bq CLI quickstart (G5)

Runnable five-command bq sequence against a fresh bqemulator instance: create a dataset, create a table, load a 3-row NDJSON file, query, and clean up.

This example is the cited reference for the Using the bq CLI guide. CI runs make test against this directory so the README's commands cannot rot.

Prerequisites

1. bqemulator on PATH (pip install bqemulator).
2. bq on PATH (install the gcloud SDK).

Run

make test

The run.sh script:

1. Spins up an ephemeral emulator on a random port via bqemulator start --ephemeral --rest-port 0.
2. Waits for /healthz to return ok.
3. Runs the five bq commands below.
4. Asserts the final query returns n=3.
5. Tears down the emulator process.

What gets exercised

# 1) Create a dataset.
bq --api=$EMU mk --dataset --location=US demo:demo_ds

# 2) Create a table with a schema.
bq --api=$EMU mk --table demo:demo_ds.customers \
    id:INTEGER,name:STRING,email:STRING

# 3) Load 3 NDJSON rows from a local file.
bq --api=$EMU load \
    --source_format=NEWLINE_DELIMITED_JSON \
    demo:demo_ds.customers /tmp/customers.ndjson

# 4) Query.
bq --api=$EMU query --use_legacy_sql=false --format=json \
    'SELECT COUNT(*) AS n FROM `demo.demo_ds.customers`'

# 5) Clean up.
bq --api=$EMU rm -r -f -d demo:demo_ds

Each command touches a different REST surface; the load command goes through the upload host endpoints, and the query command goes through jobs.insert + jobs.getQueryResults. A regression on any of those paths breaks this example — which is why CI runs it.