[jay23606]

Problem

The bounding box of the BIXI Montréal GBFS system spans from Montreal to Null Island. Its station_information feed contains three stations (304, 724, 747) with location (0, 0). reverse_geolocation_process builds the dataset bounding box from the min/max of all station coordinates, so these placeholder points stretch the box to (0, 0).

Change

In functions-python/reverse_geolocation/src/reverse_geolocation_processor.py, the stop filter that already drops missing coordinates now also drops rows where both stop_lat and stop_lon are 0. Those points are excluded from the bounding box and from reverse geolocation. Obvious outliers other than (0, 0) are not handled here.

Testing

Added test_valid_request_excludes_zero_zero_coordinates in functions-python/reverse_geolocation/tests/test_reverse_geolocation_processor.py. It passes a stop DataFrame with two (0, 0) rows and two valid rows and asserts that update_dataset_bounding_box receives only the two valid rows.

The test was not run in the authoring environment: its Python dependencies (pandas, etc.) are not installed there, so CI is the verification. The Windows Python check python -c "import pandas, shapely, geoalchemy2, flask, sqlalchemy; print('ok')" exited with ModuleNotFoundError: No module named 'pandas'.

Regression test: untested: CONTRIBUTING does not mention tests

Fixes #1887

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}