Code review comment for lp:~zorba-coders/zorba/xml-in-json-indices

Hi Matthias,

Actually, absolutely all items in a tree that is in a collection share a unique CollectionTreeInfo instance, so that the TreeId, position, etc, exist only once physically. JSON items point to it (as before), and node items have their XML tree point to it.

JSON items only have a non-NULL pointer to CollectionTreeInfo if they are in a collection. The CollectionTreeInfo instance is created upon an attachToCollection call and delete upon detachFromCollection.

XML nodes also have a pointer if they are not in a collection, because the tree ID (among others) is also needed at some places (like structural URIs) even if they are not in a collection. In such a case, i.e., the XML tree is not in a collection, the XML tree needs to instantiate its own CollectionTreeInfo (with a NULL collection pointer). It will be deleted upon insertion in a collection and replaced with the collection-tree-wide instance. It will be recreated upon quitting the collection.

Finally, note that the root of an XML tree (mostly used for garbage collection if I am correct) might be different from the overall root of the tree in the collection (which might be a JSON item). An XML tree can "become" in a collection either directly (through attachToCollection), in which case these roots are the same, or indirectly (through setCollectionTreeInfo) if it was inserted in a JSON item that is in a collection.

I hope this helps. Does it make sense?