to_xarray download fallback

[guillaumeeb]

Is your feature request related to a problem? Please describe.
Using to_xarray on an example from CNES S3 Datalake seems to trigger concurrent downloading of data:

This is lasting for about a minute, and generates high memory spikes (more than GiB temporarily for an entire product).

Describe the solution you'd like
to_xarray should not load the data. open_rasterio from rioxarray only needs to read the metadata of the rasters to create the xarray structure. Plus, the ipywidget in a Jupyter env seems to indicate some temporary download?

Describe alternatives you've considered
Use directly rioXarray, e.g.

import rioxarray as rxr
import xarray as xr

with env:
    xr_da_b2 = rxr.open_rasterio(product.assets["B02"]["href"]).squeeze() # remove band dimension
    xr_da_b3 = rxr.open_rasterio(product.assets["B03"]["href"]).squeeze()
    xr_da_b4 = rxr.open_rasterio(product.assets["B04"]["href"]).squeeze()
    xr_da_b8 = rxr.open_rasterio(product.assets["B08"]["href"]).squeeze()
    xr_ds = xr.Dataset(data_vars={'blue':xr_da_b2, 'green':xr_da_b3, 'red':xr_da_b4, 'nir': xr_da_b8})
xr_ds

[sbrunato]

Hello @guillaumeeb , there must have be an error during the attempt to remotely read the data which results to its download as fallback. Could you please use setup_logging(3) to increase logging verbosity and report the obtained logs ? Thanks

[guillaumeeb]

Interesting, I wasn't aware of this. May I suggest to add a Warning if the remote read of data does not work?

Here are the logs, which suggest you are right:

2025-10-03 15:20:19,493 eodag-cube.utils.xarray          [DEBUG   ] Could not re-open file: Parameter validation failed:
Invalid bucket name "zip+s3:": Bucket name must match the regex "^[a-zA-Z0-9[.\-_](https://jupyterhub.sis.cnes.fr/user/eynardbg/lab/tree/home/eh/eynardbg/git/pluto-tuto-cnes/datalake/-_)]{1,255}$" or be an ARN matching the regex "^arn:(aws).*:(s3|s3-object-lambda):[a-z\-0-9]*:[0-9]{12}:accesspoint[/:][a-zA-Z0-9\-.]{1,63}$|^arn:(aws).*:s3-outposts:[a-z\-0-9]+:[0-9]{12}:outpost[/:][a-zA-Z0-9\-]{1,63}[/:]accesspoint[/:][a-zA-Z0-9\-]{1,63}$"
2025-10-03 15:20:19,518 eodag-cube.utils.xarray          [DEBUG   ] Cannot open zip+s3://sentinel2-l1c/31/T/CH/2023/07/14/S2A_MSIL1C_20230714T105031_N0509_R051_T31TCH_20230714T143218.zip!S2A_MSIL1C_20230714T105031_N0509_R051_T31TCH_20230714T143218.SAFE/GRANULE/L1C_T31TCH_A042090_20230714T105030/IMG_DATA/T31TCH_20230714T105031_B01.jp2 with ('s3', 's3a') + rasterio: '/vsizip/vsis3/sentinel2-l1c/31/T/CH/2023/07/14/S2A_MSIL1C_20230714T105031_N0509_R051_T31TCH_20230714T143218.zip/S2A_MSIL1C_20230714T105031_N0509_R051_T31TCH_20230714T143218.SAFE/GRANULE/L1C_T31TCH_A042090_20230714T105030/IMG_DATA/T31TCH_20230714T105031_B01.jp2' does not exist in the file system, and is not recognized as a supported dataset name.
2025-10-03 15:20:19,518 eodag-cube.api.product           [DEBUG   ] Cannot open EOProduct(id=S2A_MSIL1C_20230714T105031_N0509_R051_T31TCH_20230714T143218, provider=geodes_s3) B01: None of the engines ['rasterio'] could open the dataset at zip+s3://sentinel2-l1c/31/T/CH/2023/07/14/S2A_MSIL1C_20230714T105031_N0509_R051_T31TCH_20230714T143218.zip!S2A_MSIL1C_20230714T105031_N0509_R051_T31TCH_20230714T143218.SAFE/GRANULE/L1C_T31TCH_A042090_20230714T105030/IMG_DATA/T31TCH_20230714T105031_B01.jp2.

Interestingly, I got this error messages with rasterio without setting the proper variables, outside of eodag-cube.

Indeed, adding the correct variables for CNES S3, this works lazily.

env = product.rio_env()
env.options["GDAL_HTTP_UNSAFESSL"] = 'YES'
env.options["GDAL_DISABLE_READDIR_ON_OPEN"] = 'FALSE'


with env:
    ds = product.assets["B01"].to_xarray()
    ds

So it might just be some environment variable missing for GDAL/Rasterio in CNES S3 Datalake context?