#434 Enable Lazy service for phpredis

[B-Galati]

The service will be lazy if symfony/proxy-manager-bridge is installed
As a consequence it will allow to warmup the cache even if Redis server is not reachable

fix #434

[B-Galati]

@dkarlovi Would you be able to test it out ? (or you can just edit your vendors with this little change)

[B-Galati]

I guess I should also edit the doc about symfony/proxy-manager-bridge and phpredis behavior

[B-Galati]

Doc updated

[dkarlovi]

@B-Galati I'll test it tomorrow. Would this be the correct approach?

[B-Galati]

@dkarlovi What do you wean exactly ?

From my knowledge, I do think lazy service is a good solution as it's something easy, powerful and well supported by Symfony.

[dkarlovi]

@B-Galati I mean, would making the service lazy have the desired behaviour and/or bad consequences elsewhere?

From discussions at Symfony's Slack, lazy=true is almost considered an anti-pattern and (some) users called for the feature to be completely deprecated by Symfony.

[B-Galati]

@dkarlovi Thank you. Can you give the reasons behind this thinking? What would be the alternatives?

I am not aware about any bad consequences for lazy services. Certainly we gonna lose some very little performances in some situation but that's all I can see. In other situation where for example, the redis client is not used like the cache warmup, performances will be better.

IMHO lazy services are just an infrastructure detail.

Perhaps I am wrong so please let me know for anything.

[OskarStark]

Symfony just decided to make all doctrine listeners lazy by default in 4.2

[dkarlovi]

@B-Galati sorry for the delay. This does indeed fix #434 (as in, allows the container to be prebuilt). There doesn't seem to be any side-effects that are noticeable to me.

About lazy services, this might have been a "storm in a teacup" by a group of loud users, but the prominent opinion was lazy services are almost always a bad idea. This might be considered an exception. :)

Thank you for working on this!

[B-Galati]

@dkarlovi Thanks for the feedback :-)

Basically PR #424 is doing the same thing but the implementation is different.

[dkarlovi]

@curry684 I see you've merged this, would you consider a bugfix release?

[curry684]

https://github.com/snc/SncRedisBundle/releases/tag/2.1.5

[piotrkochan]

Lazy loading for this service makes PHP 5.6 fail because of invalid cached Redis class. See #465