019/01/15/test-in-javascript.html
>
> Maybe we can open the discussion on how keycloak.js should be
tested.
> Without any fast and automated tests, in JavaScript, the
refactor of the
> keycloak adapter will not be easy at all.
>
> wdyt?
Several thoughts:
* Basically, I agree. It makes sense to test javascript with
javascript. I like where you are going with this.
* An important point in any discussion of testing is that the only
useful test is a test that uncovers a bug. We never write tests
just to say we have lots of tests. Coverage is what matters. I'm
not criticizing your blog. It's just something I like to keep
in mind.
* You mention TypeScript in your blog, but test.js appears to be
written in plain javascript. IMO, any javascript we write
(with the
possible exception of keycloak.js) should be written in
TypeScript.
Both internally and externally, developers are moving more and
more
to TypeScript. Also, the Java developers on our team will be much
more comfortable and confident with a strongly typed language that
works well with an IDE.
* We need to know a little more about the current test coverage
of the
javascript adapter. Much of it is tested through indirect means.
* We need to understand how javascript tests will integrate into our
builds.
* We need to standardize on a javascript test package. I don't want
the adapter to be tested with one library while the new account
management console is tested with another.
_______________________________________________
keycloak-dev mailing list
keycloak-dev(a)lists.jboss.org <mailto:keycloak-dev@lists.jboss.org>
https://lists.jboss.org/mailman/listinfo/keycloak-dev
--
Guillaume Vincent
Senior Software Engineer