PATH ridership stats

Cleaned + plotted PANYNJ PATH faregate + hourly ridership.

Live site: path.hudcostreets.org

Data

• data/all.pqt — per (month, station): total + avg-per-day for weekday / weekend / holiday
• data/all.xlsx — Excel copy of the above
• Google Sheet
• data/YYYY-hourly.pqt — per (station, hour, month), avg per weekday / Sat / Sun / holiday
• www/public/entries_vs_exits.pqt — per (ym, station), avg entries + exits per day-type
• www/public/hourly.pqt — the browser-served hourly parquet (zstd, int32-downcast)

Larger artifacts (parquets, PDFs, the pie-map GIF/MP4) are DVX-tracked (.dvc pointers in git, blobs on S3); dvx pull fetches the current versions.

Pipeline

The daily update-path-data.yml cron runs path-data gha-update:

1. path-data refresh — download the latest PANYNJ ridership PDFs
2. dvx run — re-parse any changed years and rebuild derived artifacts (path-data monthly -y YYYY, path-data parse-hourly -y YYYY, path-data combine, path-data combine-hourly, path-data entries-vs-exits)
3. dvx add + dvx push — snapshot new outputs to S3
4. git commit and, if www/public/**/*.dvc actually changed, gh workflow run www.yml to redeploy the site

Local dev:

git clone https://github.com/hudcostreets/path
cd path
pip install -e .
path-data --help

Web frontend lives at www/ — Vite + React + Plotly + Leaflet, deployed to GitHub Pages via .github/workflows/www.yml. Any push touching www/** (including new .dvc pointers to fresh data) redeploys automatically.

Bridge & Tunnel

Same repo also serves /bt — PANYNJ B&T traffic (Lincoln + Holland tunnels, GWB, Bayonne + Goethals + Outerbridge). Merge per-year traffic-e-zpass-usage-*.pdf into one PDF for parsing:

gs -o merged.pdf \
   -sDEVICE=pdfwrite \
   -dPDFFitPage \
   -g12984x10033 \
   -dPDFSETTINGS=/prepress \
   traffic-e-zpass-usage-20*

(cf. SO)