Skip to content
Docs

Download one month or day of a large table

Large tables are split by time: one file a month (`2026-08.parquet`) or one a day for the busiest feeds (`2026-10-05.parquet`).

GET https://data.transitlab.nyc/v1/{table}/year={year}/month={month}/{file}.parquet No key

A month that would make a file over 450 MB (fhv_trips) comes as two parts, 2025-04.part-0.parquet (days 1 to 15) and 2025-04.part-1.parquet (day 16 on). Take the exact URLs from the catalog’s files list. Supports HTTP Range requests, so a client such as DuckDB can read only the parts it needs.

curl https://data.transitlab.nyc/v1/subway_segment_times/year=2026/month=08/2026-08.parquet \
-o 2026-08.parquet
import requests
r = requests.get(
"https://data.transitlab.nyc/v1/subway_segment_times/year=2026/month=08/2026-08.parquet",
stream=True,
timeout=60,
)
r.raise_for_status()
with open("2026-08.parquet", "wb") as f:
for chunk in r.iter_content(1 << 20):
f.write(chunk)
const res = await fetch("https://data.transitlab.nyc/v1/subway_segment_times/year=2026/month=08/2026-08.parquet")
if (!res.ok) throw new Error(`HTTP ${res.status}`)
const bytes = new Uint8Array(await res.arrayBuffer())

No key needed.

  • table string path required

    Table name. One of subway_segment_times, subway_stop_events, bus_vehicle_positions, bus_predictions, transit_alerts, elevator_outages, elevator_equipment, taxi_yellow_trips, taxi_green_trips, fhv_trips, subway_hourly_ridership, bus_hourly_ridership.

  • year string path required

    Four-digit year.

  • month string path required

    Two-digit month.

  • file string path required

    File name without extension: YYYY-MM for a month, YYYY-MM-DD for a day, YYYY-MM.part-N for one part of a split month.

The Parquet file.

Headers:

  • Accept-Ranges string

  • ETag string

Body:

  • application/vnd.apache.parquet: The file.
  • application/octet-stream: The file.
StatusWhen
404Not found. No such file.