A Paris public-health mapping team receives a contractor's pharmacy table before a neighbourhood service review. The directory loader expects unique references, names without padding, and WGS 84 points inside the survey window; rejected rows must not become invented locations.
Prepare the raw pharmacy handover as a point layer. The source is a teaching extract of OpenStreetMap pharmacy nodes with deliberate handover faults.
| Incoming field or fault | Required output |
|---|---|
site_id repeats | Keep the first occurrence in file order. |
first_coord, second_coord are text; one may use a comma decimal | Parse them as numbers. Interpret them using coord_order (lon_lat or lat_lon). |
| Blank coordinates or coordinates outside the survey window | Omit that record; do not turn a blank into zero. |
name_raw includes possible padding | Trim its outer whitespace into name. |
Return exactly site_id (text), name (text), and point geometry in EPSG:4326. Every point must lie within [2.29, 48.84, 2.31, 48.85]. In Python read /data/paris-pharmacy-handover-raw.geojson and assign the prepared GeoDataFrame to result.
These are where the teaching is. Read them twice.
Modelled on municipal point-data intake. ISO 19157-1:2023 requires quality measures against a product specification; the file's acceptance rules are stated here. Source: © OpenStreetMap contributors, ODbL 1.0.
Files: /data/paris-pharmacy-handover-raw.geojson · assign result
6 scored, 0 informational
Prepared records agree with the handover rules
50%One point per source reference
10%Every delivered geometry is valid
10%Points fall within the stated survey window
10%Handover fields match the directory schema
10%References and names remain text
10%
