Hi Martin, Thanks for proposing this feature. Will the writer generate cloud-optimized GeoTiFF. I am asking because Cogs are becoming industry standards for storing raster in object stores.
-Mo On Thu, Mar 23, 2023 at 6:53 AM Martin Andersson (Jira) <j...@apache.org> wrote: > Martin Andersson created SEDONA-268: > --------------------------------------- > > Summary: Add support for raster types in Geotiff writer > Key: SEDONA-268 > URL: https://issues.apache.org/jira/browse/SEDONA-268 > Project: Apache Sedona > Issue Type: Improvement > Reporter: Martin Andersson > > > As discussed in: [ > https://lists.apache.org/thread/kbwqnj7kn9omtpsoyzbn0zsslvd8cm5t] > > With the introduction of raster types in SEDONA-251, it is now possible to > enhance the existing Geotiff writer to directly support writing Geotiffs > from rasters. > > Currently, the Geotiff writer requires six columns to create Geotiff > files: origin, geometry, width, height, nBands, and data. However, for > rasters, only two columns would be necessary: origin and raster. > > To achieve this, we propose modifying the Geotiff writer to first attempt > to use the existing six-column format. If the input DataFrame does not meet > the necessary criteria, the writer should instead look for the origin > column and a column with the raster UDT. > > For the origin column we could use the existing configuration parameter > (fieldOrigin). The raster column could be detected by type. If there are > several columns of type raster, the writer would throw an exception. For > DataFrames with several raster columns users would need to select one at a > time. Example: > {code:java} > df_many_rasters = ... > df_many_rasters.select("origin","raster1").write... > df_many_rasters.select("origin","raster2").write... > {code} > > > > > > > > > > > -- > This message was sent by Atlassian Jira > (v8.20.10#820010) >