Home

Awesome

ZetaSQL - Analyzer Framework for SQL

ZetaSQL defines a SQL language (grammar, types, data model, semantics, and function library) and implements parsing and analysis for that language as a reusable component. ZetaSQL is not itself a database or query engine. Instead, it's intended to be used by multiple engines, to provide consistent language and behavior (name resolution, type checking, implicit casting, etc.). Specific query engines may implement a subset of features, giving errors for unsupported features. ZetaSQL's compliance test suite can be used to validate query engine implementations are correct and consistent.

ZetaSQL implements the GoogleSQL language, which is used across several of Google's SQL products, both publicly and internally, including BigQuery, Spanner, F1, BigTable, Dremel, Procella, and others.

GoogleSQL and ZetaSQL have been described in these publications:

Some other documentation:

Project Overview

The main components and APIs are in these directories under zetasql/:

Multiplatform support is planned for the following platforms:

We do not provide any guarantees of API stability and cannot accept contributions.

Running Queries with execute_query

The execute_query tool can parse, analyze and run SQL queries using the reference implementation.

See Execute Query for more details on using the tool.

You can run it using binaries from Releases, or build it using the instructions below.

There are some runnable example queries in zetasql/examples/tpch and zetasql/examples/pipe_queries.

Getting and Running execute_query

Pre-built Binaries

ZetaSQL provides pre-built binaries for execute_query for Linux and MacOS on the Releases page. You can run the downloaded binary like:

chmod +x execute_query_linux
./execute_query_linux --web

MacOS users may see the error execute_query_macos cannot be opened because the developer cannot be verified.. You can right click the execute_query_macos file, click "open", and then you should be able to run the binary.

Note the prebuilt binaries require GCC-9+ and tzdata. If you run into dependency issues or if the binary is incompatible with your platform, you can try running execute_query with Docker. See the Run with Docker section.

Running from a Bazel Build

You can build execute_query with Bazel from source and run it by:

bazel run zetasql/tools/execute_query:execute_query -- --web

Run with Docker

You can run execute_query using Docker. Download the pre-built Docker image file zetasql_docker.tar.gz from the Releases page, and load the image using:

sudo docker load -i /path/to/the/downloaded/zetasql_docker.tar.gz

The Docker image name is zetasql. (You can also build a Docker image locally using the instructions in the Build with Docker section.)

You can then run execute_query using:

sudo docker run --init -it -h=$(hostname) -p 8080:8080 zetasql execute_query --web

If you are using MacOS with an Apple M1/M2 chip, add the additional argument --platform=linux/amd64:

sudo docker run --init -it -h=$(hostname) -p 8080:8080 --platform linux/amd64 zetasql execute_query --web

Argument descriptions:

-h=$(hostname) and -p 8080:8080 together make the URL address of the web server accessible from the host machine.

Alternatively, you can run this to start a bash shell, and then run execute_query inside:

sudo docker run --init -it -h=$(hostname) -p 8080:8080 my-zetasql-image

# Inside the container bash shell
execute_query --web

How to Build

Build with Bazel

ZetaSQL uses Bazel for building and dependency resolution. Instructions for installing Bazel can be found in https://bazel.build/install. The Bazel version that ZetaSQL uses is specified in the .bazelversion file.

Besides Bazel, the following dependencies are also needed:

tzdata provides the support for time zone information. It is generally available on MacOS. If you run Linux and it is not pre-installed, you can install it with apt-get install tzdata.

Once the dependencies are installed, you can build or run ZetaSQL targets as needed, for example:

# Build everything.
bazel build ...

# Build and run the execute_query tool.
bazel run //zetasql/tools/execute_query:execute_query -- --web

# The built binary can be found under bazel-bin and run directly.
bazel-bin/zetasql/tools/execute_query/execute_query --web

# Build and run a test.
bazel test //zetasql/parser:parser_set_test

Some Mac users may experience build issues due to the Python error ModuleNotFoundError: no module named 'google.protobuf'. To resolve it, run pip install protobuf==<version> to install python protobuf. The protobuf version can be found in the zetasql_deps_step_2.bzl file.

Build with Docker

ZetaSQL also provides a Dockerfile which configures all the dependencies so that users can build ZetaSQL more easily across different platforms.

To build the Docker image locally (called my-zetasql-image here), run:

sudo docker build . -t my-zetasql-image -f Dockerfile

Alternatively, ZetaSQL provides pre-built Docker images named zetasql. See the Releases page. You can load the downloaded image by:

sudo docker load -i /path/to/the/downloaded/zetasql_docker.tar.gz

To run builds or other commands inside the Docker environment, run this command to open a bash shell inside the container:

# Start a bash shell running inside the Docker container.
sudo docker run -it my-zetasql-image

Replace my-zetasql-image with zetasql if you use the pre-built Docker image.

Then you can run the commands from the Build with Bazel section above.

Differential Privacy

For questions, documentation, and examples of ZetaSQL's implementation of Differential Privacy, please check out (https://github.com/google/differential-privacy).

Versions

ZetaSQL makes no guarantees regarding compatibility between releases. Breaking changes may be made at any time. Our releases are numbered based on the date of the commit the release is cut from. The number format is YYYY.MM.n, where YYYY is the year, MM is the two digit month, and n is a sequence number within the time period.

License

Apache License 2.0

Support Disclaimer

This is not an officially supported Google product.