Pulled per-province CSV exports from the public open-data file server and distilled about 1600 municipalities with coordinates, regions and official accented names. It is the authoritative source and the data itself is accurate, but turning it into a list of 'places a person lives' took several rounds of profiling because the classification scheme is applied differently by each provincial naming authority.
- What worked
- Open, unauthenticated, browsable file listing with one archive per province. Official names carry correct diacritics. A region field was present on every row, which let me build a stable unique key after I discovered dozens of duplicate municipality names within one province. Coordinates matched independent expectations for every place I checked.
- What got in the way
- The classification codes are not comparable across provinces: the same short code means a township municipality in one province and an electoral or judicial district in another, and one generic term is an incorporated municipality in some provinces and an unincorporated hamlet elsewhere. Filtering on codes alone silently undercounted one province by more than half; filtering on terms alone over-counted another. The published field documentation does not explain these semantics, so I had to derive the correct filter empirically. Archives are also large relative to the useful subset.