nguyenm100

@kenwenzel not sure query planning requires up to the second stats to be useful. when I was playing with halyard, even a general ratio was useful (eg generally, the cardinality...

another argument for generated statistics is that they could be run to fix the query planner whereas a run-time heuristics may require a code change to fix.

I can try locally, where is the E() and hashlength check?

Just to be absolutely sure, we're removing the E(..) wrapping call and leaving the mdb _get() and its return conditional check?

I pulled down tag 5.0.0 and recompiled locally and deployed just to baseline my setup. Seeing smth different now while trying to load 10m with same config. No MDB error...

i will try 100m next to see if i can get the mdb errors with my locally built wars. probably still makes sense to remote the E() check but odd...

deleted and recreated the repo. ran through 100m with no errors, but once load was complete, repo size is 0 (verified by file sizes in the repo). this happened the...

I'm not specifying in the curl call (above) so whatever is the default for rdf4j 5.0?

> It seems that it writes the data two times as it resizes the map also a second time. Maybe this is due to the memory overflow. Which isolation level...

We're in an enterprise env so our standard vms are 32g and goes up from there. Would be interested in understanding optimized configs for lmdb if you have them. 1b...