Add user option to Redis adapter to authenticate with a specific user

[vcg-development]

Q	A
Documentation	no
Bugfix	no
BC Break	no
New Feature	yes
RFC	no
QA	no

Description

Description
This PR is adding the possibility to authenticate with Redis with a custom username. The ACL feature is available in Redis since Version 6.
The username can be passed as an resource option.

[boesing]

Sorry for the late feedback, I didn't felt like having a look into it.
However, thanks for this contribution. I wonder if it is also available for RedisCluster but I do not have a problem if that is a Redis-only feature for now, can be implemented for RedisCluster once requested.

I do have some nitpicks and some change requests.

• especially the native return types and argument types should be added
• we should have a test which verifies that the user can be extracted from the server URI

I think its time to have a proper resource array shape declared in the RedisResourceManager so that we do not have that weird array with "whatever" array keys in it. Once I find some time, I might implement that myself and push it to this patch - if you think you want to that on your own, I'd be happy to review that change again.

We recently introduced SSLContextArrayShape in SslContext. You can take that as an example on how to declare array shapes if you do not have an idea on how that might work.

Once we defined the RedisResourceArrayShape, we can use that for RedisResourceManager::$resources. That will most probably remove a bunch of baseline entries and errors within the RedisResourceManager so that would be a huge improvement.

[vcg-development]

Thank you very much for the detailed review. I changed the code according to your comments.

I also added the suggested ResourceArrayShape. I`d be happy, if have another look at my changes.

[boesing]

Nice, thanks for updating and providing proper type-hints for the new methods!
I put some suggestions and comments to your latest changes regarding the array shape. If you apply these changes, I guess psalm will stop reporting some errors.

You might want to have a look on these psalm issues as they might tell you why some types may probably not correct 👍🏼

If you need help, just let me know.

[vcg-development]

Thank you again for having a deep look at my commits again. The types declared in the array shape and related return types should be correct now.
I thing the failing psalm checks come from the array_merge in line 416. At the moment I am not sure how this could be fixed, so that they will not fail anymore.

Also I think the Assertment in Line 177 is not necessary anymore, because the instance is already checked in line 169.

[boesing]

Yes you are right, one of the issues were due to the array_merge (psalm has issues with that while array_replace does work. In this case, array_replace should suffice and thus I changed the underlying code.

I also removed the fixed issues with psalm from the baseline by updating the baseline using psalm --update-baseline. That will automatically remove all fixed issues which were fixed by this PR.

I will have a closer look on the integration tests, it seems that the password test is actually introducing problems when used on a server without a configured password.

[boesing]

Okay, that was a tough one. I was wondering why stuff started to went wrong after my changes.
I found out that in one of my changes, I removed the property reference which had side-effects in some tests.

Another good reason on why having these property references removed from our code.
I've created #93 so that we can tackle further refactorings in v3.0.0.

[boesing]

Thanks @vcg-development and @cbraunVBM!
I will create an issue in laminas-cache so that we can document the new option as that has to be done there.