build collecton/item metadata system

[hrodmn]

Objective: Build a stactools package for the GLAD Land Cover Land Use 2020 dataset.

Notes:

• I split the data into two collections because the annual and change items would overlap in time and have different asset properties:
  • annual data products: glad-glclu2020
  • 2000-2020 change product: glad-glclu2020-change
• We could do it in a single collection but users would need to apply a specific filter to avoid retrieving two items with every query!
• The default setting is to generate metadata for the original assets but those are not COGs so be careful out there, folks.
• Users can provide their own href_format to generate metadata for new copies of the assets (e.g. COGs in S3).

Checklist

• Tests pass (run scripts/test)
• Documentation has been updated to reflect changes, if applicable
• Examples have been updated to reflect changes, if applicable
• Changes are added to the CHANGELOG, if applicable

[wildintellect]

Missing:

• providers, that MAAP is the host (Processor/Host), also our copy is going to be COG (this is our processing), and that the original data is from the cited source (Producer/Licensor)
• projection extension is in the items but not declared in the collection
• Assets at the collection level don't have the data object specified only thumbnail.
• Should we use the raster extension? not sure where else resolution is specified

[hrodmn]

Thanks for the quick feedback @wildintellect.

• providers, that MAAP is the host (Processor/Host), also our copy is going to be COG (this is our processing), and that the original data is from the cited source (Producer/Licensor)

I will add a provider entry for the Producer/Licensor. I wrote the package to build metadata for the original data which are not COGs and imagined that when we build metadata for the MAAP-hosted data we can add the Processor/Host entry for MAAP.

• Assets at the collection level don't have the data object specified only thumbnail.

I filled out the item_assets section with the actual data asset information. I think the thumbnail is the only proper collection-level asset?

• Should we use the raster extension? not sure where else resolution is specified

Yes, I'll add this! I didn't try it yet but maybe this is a chance to try STAC v1.1.0? pystac doesn't support 1.1 yet so I'll hold off stac-utils/pystac#1375

[wildintellect]

@hrodmn

Thanks for the quick feedback @wildintellect.

• providers, that MAAP is the host (Processor/Host), also our copy is going to be COG (this is our processing), and that the original data is from the cited source (Producer/Licensor)

I will add a provider entry for the Producer/Licensor. I wrote the package to build metadata for the original data which are not COGs and imagined that when we build metadata for the MAAP-hosted data we can add the Processor/Host entry for MAAP.

There should be an easy flag, arg to pass this info into the generation. I've done this before, let me try to recall the example https://github.com/stactools-packages/cop-dem/blob/852ed13abc95513e4113369004cfb6110bde0081/src/stactools/cop_dem/constants.py#L46-L57

• Assets at the collection level don't have the data object specified only thumbnail.

I filled out the item_assets section with the actual data asset information. I think the thumbnail is the only proper collection-level asset?

But the collection is also supposed to show what assets are available on items (this might be the item-asset extension)

• Should we use the raster extension? not sure where else resolution is specified

Yes, I'll add this! I didn't try it yet but maybe this is a chance to try STAC v1.1.0? pystac doesn't support 1.1 yet so I'll hold off stac-utils/pystac#1375

I also wasn't sure if pgstac supports 1.1 yet either...

[hrodmn]

@wildintellect I added the providers and the extension projection and raster extension entries to the collection. The item-assets extension is handling the "common assets" description that you mentioned.