Lazy vs. Eager Execution

In eager mode (the default), operations run immediately:

df = pl.read_csv("players.csv")
result = df.filter(pl.col("level") > 25)

Which, to be fair, is what you'd probably expect from most Python code.

For higher performance, enter lazy mode with df.lazy(), then chain your operations and execute the query plan with .collect():

df = pl.read_csv("players.csv")
result = (
    df.lazy()  # Start lazy mode
    .filter(pl.col("level") > 25)
    .select(["player", "level"])
    .collect()  # Execute the query
)

Lazy mode lets Polars optimize the entire query before it runs. By the time .collect() is called, Polars understands the whole pipeline, not just one step at a time.

For large datasets, this can make a huge difference.

SnackStack's export of device status alerts is getting large enough that the analytics team wants the report built lazily.

Complete the build_lazy_alert_report function. It accepts a Polars DataFrame of device alerts and returns a new report DataFrame.

1. Start lazy mode.
2. Keep only rows where duration_s (i.e., seconds) is greater than 300.
3. Add a column named duration_m (i.e., minutes) by dividing duration_s by 60.
4. Select only the alert_id, device_id, duration_s, and duration_m columns.
5. Execute the lazy query and return the resulting DataFrame.