Packaging

Check the name-cleaned output of step 5 with validate first. It runs schema-detect, topo-detect, code-detect and name-detect on the layer and writes each report plus a summary into checks/, exiting with status 1 if any report has an error.

topo-tools validate admin3_names_edited.parquet --output-dir checks \
  --name-field "adm{n}_name" --code-field "adm{n}_code"

Open checks/admin3_names_edited_validate_summary.csv for one row per stage and kind found; each row names the report with the details.

Then run package on the same layer.

topo-tools package admin3_names_edited.parquet --output "release/{x}.parquet" \
  --name-field "adm{n}_name" --code-field "adm{n}_code" \
  --output-code-field "adm{n}_pcode"

Earlier steps work with adm{n}_code; --output-code-field writes the release’s code columns as adm{n}_pcode.

Writes one release/admin{n}.parquet per detected level, release/points.parquet (one label point per admin unit), and release/lines.parquet (the deduplicated boundary network).

topo-clean, edge-match, and code-create each write an issues file only when they find something to report; load one as a map layer, not just a table, to see exactly which polygons it touched.