How-to

Run schema-map on each supplied level below admin0:

topo-tools schema-map admin2.parquet --map-only

This writes admin2_crosswalk.csv without changing the input. schema-map matches columns by their values, not their names, so check every row:

source_column, target_column, unique_count, note
NAME_2,        adm2_name,     8,
PCODE_2,       adm2_code,     8,
NAME_1,        adm1_name,     2,
PCODE_1,       adm1_code,     2,

schema-map maps every column whose values fit a level, so check each numbered sibling (adm2_code1, adm2_name2). It can be a translation or alternate spelling, a reference number, or a code or name from another source or an older version, and a code column can get a name target. Decide whether the release keeps each one: blank out its target_column to drop it, or correct its target to the next free adm2_code1/adm2_name1. A per-level extra such as a unit type takes the level's own prefix (adm2_type). Output columns follow the crosswalk's row order, so move rows to rearrange them. Then apply it:

topo-tools schema-map admin2.parquet admin2_mapped.parquet --csv admin2_crosswalk.csv

The output keeps only the columns with a target_column, plus geometry. A column with one value in every row, such as the country's name and code, stays blank and is dropped.

With more than one level below admin0, copy each parent's codes and names onto its children, coarsest first, overwriting each child in place:

topo-tools schema-join admin2_mapped.parquet admin1_mapped.parquet admin2_mapped.parquet

If any child's value differs from its parent's, schema-join keeps the child's column and adds the parent's as the next free sibling (adm2_name1), filled on every row. Its issues file lists:

• value-mismatch: a differing name or code, settled in review names;
• no-overlap, low-overlap: a unit outside or mostly outside its parent, settled with the data provider before coding.