As part of a comparison of Pandas 1.5.1 and 1.5.2 output for the ETL, it appears that many tables in the data/grapher/un_sdg/2022-07-07/un_sdg dataset get minor renames like below.
My interpretation is that the table names are being generated by exploding dimensions, but the explosion is not done in a deterministic order.
Impact
It's not a big deal for our usage in Grapher, but it makes QA on them more difficult, and it means the short names aren't stable handles for these variables at the moment.
Problem
As part of a comparison of Pandas 1.5.1 and 1.5.2 output for the ETL, it appears that many tables in the
data/grapher/un_sdg/2022-07-07/un_sdg
dataset get minor renames like below.Before:
data/grapher/un_sdg/2022-07-07/un_sdg/_10_3_1__vc_vov_gdsd__no_breakdown_by_disability__age__all_areas__both_sexes.feather
After:
data/grapher/un_sdg/2022-07-07/un_sdg/_10_3_1__vc_vov_gdsd__all_areas__no_breakdown_by_disability__age__both_sexes.feather
My interpretation is that the table names are being generated by exploding dimensions, but the explosion is not done in a deterministic order.
Impact
It's not a big deal for our usage in Grapher, but it makes QA on them more difficult, and it means the short names aren't stable handles for these variables at the moment.