Inconsistent EO and EO3 Product Definition File Format handling

Question

Is there a common mapping from the EO to the EO3 data formats for product defintion yaml files. Are the following logically the same as they both cause differing problems during ODC data-in process.

metadata_type: eo
metadata:
    platform:
        code: LANDSAT8
    instrument:
        name: OLI
    product_type: level2
    format:
        name: GeoTIFF

metadata_type: eo3
metadata:
product:
    name: landsat8
properties:
    eo:platform: LANDSAT8
    eo:instrument: OLI
    odc:product_family: level2
    odc:file_format: GeoTIFF

Firstly, if we use the EO format then the ODC Web-UI loads with 'Platform', 'Instrument' & 'Product Type' columns correctly populated, whereas the EO3 approach leaves them empty. This causes other problems later on also when we use the eodatasets3 DatasetAssembler package to generate the yamls for the data sets as these are in EO3 format.

So the second approach of starting out with the EO3 approach on product definition leaves the Web-UI missing data but also causes another issue after we generate our dataset yaml and try and run the 'datacube dataset add' as follows:

datacube-dataset ERROR Dataset metadata did not match product signature.
Dataset definition:
 {
    "product": {
        "name": "landsat8"
    },
    "properties": {
        "datetime": "2020-05-20 11:22:00Z",
        "odc:file_format": "GeoTIFF",
        "odc:processing_datetime": "2021-06-01 14:14:32.626582Z",
        "odc:product_family": "landsat8"
    }
}

Product signature:
 {
    "product": {
        "name": "landsat8"
    },
    "properties": {
        "eo:platform": "LANDSAT8",
        "eo:instrument": "OLI",
        "odc:file_format": "GeoTIFF",
        "odc:product_family": "level2"
    }
}

Which suggests the create yaml process from the eodatasets3 DatasetAssembler package isn't reading the product definition properly to add it into the dataset yaml file. We can get around this by editing the dataset yaml file to make the Metadata match but I reckon we have our steps wrong, although we can code around these seeming mismatches....

Answer 1

When it comes to Datacube UI I think you'll have to stick with EO metadata types.

That application is not particularly robust to different types and it is fairly hard to use when you do have your metadata sorted well.

Personally, I don't recommend the application. I do recommend Datacube OWS, Datacube Explorer and using Datacube in Jupyter notebooks.