A simple yet efficient library enabling aggregation queries over large sets of things
A free
None
HolyMolap shall be usable in production for very large instances. Such workloads may be loaded into NUMA hardware.
JDK14 enables G1 to be NUMA-aware: https://openjdk.org/jeps/345
All keys are Strings. Values may be anything (e.g. String, int, Number, double[], List, ...). All public data-structures may be bigger than Integer.MAX_VALUE, but always smaller than Long.MAX_VALUE. Implementations would generally be partitionned. All data-structures are append-only. Removal can be done through masking (i.e. queries can be executed over a chain, where elements may be removals, hiding some templates from the rest of the chain). We prefer consistency over performance: while we expect generally good performances, we prefer lower performances but easy monitoring of the result progress, support for cancellation, etc.