Which modes to support in bulk API store?

[enikao]

In bulk API store command, we consider different modes.
Their names are inspired by SQL, and might change.

We consider a node to be updated if a node with the same id already exists in the repository. Otherwise, we consider that node to be new.

replace

Accept both new and updated nodes.

Pro:

• Main idea of bulk API: "this is how the world should be"

create

Only accept new nodes. The call fails if any node id from nodes already exists in the repository. The call accepts already existing node ids only referenced as child, reference, or annotation.

Pro:

• Easy to check for the repository
• Prevents against accidental changes to nodes with ids we assumed to be free, but aren't

Con:

• Effectively never useful: According to When moving a node, do we need to mention both source and target parent? #227 and In bulk API store(), a node's parent field may only be null for root nodes #228, we must change the parent when adding a new child. We cannot change an existing node with this mode. According to How to create and delete partitions #216, we create new partitions explicitly. So this mode can never be used.

update

Only accept existing nodes. The call fails if any node id does not exist in the repository.

Pro:

• Easy to check for the repository
• Prevents from accidentally creating new nodes

Con:

• Rare use case

Decision: Don't support modes in bulk API store