Create raster tables with geography and metadata

[jgoizueta]

Issue

sc-287177

Raster tables shouldn't require any post-processing to build block GEOGRAPHIEs, and metadata should be available as efficiently as possible.

Proposed Changes

• Generate geog GEOGRAPHY column with block boundaries for generic rasters
• Add a metadata row (with NULL 'geog')
• Change attrs column type to JSON (not supported by the Python client at the moment)

Pull Request Checklist

• I have tested the changes locally
• I have added tests to cover my changes (if applicable)
• I have updated the documentation (if applicable)

Additional Information

[Anything else you'd like to include.]

[shortcut-integration]

This pull request has been linked to Shortcut Story #287177: Do not update the user's input table.

[Jesus89]

Good job!

I have found a difference in the two attributes of the quadbin raster. I'll paste the example after running RASTER_GETMETADATA:

Production

  "raster_boundary": "POLYGON((-4.21875 50.7364551370107, -4.21875 49.8379824530848, -4.21875 48.9224992637582, -2.8125 48.9224992637582, -1.40625 48.9224992637582, -1.40625 49.8379824530848, -1.40625 50.7364551370107, -2.8125 50.7364551370107, -4.21875 50.7364551370107))",
  "avg_pixel_area": "155209.87843240262",

New changes

  "raster_boundary": "POLYGON((-4.21874999999998 48.9224992637582, -4.04296874999999 48.9224992637582, -3.86718749999999 48.9224992637582, -3.69140624999999 48.9224992637582, -3.51562499999999 48.9224992637582, -3.33984374999999 48.9224992637582, -3.16406249999999 48.9224992637582, -2.98828124999999 48.9224992637582, -2.81249999999999 48.9224992637582, -2.63671874999999 48.9224992637582, -2.46093749999999 48.9224992637582, -2.28515624999999 48.9224992637582, -2.10937499999999 48.9224992637582, -1.93359374999999 48.9224992637582, -1.75781249999999 48.9224992637582, -1.58203124999999 48.9224992637582, -1.40625 48.9224992637582, -1.40625 49.1492437479148, -1.40625 49.3759882320713, -1.40625 49.6027327162279, -1.40625 49.8294772003844, -1.40625 50.056221684541, -1.40625 50.2829661686975, -1.40625 50.5097106528541, -1.40625 50.7364551370106, -1.58203124999999 50.7364551370106, -1.75781249999999 50.7364551370106, -1.93359374999999 50.7364551370106, -2.10937499999999 50.7364551370106, -2.28515624999999 50.7364551370106, -2.46093749999999 50.7364551370106, -2.63671874999999 50.7364551370106, -2.81249999999999 50.7364551370106, -2.98828124999999 50.7364551370106, -3.16406249999999 50.7364551370106, -3.33984374999999 50.7364551370106, -3.51562499999999 50.7364551370106, -3.69140624999999 50.7364551370106, -3.86718749999999 50.7364551370106, -4.04296874999999 50.7364551370106, -4.21874999999999 50.7364551370106, -4.21874999999998 50.5097106528541, -4.21874999999998 50.2829661686975, -4.21874999999999 50.056221684541, -4.21874999999999 49.8294772003844, -4.21874999999998 49.6027327162279, -4.21874999999998 49.3759882320713, -4.21874999999999 49.1492437479148, -4.21874999999998 48.9224992637582))",
  "avg_pixel_area": "31572.769599917083",

So I think there is a bug in the pixel_area formula, and something more with the boundary (seems tesselated).

[jgoizueta]

Production

  "raster_boundary": "POLYGON((-4.21875 50.7364551370107, -4.21875 49.8379824530848, -4.21875 48.9224992637582, -2.8125 48.9224992637582, -1.40625 48.9224992637582, -1.40625 49.8379824530848, -1.40625 50.7364551370107, -2.8125 50.7364551370107, -4.21875 50.7364551370107))",
  "avg_pixel_area": "155209.87843240262",

New changes

  "raster_boundary": "POLYGON((-4.21874999999998 48.9224992637582, -4.04296874999999 48.9224992637582, -3.86718749999999 48.9224992637582, -3.69140624999999 48.9224992637582, -3.51562499999999 48.9224992637582, -3.33984374999999 48.9224992637582, -3.16406249999999 48.9224992637582, -2.98828124999999 48.9224992637582, -2.81249999999999 48.9224992637582, -2.63671874999999 48.9224992637582, -2.46093749999999 48.9224992637582, -2.28515624999999 48.9224992637582, -2.10937499999999 48.9224992637582, -1.93359374999999 48.9224992637582, -1.75781249999999 48.9224992637582, -1.58203124999999 48.9224992637582, -1.40625 48.9224992637582, -1.40625 49.1492437479148, -1.40625 49.3759882320713, -1.40625 49.6027327162279, -1.40625 49.8294772003844, -1.40625 50.056221684541, -1.40625 50.2829661686975, -1.40625 50.5097106528541, -1.40625 50.7364551370106, -1.58203124999999 50.7364551370106, -1.75781249999999 50.7364551370106, -1.93359374999999 50.7364551370106, -2.10937499999999 50.7364551370106, -2.28515624999999 50.7364551370106, -2.46093749999999 50.7364551370106, -2.63671874999999 50.7364551370106, -2.81249999999999 50.7364551370106, -2.98828124999999 50.7364551370106, -3.16406249999999 50.7364551370106, -3.33984374999999 50.7364551370106, -3.51562499999999 50.7364551370106, -3.69140624999999 50.7364551370106, -3.86718749999999 50.7364551370106, -4.04296874999999 50.7364551370106, -4.21874999999999 50.7364551370106, -4.21874999999998 50.5097106528541, -4.21874999999998 50.2829661686975, -4.21874999999999 50.056221684541, -4.21874999999999 49.8294772003844, -4.21874999999998 49.6027327162279, -4.21874999999998 49.3759882320713, -4.21874999999999 49.1492437479148, -4.21874999999998 48.9224992637582))",
  "avg_pixel_area": "31572.769599917083",

After the fix in 7bff268

  "avg_pixel_area": "155209.87729301196",

[Jesus89]

LGTM!

I have tested all the raster files and found no more regression after some iterations.

Also, the upload time is similar to v0.2 including all the pre-computation of the metadata.