Potential performance improvement: Turn partial publish into full publish if outcome is the same

[bwaidelich]

TL;DR: A partial publish is much slower and more complex than a full publish of a workspace – we can detect if a partial publish already contains all pending events and turn it into a full publish in this case.

For Neos the default publishing mode is to only publish events of the currently selected document (IMHO a full publish should be the default but that discussion is out of scope!).
A partial publish is much more complex than a full publish because we first have to split affected and remaining events and publish those to different newly forked streams...

During the splitting of those events we could handle the PublishIndividualNodesFromWorkspace like the corresponding PublishWorkspace command if $remainingCommands is empty.

The same should be true for DiscardIndividualNodesFromWorkspace vs DiscardWorkspace

Related: #4388

[mhsdesign]

jup @kitsunet and me had the same thoughts :) it will be implemented via #5301

[mhsdesign]

Funnily after thinking about this #5301 doesnt really solve this case really. As a full publish will with that pr also do the simulation and apply the commands. We only said "do a full publish by emitting WorkspaceWasPublished this is simpler for the projections to handle" which is not true either. The projections handle the events the same. So just for completion: We always have deserialise all the commands that need to be reapplied and splitting them by ids doesnt cost anything. The costly simulation will be carried out for both partial and full publish.

And just for completeness with #5385 will the WorkspaceWasPartially published event actually removed as we only need a WorkspaceWasPublished event at all.

[bwaidelich]

We only said "do a full publish by emitting WorkspaceWasPublished this is simpler for the projections to handle"

I don't know who said that, but it wasn't me. My point was: If is much easier to reason about the events if we can eliminate the last case of:

• all changes were published
• some changes were published
• all changes were published but through a different button

with #5385 will the WorkspaceWasPartially published event actually removed

That would be the great ofc, but I really don't understand how #5385 is supposed to work. And IMO the difference between a partial and a full publish has a different meaning and should not be distinguished just by some metadata