gtfs-dagster/user_code/sensors/gtfs_realtime.py

77 lines
2.7 KiB
Python

from dagster import (
sensor,
SensorEvaluationContext,
RunRequest,
SkipReason,
DefaultSensorStatus,
AssetKey,
)
from dagster_duckdb import DuckDBResource
@sensor(
name="gtfs_rt_vehicles_sensor",
minimum_interval_seconds=60,
asset_selection=["gtfs_rt_vehicles_downloads"],
default_status=DefaultSensorStatus.RUNNING
)
def gtfs_rt_vehicles_sensor(
context: SensorEvaluationContext,
duckdb: DuckDBResource,
) -> list[RunRequest] | SkipReason:
"""
Sensor that triggers gtfs_rt_vehicles_downloads every 60 seconds.
Fetches feed metadata once and passes it to each partition run.
"""
# Check if upstream asset has been materialized at least once
upstream_asset_key = AssetKey("gtfs_rt_vehicles_partitions")
latest_materialization = context.instance.get_latest_materialization_event(upstream_asset_key)
if latest_materialization is None:
return SkipReason(
"Waiting for upstream asset 'gtfs_rt_vehicles_partitions' to be materialized. "
"Run the upstream assets first."
)
try:
with duckdb.get_connection() as conn:
# Get all active feeds with their metadata in one query
feeds = conn.execute("""
SELECT feed_id, provider, producer_url
FROM gtfs_rt_vehicles_metadata
WHERE producer_url IS NOT NULL AND producer_url != ''
ORDER BY feed_id
""").fetchall()
if not feeds:
return SkipReason("No GTFS-RT vehicle feeds configured")
# Create a RunRequest for each partition with metadata
run_requests = [
RunRequest(
partition_key=feed_id,
run_config={
"ops": {
"gtfs_rt_vehicles_downloads": {
"config": {
"provider": provider,
"producer_url": producer_url,
}
}
}
},
tags={
"feed_id": feed_id,
"sensor": "gtfs_rt_vehicles_sensor"
}
)
for feed_id, provider, producer_url in feeds
]
context.log.info(f"Triggering downloads for {len(run_requests)} GTFS-RT vehicle feeds")
return run_requests
except Exception as e:
# Handle case where table doesn't exist yet or other DB errors
context.log.warning(f"Database query failed: {e}")
return SkipReason(f"Database not ready or query failed: {e}")