Hello Oliver,
The cost of "variable P" heavily depends on the sample query.
1. In good case the overhead is zero. It's possible when graphs (G) are
known or almost all data of the dataset are in very numerous small
graphs, so it's cheaper to read whole graphs than to try to use
additional indices
Hi Hugh.
I tried once more with the bulk loading procedure and everything worked
smoothly, or at least no error occurred. Therefore, I consider the issue
solved. Perhaps after restarting the virtual machine, the transaction
rolled back and the problem with the indexes has been resolved.
Thank you