PowerSQL

PowerSQL, the data transformation tool.

Features:

• Supports many SQL dialects
• Plain SQL helps to integrate with your favourite database tools, formatters and linters: simple add your CREATE VIEW or CREATE TABLE AS statements.
• Syntax & type checking avoid errors
• Automatically creates and executes a dependency graph.

Install the latest version using cargo (curl https://sh.rustup.rs -sSf | sh).

# For PostgreSQL
cargo install powersql --features postgres
# For BigQuery
cargo install powersql --features bigquery

To get started with PostgreSQL, simply create a new project in a file called powersql.toml:

[project]
name = "my_project"
models = ["models"]
tests = ["tests]

Now create one or more models in the models directory:

CREATE VIEW my_model AS SELECT id, category from my_source;
CREATE TABLE category_stats AS SELECT COUNT(*) category_count FROM my_model GROUP BY category;

PowerSQL automatically will create a DAG based on the relations in your database.

To run against the database, provide the following environment variables:

• PG_HOSTNAME
• PG_USERNAME
• PG_PORT
• PG_DATABASE
• PG_PASSWORD

To run against the database, provide the following environment variables:

• GOOGLE_APPLICATION_CREDENTIALS
• PROJECT_ID
• DATASET_ID
• LOCATION

GOOGLE_APPLICATION_CREDENTIALS should refer to an service account key file (this can be set by an appliation rather than locally).

PROJECT_ID is the id (not number) of the project and DATASET_ID is the name of the dataset that is used by default.

LOCATION is an (optional) datacenter location id where the query is being executed.

• powersql check: This will load all your .sql files in the directories listed in models. It will check the syntax of the SQL statements. After this, it will check the DAG and report if there is a circular dependency. Finally, it will run a type checker and report any type errors.
• powersql run: Loads and runs the entire DAG of SQL statements.
• powersql test: Loads and runs the data tests.

Data tests are SQL queries that you can run on your database tables and views and perform checks on data quality, recency, etc. The test fails if the query returns 1 or more rows.

Some examples:

-- NULL check
SELECT 1 FROM t WHERE column IS NULL;
-- Check values
SELECT 1 FROM t WHERE amount < 0;
-- Check relations
SELECT 1 FROM t LEFT JOIN u ON t.id = u.id WHERE u.id IS NULL;
-- Prefix check
SELECT 1 FROM t WHERE NOT STARTS_WITH(str_column, "http");