arcli sample

arcli sample puts a real dataset into a database you own, then hands you queries worth running against it. A synthetic data point proves the write path works; it does not show you what Arc is for.

Quick reference

arcli sample list                    # what is published
arcli sample show citibike           # details, licence, time range
arcli sample load citibike           # download, verify, import

Loading a dataset

$ arcli sample load citibike
Created database "nyc"
Dataset:     NYC Citi Bike trips (citibike), 2025-12
Source:      https://samples.basekick.net
Download:    31 files, 39.4 MiB, 1003770 rows
Directory:   /tmp/arcli-sample/citibike/2025-12
Target:      nyc.citibike_trips
Licence:     NYCBS Data Use Policy - https://ride.citibikenyc.com/data-sharing-policy

Verified 31 files against the published checksums.
Imported 31 files, 1003770 rows into nyc.citibike_trips

Try these:

  # Busiest stations
  arcli query --database nyc 'SELECT start_station_name AS station, count(*) AS trips FROM citibike_trips GROUP BY 1 ORDER BY trips DESC LIMIT 5'

  # Departures per hour, with a 3-hour moving average
  arcli query --database nyc 'WITH hourly AS (...) SELECT hr, trips, round(avg(trips) OVER (ORDER BY hr ROWS BETWEEN 2 PRECEDING AND CURRENT ROW), 1) AS moving_avg_3h FROM hourly ORDER BY hr LIMIT 24'

The printed queries are ready to paste. The first returns W 21 St & 6 Ave at the top with 4,070 trips; the second shows the commute curve, peaking at 08:00 and 17:00.

Importing needs an admin token when the Arc server has authentication enabled. That is checked before anything is downloaded, so a permissions problem costs you nothing.

Flags

Behaviour worth knowing

Every file is checksum-verified. Each part's SHA-256 is compared against the published manifest before anything is imported. A file that does not match is re-downloaded once, and if it still does not match the command fails rather than importing data that does not match the manifest.

Downloads resume. A file already present with a matching checksum is not fetched again, so an interrupted run picks up where it left off. Files download four at a time, and one failure stops the rest.

A second run asks first. If the target measurement already holds rows, load reports the count and asks before adding more, so a re-run cannot silently double the numbers the showcase queries report. --yes skips the prompt; a non-interactive shell refuses rather than hanging.

Inspecting without downloading

arcli sample show <dataset> -o json prints the full manifest entry, including every file URL and SHA-256:

arcli sample show citibike -o json

That is also the escape hatch if you would rather not fetch from Basekick at all: download the files however you like, verify them against those checksums, and load them with arcli import parquet.

About the data

NYC Citi Bike trip records, published by NYC Bike Share under the NYCBS Data Use Policy, one month per dataset entry. The full archive lives at the NYC source.

The time column carries local New York wall-clock time, which is what makes the hourly curve peak where a reader expects. Note that Arc stores every timestamp column other than time as integer microseconds, so started_at and ended_at need to_timestamp(x/1000000) if you want to read them as timestamps.

Privacy

These are the only arcli commands that contact a Basekick-controlled host. They send the same User-Agent and installation id as any other arcli request, honouring DO_NOT_TRACK=1 and send_installation_id = false, and never your token, your Arc endpoint or anything about your data. See Privacy for the detail, including what any file download necessarily reveals.