Polygon tool library

[mpaget]

Polygon tool library

• Checks and balances
• Accepted input/output types
• Simple polygons
• Geopolygons
• Statistic types
  • median cover
  • benchmark analyses. See Figs 7-10 (Ground cover milestone 3 report, 2022-12-21, p.12-14) - Not covered by the project (future work)
    • Request a code/tech review when appropriate
  • NRM/LGA reports
• Donuts

[mpaget]

datacube-wps

• input_data()
  1. dc "bag" for polygon > find_datasets
  2. box (from bag) for output resolution > xarray
  3. get mask rasterio.features.geometry_mask()
     • all_touched = False
     • invert = True
  4. apply to xarray (where)
  5. return xarray

rasterio.features.geometry_mask()

• [geom.to_crs(geobox.crs)], # what if geom is a list already?
• geometries = iterable geoms
• out_shape = target nd.array shape
• all_touched = False (centre within geom), True (any part of pixel in geom)
• invert = False (mask True is outside geo), True (mask True is inside geom) # check apply mask

To do

• Check geom input
  • get "features" from request
  • multipolygon "feature" not supported
  • use CRS in request or geom, else default is WGS84
  • create a datacube.utils.geometry.Geometry object
    • convert to 2D (take first 2 values from each coord sequence)
    • convert to shapely geom
    • support multipolygons? Yes
  • product/tool-specific process_data() function handles the geom
• Check mask
  • set all_touched > PolygonDrill class variable
    • default in class constructor PolygonDrill().mask_all_touched = True|**False**
    • set in product/tool-specific PolygonDrill subclass if writable, else add an @x.setter or make it a param
  • invert and apply settings
    • True (mask True is inside geom): rasterio.features.geometry_mask()
    • xarray.where (mask True is preserve. Use nodata or NaN for False): PolygonDrill().input_data()

Out of scope for now (to be revisited if/when needed)

• rasterize performance
  • will check valid geom > return value/exception?
  • The shapes iterator will be materialized to a Python list and another C copy of that list will be made
  • For maximum speed, ensure that GDAL_CACHEMAX is larger than the size of out or out_shape.
• other rasterio functions we could consider