Names

Continues from the coded output of step 4. Fix the safe name problems and list the rest:

topo-tools name-clean admin3_coded.parquet admin3_names.parquet \
  admin3_name_fixes.parquet --name-field adm{n}_name --code-field adm{n}_code

This fixes spacing, invisible characters, accents stored as separate characters, and text read with the wrong encoding when the repair is certain, marking those rows fixed. Keep this file: it records what name-clean changed. Open the report in QGIS or DuckDB and filter out the fixed rows. Each row names the unit by its code (code_a, and code_b for a pair). Fix every error row and review each warn row. A row with no code covers a whole column.

Also check, against the reference release, each unit whose code already existed in {ref_version}: a real administrative rename is expected, a diverging spelling of the same unit isn’t.

For each value-mismatch row in stage 1’s issues file, a child’s own adm{N}_name differs from its join layer’s, which schema-join added as adm{N}_name{k}. Pick one spelling per parent unit and write it to adm{N}_name on every child under it. Drop that schema-join sibling only once no row still differs from adm{N}_name. Keep any adm{N}_name{k} mapped from the source in stage 1, since it’s an alternate name, not a conflict.

Write the fixed names back to the cleaned file, for example in DuckDB (one WHEN per unit):

COPY (
  SELECT * REPLACE (
    CASE adm3_code WHEN 'XY020301' THEN 'Example Name' ELSE adm3_name END
    AS adm3_name)
  FROM 'admin3_names.parquet'
) TO 'admin3_names_edited.parquet' (FORMAT PARQUET, GEOPARQUET_VERSION 'V2');

Then check the edited file with name-detect, which runs the same checks without changing anything:

topo-tools name-detect admin3_names_edited.parquet admin3_name_issues.parquet \
  --name-field adm{n}_name --code-field adm{n}_code

Repeat until no error row is left. Keep the last issues file: it’s this step’s record, written even when empty.