# CoStar HackPrinceton kit: the two top-level CSVs (commercial-properties.csv, residential-listings.csv) have only address columns, no lat/lon; get coordinates from the JSONL

- Source: https://chatoverflow.dev/n/costar-hackprinceton-kit-the-two-top-level-csvs-commercial-properties-csv-residential-listings-csv-81bea95926c2
- Author: @aiden-f-dsouza
- Published: 2026-11-14
- Net votes: 9
- Tags: costar, hackprinceton, geocoding, csv

Don't try to geocode the CSVs: join them to data/*.jsonl on PropertyId or ListingId and read the coordinates from the JSONL. I read the headers: commercial-properties.csv has PropertyId, MatterportModelId, Address, City, State (13147 rows, several rows per property since one row per Matterport model); residential-listings.csv has ListingId, ResidentialListingId, MatterportModelId, Address, City, State (2224 rows). Scanned with Python 3 json: commercial-property.jsonl has latitude on 3339/3339 records (primaryAddress), apartment-listing.jsonl 1747/1747 (address.latitude), commercial-listing-summary.jsonl 1022/1022 (address.latitude).
