Algorithmically feasible: blocknotify.py checks the notified block's prevhash against the hash of the last-recorded block in the graph. If they differ, it implies that a re-org has occurred and blocknotify flips into “re-org” mode, iterating down from currentheight, checking the recorded nexthash value against the node's nexthash value and, until they match, removing the orphaned blocks from the graph and then finally, when the nexthashes match, catching up from there to current.

I'm mapping the Gapcoin blockchain atm, I changed the catchup procedure so that it cycles in 100K block steps, it's a tad kinder, and seeing as the initial mapping is usually a one-off task, I left the output files of ntriples in place, ready for concatenation and feeding through to Jena's TDB cli.

Point taken about filenames scripts and roles, needs to be clearer. The unit test inclusion is just my perverted interpretation of TDD, by binding the ACME virtualenv python to a SublimeText Build, I can develop the code using a restricted subset of the data and just hitting ^B runs the test, very direct prototyping. When I've got to a suitable stopping point, I release the restrictions and can just run it as a specific test. Just a development foible of mine.

I've also adjusted Fuseki on the server to increase the max JVM heap space to 8Gb, responses seem faster but that may be fond imagining on my part.

Cheers

Graham