Punctuality

# pip install polars requests
import io
import requests
import polars as pl

from datetime import datetime, timedelta, timezone

# Yesterday's Brussels-day file (the harvester anchors windows to local midnight,
# so any timestamp inside that day returns the same parquet).
yesterday_noon = (datetime.now(timezone.utc) - timedelta(days=1)).replace(
    hour=12, minute=0, second=0, microsecond=0
)
params = {"timestamp": yesterday_noon.isoformat()}  # ISO accepted; epoch also works

url = "https://api.mobilitytwin.brussels/de-lijn/punctuality"
data = requests.get(url, headers={
    "Authorization": "Bearer [MY_API_KEY]"
}, params=params).content

df = pl.read_parquet(io.BytesIO(data))

print(f"rows: {df.height:,}, columns: {df.columns}")

# Average arrival delay per route, ignoring cancellations and no-data rows.
delays = (
    df.filter(pl.col("arrival_delay").is_not_null() & ~pl.col("cancelled"))
      .group_by("route_id")
      .agg([
          pl.col("arrival_delay").mean().round(1).alias("mean_delay_s"),
          pl.col("arrival_delay").quantile(0.9).alias("p90_delay_s"),
          pl.len().alias("n"),
      ])
      .sort("mean_delay_s", descending=True)
)
print("\nWorst 10 routes by mean arrival delay:")
print(delays.head(10))

# How many trips were cancelled and on which routes?
cancellations = (
    df.filter(pl.col("cancelled"))
      .select(["route_id", "trip_id", "start_date", "start_time"])
      .unique()
)
print(f"\ncancelled trip-instances: {cancellations.height:,}")
print(cancellations.group_by("route_id").len()
      .sort("len", descending=True).head(10))