Closed jdries closed 1 year ago
Sentinel 2 cube logging:
Created cube for urn:eop:VITO:TERRASCOPE_S2_TOC_V2 with metadata TileLayerMetadata(int16ud32767,LayoutDefinition(Extent(652910.0, 5679370.0, 655470.0, 5681930.0),CellSize(10.0,10.0),2x2 tiles,256x256 pixels),Extent(652915.9230789405, 5680678.468706609, 654568.4352774331, 5681921.294014332),EPSG:32631,KeyBounds(SpaceTimeKey(0,0,1614556800000),SpaceTimeKey(1,0,1635465600000))) and partitioner Some(SpacePartitioner(KeyBounds(SpaceTimeKey(0,0,1614556800000),SpaceTimeKey(1,0,1635465600000))))
Cube partitioner index: SparseSpaceTimePartitioner 54 false
aggregate_temporal results in 48 keys, using partitioner index: org.openeo.geotrelliscommon.package$SpaceTimeByMonthPartitioner$@79c654b8
Created cube for urn:eop:VITO:CGS_S1_GRD_SIGMA0_L1 with metadata TileLayerMetadata(float32,LayoutDefinition(Extent(652910.0, 5679370.0, 655470.0, 5681930.0),CellSize(10.0,10.0),2x2 tiles,256x256 pixels),Extent(652915.9230789405, 5680678.468706609, 654568.4352774331, 5681921.294014332),EPSG:32631,KeyBounds(SpaceTimeKey(0,0,1614816000000),SpaceTimeKey(1,0,1635638400000))) and partitioner Some(SpacePartitioner(KeyBounds(SpaceTimeKey(0,0,1614816000000),SpaceTimeKey(1,0,1635638400000))))
Cube partitioner index: SparseSpaceTimePartitioner 41 false
aggregate_temporal results in 48 keys, using partitioner index: org.openeo.geotrelliscommon.package$SpaceTimeByMonthPartitioner$@79c654b8
Merging cubes with spatial indices: org.openeo.geotrelliscommon.package$SpaceTimeByMonthPartitioner$@79c654b8 - org.openeo.geotrelliscommon.package$SpaceTimeByMonthPartitioner$@79c654b8
Fixes committed and reproduced in unit test. The partitioner keybounds was set incorrectly by aggregate_temporal in the particular case where the target label bounds exceeded the bounds of the original cube. My hypothesis is that the incorrect bounds then caused a mismatch in the partitions themselves, and the join based on SpacePartitioner to be incorrect.
A complex combination leads to a datacube where the first S2 image in the timeseries is nodata: