Baseline compatability with STAC?

[robmarkcole]

I haven't spent much with with terracotta yet but its great that it is agnositc about the file structure of imagery. However you might also want to have a baseline for documenting etc, so perhaps could consider adhering to the STAC spec? https://github.com/radiantearth/stac-spec

[dionhaefner]

Very interesting.

Part of the point of Terracotta is that we do not want to include all metadata that you could conceivably use to describe a raster file to build a catalogue, but instead hand-pick only the relevant keys that make sense for your given deployment. We made this choice because of two main reasons:

1. To reduce cognitive load on users and frontend devs (who might not be familiar with the intricacies of geodata).
2. Because I personally believe that every attempt of a spec is doomed, because the geodata universe is so rich and heterogeneous.

I didn't know about the STAC spec, and I do think it looks reasonably small and general to be useful. However, I don't see how it could fit with the current Terracotta data model.

That being said, in Terracotta you can always roll your own conventions:

import terracotta as tc

driver = tc.get_driver('mydb.sqlite')

raster_file = 'myraster.tif'
raster_keys = ['S2', '20180101', 'B04']
stac_metadata = {
    'asset': ...,
    'geometry': ...,
    'link': ...,
    # other STAC data
}

tc_metadata = driver.compute_metadata(raster_file, extra_metadata=stac_metadata)
driver.insert(keys, raster_file, metadata=tc_metadata)

Then, you can retrieve the extra metadata verbatim from the frontend:

$ curl localhost:5000/metadata/S2/20180101/B04
{
  ...
  metadata: {
    asset: ...,
    geometry: ...,
    link: ...
  }
}

Hope that clears it up a bit.