Create Test Datasets Agent Skill

Create Test Datasets Skill

Set up test datasets by executing standard creation scripts. Supports hits, visits, tpcds, and tpch.

Arguments

• $ARGS (optional): Space-separated list of dataset names from {hits, visits, tpcds, tpch}. If no arguments are provided, defaults to hits visits.

Examples:

• /create-test-datasets — sets up hits and visits (default)
• /create-test-datasets hits — sets up only hits
• /create-test-datasets tpch tpcds — sets up TPC-H and TPC-DS
• /create-test-datasets hits visits tpch — sets up all three

Process

1. Parse arguments

Parse the argument string into a set of requested datasets. Valid values are hits, visits, tpcds, tpch. If no arguments are provided, use {hits, visits} as the default. If any argument is not in the valid set, report an error and stop.

2. Verify the server is running

clickhouse-client -q "SELECT 1" 2>&1

If the server is not reachable, report an error and stop. Do not attempt to start the server — ask the user to start it first.

3. Check for conflicting tables

Before making any changes, check whether any target tables already exist. Query system.tables once:

clickhouse-client -q "SELECT database || '.' || name FROM system.tables WHERE
    (database = 'test' AND name IN ('hits', 'visits'))
    OR (database = 'tpcds')
    OR (database = 'tpch')"

Build a list of conflicting tables based on what was requested:

• If hits is requested: test.hits must not exist.
• If visits is requested: test.visits must not exist.
• If tpcds is requested: the tpcds database must have no tables.
• If tpch is requested: the tpch database must have no tables.

If any conflicts are found, report all of them to the user and stop. Do not drop or overwrite anything. Suggest the user drop the conflicting tables manually and rerun the skill.

4. Set up hits and/or visits

Only execute this step if hits or visits (or both) are in the requested set.

The file tests/docker_scripts/create.sql contains two CREATE TABLE statements: one for datasets.hits_v1 and one for datasets.visits_v1. For each requested dataset, extract only the corresponding statement, replace the table name to create it directly as test.hits or test.visits, and execute it. This avoids creating unnecessary intermediate tables.

clickhouse-client -q "CREATE DATABASE IF NOT EXISTS test"

If hits is requested — extract the datasets.hits_v1 statement, replace the name, and execute:

clickhouse-client --multiquery <<< "$(sed -n '/^CREATE TABLE datasets\.hits_v1/,/);/p' tests/docker_scripts/create.sql | sed 's/datasets\.hits_v1/test.hits/')"

If visits is requested — extract the datasets.visits_v1 statement, replace the name, and execute:

clickhouse-client --multiquery <<< "$(sed -n '/^CREATE TABLE datasets\.visits_v1/,/);/p' tests/docker_scripts/create.sql | sed 's/datasets\.visits_v1/test.visits/')"

5. Set up TPC-DS

Only execute this step if tpcds is in the requested set.

bash tests/docker_scripts/create_tpcds.sh

6. Set up TPC-H

Only execute this step if tpch is in the requested set.

bash tests/docker_scripts/create_tpch.sh

7. Verify

For each dataset that was set up, confirm it exists and report row counts:

• If hits: SELECT 'test.hits', count() FROM test.hits
• If visits: SELECT 'test.visits', count() FROM test.visits
• If tpcds: SELECT name, total_rows FROM system.tables WHERE database = 'tpcds' ORDER BY name
• If tpch: SELECT name, total_rows FROM system.tables WHERE database = 'tpch' ORDER BY name

Report all results to the user.