Hi, I was playing around with rf_local_add and came across the following results. I expected that with local tile arithmatic, no_data + data = no_data.
This code snippet creates a tile with 0's and 1's, and a tile with all 0's. Then I set 1 as the no data value. This makes the tile with 0's and 1's to now have 0's and "no datas" . Finally, I add the tiles together. The tile with with 0's and 1's does have no data as expected, but the sum does not have any nodata.
tile_size = 100
x = Tile(np.eye(tile_size, dtype='int16'))
y = Tile(np.zeros((tile_size, tile_size), dtype='int16'))
rf = spark.createDataFrame(pd.DataFrame(data =
{'eye_tile': [x],
'zero_tile': [y]}))
mask_ct = CellType('uint16').with_no_data_value(1)
masked_rf = rf.withColumn('eye_tile_nd',
rf_convert_cell_type('eye_tile', mask_ct.cell_type_name)) \
.withColumn('zero_tile_nd',
rf_convert_cell_type('zero_tile', mask_ct.cell_type_name)) \
.withColumn('masked', rf_local_add('eye_tile_nd', 'zero_tile_nd'))
However, when doing the sum directly on the tiles, I do get a result with nodatas.
r = masked_rf.collect()
r[0].eye_tile_nd + r[0].zero_tile_nd
The second behavior is what I expected, but regardless I think the behavior should be consistent.
Hi, I was playing around with
rf_local_addand came across the following results. I expected that with local tile arithmatic, no_data + data = no_data.This code snippet creates a tile with 0's and 1's, and a tile with all 0's. Then I set 1 as the no data value. This makes the tile with 0's and 1's to now have 0's and "no datas" . Finally, I add the tiles together. The tile with with 0's and 1's does have no data as expected, but the sum does not have any nodata.
However, when doing the sum directly on the tiles, I do get a result with nodatas.
The second behavior is what I expected, but regardless I think the behavior should be consistent.