[RFC] Remove coupling to SonataAdminBundle

[core23]

This might be a controversial topic, but what do you think about removing the hard backend dependency to the SonataAdminBundle from our feature bundles and create new backend bundles (e.g. SonataNewsAdminBundle)? This would help to decouple our projects more and more.

The following projects are affected:

• SonataClassificationBundle
• SonataCommentBundle
• SonataMediaBundle
• SonataNewsBundle
• SonataUserBundle

:+1: Pro

• We have lightweight bundles
• Easier extension
• Creating API driven bundles (interop with EasyAdmin, api-platform, ...)
• Faster release process
• Getting rid of require-dev and conflicting dependencies

:-1: Contra

• Doing a major release of a non-admin major release (e.g. Entity change) requires a major release of the admin release
• More projects
• More maintenance / release overhead
• Requires active communicate for the next major release ("Why is my feature now shown inside the admin dashboard?")

[VincentLanglet]

It would be interesting to know which % of sonata users are using those bundle without any interest in the AdminBundle. My personal opinion is "none or not a lot". This extra work, extra projects, extra maintenance, ... would be only useful for those users ; so I don't see a real interest.

This "remove coupling to SonataAdminBundle" was already started with the creation of the DatagridBundle.

• The Datagrid bundle will contains the classes needed by the backend like Datagrid, Pager, Filter, ProxyQuery
• When using a feature bundle like SonataMediaBundle without the SonataAdminBundle, only the DatagridBundle will be needed (for instance for the REST API)

This now have tons of issues:

• DatagridBundle and AdminBundle have some duplicated classes
• Worst, DatagridBundle and AdminBundle duplicated classes are slightly different/incompatible
• DatagridBundle is stick to a persistence bundle since it provide some ORM implementation of the interface
• Worst again, lot of improvements/fixes done in SonataDoctrineORMAdminBundle are not present in the DatagridBundle.

In a perfect world, we would have

• A DatagridBundle
• A SonataAdminBundle who require DatagridBundle
• A DatagridORMBundle who require DatagridBundle
• A SonataORMAdminBundle who require DatagridORMBundle and SonataAdminBundle

Controversial opinion: this is too much work/maintenance for the maintainers we have with too low benefit for the users we have.

Here, if I understand correctly for the SonataMediaBundle (and the other), we would have

• SonataMediaBundle with entities
• SonataMediaRestBundle with Apis, requiring SonataMediaBundle and SonataDatagridBundle
• SonataMediaAdminBundle with Admin, requiring SonataMediaBundle and SonataAdminBundle

It's a nice idea on the paper. But I wonder if this would be really useful for the community we have. It could be interesting to create a poll on twitter with the sonata-project account in order to know which % of users are not using the AdminBundle.

Also if we plan to keep splitting bundle, a mono-repository would be better.

[core23]

Controversial opinion: this is too much work/maintenance for the maintainers we have with too low benefit for the users we have.

As we have a lack of active core members at the moment, we should not spend any more time on this issue.

Off topic: I wonder if sonata is still a fancy project that is used for new projects / developers or if EasyAdmin is more famous. Symfony stopped listing us on their project list and it feels like we core developers are the only people that open new pull requests and fix issues.

[VincentLanglet]

Off topic: I wonder if sonata is still a fancy project that is used for new projects / developers or if EasyAdmin is more famous. Symfony stopped listing us on their project list and it feels like we core developers are the only people that open new pull requests and fix issues.

The list from Symfony is pretty reduced, I assume they removed sonata because they want to promote EasyAdmin. I merged multiple bug fix coming from non-core-developer and if we look at the stats it seems used a lot. https://packagist.org/packages/sonata-project/admin-bundle/stats

The fact we didn't support Symfony 5 for a long time hurt sonata. New projects/developers had to use Symfony 4 in order to use SonataAdmin (that's why we got issue about sf5/twig3).

But the stable release of the v4 should come soon.