Include collections in the geoparquet metadata

[gadomski]

https://github.com/stac-utils/stac-geoparquet/blob/main/spec/stac-geoparquet-spec.md#including-a-stac-collection-json-in-a-stac-geoparquet-collection

I think we'd support this on the way in either by:

• A builder, or
• Expecting it on the input item collection (cleaner, I think)

Way out would be similar-ish, where if we didn't do it on an item collection we'd just hit the parquet file twice, which seems bad.

Probably just add a collection field to an ItemCollection, eh?

[bitner]

It should probably be a collections array field rather than singular to accommodate geoparquet that may have Items from multiple collections.

[gadomski]

@bitner should we update the spec to reflect this? It seems to suggest that only a single collection should be included.

[bitner]

I see two use cases here --

1. Geoparquet as an archive format that is generally going to be partitioned by Collection in which case there would be a single Collection per parquet file
2. Geoparquet as an output format for an API search in which case there would be potentially multiple Collections.

It would seem to me that we would want to update the spec to accommodate both.

[gadomski]

🤔 I think my instinct is to make collections a map in the metadata so you can always do a quick key lookup (in psuedo-json):

{
  "collections": {
    "collection-a": { ... },
    "collection-b": { ... }
  },
  "items": [
    { "id": "item-a", "collection": "collection-a", ... },
    { "id": "item-b", "collection": "collection-b", ... }
  ]
}

[gadomski]

RFC for multiple collections: stac-utils/stac-geoparquet#88