OAK-9757 : increased node name limit for mongo 4.2 version

[rishabhdaim]

No description provided.

[reschke]

Wow. I would say that if we need to modify 109 files, there's something very wrong with this change.

[reschke]

Idea: Move (or copy) the util functions to the DocumentStore API, where they have access to metadata.

[rishabhdaim]

Thanks @reschke, that's a good idea to move util function to DocumentStore and that would help us avoiding in calling metadata multiple times.
Let me make changes and raise the PR again, but I doubt that would restrict the code changes to less file since most of the changes are in tests cases and the changes are to pass the parameters to utils class.

[mreutegg]

I suggest we take a slightly different approach. To me it feels wrong that Utils.isLongPath() is currently responsible for detecting when a node name is too long. I think this is the reason why so many other code changes are necessary. E.g. the method is then used in Utils.getIdFromPath(), which in turn is used in many places. But in many cases the check for the node name length is not actually needed. I my view it would be better if the node name length check is moved out of Utils.isLongPath() to a place were it is really needed. That is, when a new node is added through the DocumentNodeStore or created with an UpdateOp on the DocumentStore.

[mreutegg]

Also note that checks fails because two new files for classes MongoVersion and MongoVersionTest do not have a license header.

[rishabhdaim]

@mreutegg

To me it feels wrong that Utils.isLongPath() is currently responsible for detecting when a node name is too long

Yes, I think it should be other way around i.e. node name too long should check whether path is long or not.
I think, we should move out the logic to check node name too long to a separate method and call Utils.isLongPath() from it since node name too long make sense only for long paths.
/*** The maximum size a node name, in bytes. This is only a problem for long path.*/ int NODE_NAME_LIMIT = Integer.getInteger("oak.nodeNameLimit", 150);
As mentioned by you, we should move whole logic for node name too long here https://github.com/apache/jackrabbit-oak/blob/trunk/oak-store-document/src/main/java/org/apache/jackrabbit/oak/plugins/document/DocumentNodeState.java#L417 and then we should be able to reduce the number of changed files significantly.

[mreutegg]

The changes look pretty good now but then I noticed a problem.

The new default method on the DocumentStore interfaces is not overridden in wrapper classes like LeaseCheckDocumentStoreWrapper. This means the increased node name capability of a MongoDocumentStore may be hidden behind one of those wrapper classes.

[stefan-egli]

+1, lgtm now.
(I agree, the other DocumentStore classes in /src/test do not necessarily have to implement getNodeNameLimit - so that's fine to leave those out)