Pants adhoc_tool examples

This is an example repository to demonstrate support for unsupported languages and arbitrary 3rd-party tools in Pants.

The examples use the adhoc_tool and system_binary targets, that were added as experimental features in Pants 2.16.0a0, and will

See pantsbuild.org for much more detailed documentation of adhoc_tool and system_binary Full documentation will available before Pants 2.16 stable release.

This is only one possible way of laying out your project with Pants. See pantsbuild.org/docs/source-roots#examples for some other example layouts.

This repository demonstrates advanced uses of Pants. For more introductory use cases, consider looking at example-python or example-jvm.

Examples

Using a JVM artifact from Maven to generate Python source code

Using adhoc_tool, you can run a Maven artifact that's declared by a jvm_artifact target. We can use that to run the JVM-based antlr parser generator to transparently produce Python bindings, which can then be imported from our first-party Python code.

To see the demo in practice, run ./pants run antlr/antlr_demo.py.

This demo uses:

• jvm_artifact to declare a dependency on the Antlr parser generator
• adhoc_tool which asks Pants to run the Antlr dependency as a build step, outputting files containing Python bindings (as loose files)
• experimental_wrap_as_python_sources, which allows subsequent steps to consume the loose files as Python sources that can be imported.

Note that sources declared by experimental_wrap_as_* targets can not currently be detected using Dependency Inference.

Building a JavaScript asset for inclusion in a web application

Using system_binary, you can declare dependencies on tools that are managed externally to Pants, including basic compatibility checks by way of version constraints.

This allows you to use tools from languages that aren't directly supported by Pants. We can use that to manage a node_modules directory using yarn and node binaries that were installed onto the host system (e.g. by Homebrew or apt).

Our demo produces a simple CLI script that imports an npm dependency and functions from a first-party library and links them together using Parcel. Package resolution and tool execution is handled by yarn.

To see the demo in practice, run ./pants run javascript:run-js-app, or ./pants package javascript:packaged-js to package the JavaScript code into a zip file.

This demo uses:

• system_binary to declare dependencies on node and yarn binaries. The fingerprint* fields are used to declare version constraints that can be used to ensure builds are reproducible across multiple execution environments.
• adhoc_tool to execute the yarn install and yarn parcel commands.
• run_shell_command to run the generated JavaScript artifact with node.
• archive to package the generated JavaScript archive into a zip file

Automatically generating JavaScript bindings for a Flask web app with an OpenAPI schema

Using adhoc_tool, you can run individual first-party sources to fetch their output. This can produce inputs for other targets.

In this example, a small Flask web application can dump an OpenAPI schema, which can in turn be used to transparently generate bindings for other languages.

To see the demo in practive, run ./pants export-codegen openapi:webapp-js-bindings

This demo uses

• python_source to declare a runnable Python source, webapp.py.
• jvm_artifact to declare a dependency on the JVM-based OpenAPI client generator
• adhoc_tool to run the webapp.py source, and saving its stdout to a file using the stdout= field
• adhoc_tool to run the JVM client generator