I am interested in adding support for multi-dimensional datasets within Rasterio as per the recent changes within GDAL - https://gdal.org/development/rfc/rfc75_multidimensional_arrays.html
This will involve API changes to Rasterio and is a significant amount of work.
@sgillies is this something that you are interested in reviewing and have any guidance on how you would like this to be done?
I can start by adding the cython code if this of interest.
normanb
All 5 comments
@normanb why not use h5py? What would surfacing the new API in rasterio do for Python programmers?
sgillies
on 22 Aug 2019
@sgillies there are additional n-dimensional data formats in addition to those supported by h5py e.g. Grib, TileDB and Zarr as well as other formats that can be abstracted behind a VRT file (using COGs as an n-dimensional dataset with VRT for example). XArray and Pangeo use rasterio to load geospatial datasets and could benefit from having the n-dimensional API surfaced.
normanb
on 22 Aug 2019
This is something that should be discussed in our dev group, for sure, before we do anything about it here.
sgillies
on 22 Aug 2019
Discussion here: https://rasterio.groups.io/g/main/message/288. Next time, let's be sure to post to
[email protected] instead.
sgillies
on 23 Aug 2019
The project is not going to add a new multidimensional API.
sgillies
on 8 Jan 2020
Related issues
sgillies
路
4Comments
atorch
路
3Comments
astrojuanlu
路
4Comments
sgillies
路
3Comments
ungarj
路
4Comments
Most helpful comment
@sgillies there are additional n-dimensional data formats in addition to those supported by h5py e.g. Grib, TileDB and Zarr as well as other formats that can be abstracted behind a VRT file (using COGs as an n-dimensional dataset with VRT for example). XArray and Pangeo use rasterio to load geospatial datasets and could benefit from having the n-dimensional API surfaced.