Where the initial releases of glutil only supported tables with a [year, month, day, hour]
schema, now arbitrary partition schemas are supported. So long as they conform to either:
-
Partition keys are path segments
partition keys: [year, month, day] s3://bucket/table/prefix/2019/08/12/ => [2019, 08, 12]
-
Partition keys are path segments, with hive-format paths
partition keys: [dt] s3://bucket/table/prefix/dt=2019-08-12/ => [2019-08-12]
-
Single-key partitions, where the partition value is the path with slashes replaced with hyphens
partition keys: [dt] s3://bucket/table/prefix/2019/08/12/ => [2019-08-12]
Everything appears to be working as expected, the only change is that Partitioner.partitions_on_disk
only accepts a limit_days
key if the table's first three partition keys are [year, month, day]
.