nyc-ride-market

L09 · Metrics library

Wave 2 · Phase 2–3 · Spec §6 · Feeds L10, L11

Goal

Every SPEC §6 metric defined once, as SQL over the aggregate tables, each with a test whose expected value was worked out by hand.

Owns

• sql/metrics/*.sql, except cp_effect.sql and pay_event.sql (L10)
• pipeline/metrics.py: registers data/agg/*.parquet as views, runs a named metric with parameters, returns an Arrow table; CLI python -m pipeline.metrics <name> and --all
• tests/test_metrics_<name>.py

Tasks

• One file per metric listed in contracts/metrics.md, each with a header comment: question(s), params, grain, output columns
• Ratios of sums only; divide with NULLIF(denominator, 0)
• Filter modes as SPEC says (driver share: uber, lyft); taxi tip rate uses card trips only
• trips_per_hour_worked reads ref_indicators and is marked TLC-reported in its output
• Output columns match contracts/metrics.md exactly (tested)

Tests

• One per metric: a factory fixture, with the expected value worked out by hand in a comment

Done when

• python -m pipeline.metrics --all runs on the G2 aggregates without error
• Tests green

Not this lane

• The congestion and pay-rise studies (L10); shaping site files or writing findings (L11)

---

Status log

Only this lane writes here. Newest first.

Date	State	Note
	not started

Blocked: none

Hand-off

• Built:
• Checked:
• Known limits:
• Follow-ups: