@nxlv/python v21.0.3

@nxlv/python

@nxlv/python plugin is designed to extend the Nx features to work with Python projects based on Uv or Poetry.

Check this article for a step by step tutorial with Poetry: https://lucasvieirasilva.medium.com/poetry-python-nx-monorepo-5750d8627024

What is @nxlv/python

🔎 An Nx Plugin to generate and manage Python projects in the Nx workspace.

Main Features

• Generate Python projects using Uv or Poetry.
• Manage local and external dependencies.
• Build and publish Python packages.
• Bundle local dependencies, like ESBuild for JavaScript.
• Linting and Unit Testing.
• Nx Dependency graph integration.

Motivation

Nx is a powerful tool to manage monorepos, however, it doesn't have built-in support for Python projects, this plugin aims to fill this gap by providing a set of executors to manage Python projects in the Nx workspace, and levearge the Nx features like dependency graph, affected commands, local cache, remote cache, releases, and more to Python projects.

Why Bundle Local Dependencies

When working with monorepos, it's common to have local dependencies, like shared libraries, that are used across multiple projects and don't necessarily need to be published to a package repository, the @nxlv/python plugin provides a way to bundle these local dependencies in the same wheel file.

This approach helps to deploy the projects to Cloud services like AWS Lambda, Google Cloud Functions, Azure Functions, and others, where the deployment package needs to be a single file.

News

• We now support Uv package manager.

Nx 21 Release Notes

This plugin supports the new Nx 21 release versioning.

Migrating Nx release versioning

Update all python project.json files to use the new versionActions property.

From:

{
  "name": "app1",
  "$schema": "../../node_modules/nx/schemas/project-schema.json",
  "projectType": "application",
  "sourceRoot": "apps/app1/app1",
  "targets": {
    ...
  },
  "tags": [],
  "release": {
    "version": {
      "generator": "@nxlv/python:release-version"
    }
  }
}

To:

{
  "name": "app1",
  "$schema": "../../node_modules/nx/schemas/project-schema.json",
  "projectType": "application",
  "sourceRoot": "apps/app1/app1",
  "targets": {
    ...
  },
  "tags": [],
  "release": {
    "version": {
      "versionActions": "@nxlv/python/src/release/version-actions"
    }
  }
}

NOTE: If you have other plugins that don't support the new release versioning, keep the release.version.generator property in the project.json file as is and update the nx.json file to use the legacy versioning.

nx.json

{
  "release": {
    ...
    "version": {
      ...
      "useLegacyVersioning": true
    }
  }
}

For now, the @nxlv/python plugin generates new projects with the legacy versioning, so, if you want to use the new release versioning, add the --useNxReleaseLegacyVersioning option to the project generator.

Poetry Project:

nx generate @nxlv/python:poetry-project myproject --useNxReleaseLegacyVersioning=false

Uv Project:

nx generate @nxlv/python:uv-project myproject --useNxReleaseLegacyVersioning=false

or update the nx.json file to default the project generator to use the new release versioning.

nx.json

Poetry Project:

{
  ...
  "generators": {
    "@nxlv/python:poetry-project": {
      "useNxReleaseLegacyVersioning": false
    }
  }
  ...
}

Uv Project:

{
  ...
  "generators": {
    "@nxlv/python:uv-project": {
      "useNxReleaseLegacyVersioning": false
    }
  }
  ...
}

Getting Started

Add to an existing Nx Workspace

Install the npm dependency

npm install @nxlv/python --save-dev

Usage

1. Update nx.json to add the property plugins with @nxlv/python value.

Example:

{
  ...
  "plugins": [
    "@nxlv/python"
  ]
  ...
}

for Nx 20.x or higher, use the following pattern:

{
  ...
  "plugins": [
    ...
    {
      "plugin": "@nxlv/python"
    }
  ]
  ...
}

IMPORTANT: To use uv package manager without workspaces, set the packageManager option to uv in the nx.json file, as shown below:

{
  ...
  "plugins": [
    ...
    {
      "plugin": "@nxlv/python",
      "options": {
        "packageManager": "uv"
      }
    }
  ]
  ...
}

NOTE: The default package manager is poetry, but it's automatically detected if the repository is configured to use uv workspaces since the uv.lock filw will be present in the root directory.

💖 Support

If you find this project useful, consider buying me a coffee or sponsoring me via GitHub Sponsors!

Poetry

Add a new Python Project

nx generate @nxlv/python:poetry-project myproject

NOTE: On Nx v16.8.0 or higher, the generators are prompting the user to choose how Nx will calculate where the project should be located, by using the --projectNameAndRootFormat with as-provided or derived.

The as-provided option uses the --directory option to calculate the project root, so, if the --directory option is not provided, the project root will be the same as the project name, and the project name will be the same as provided in the --name option.

Example 1: nx generate @nxlv/python:poetry-project myproject will generate the project in the myproject folder with the name myproject. Example 2: nx generate @nxlv/python:poetry-project myproject --directory=api will generate the project in the api folder with the name myproject.

The derived option uses the combination of the workspace layout, --directory option, and --name to calculate the name and root of the project.

Example 1: nx generate @nxlv/python:poetry-project myproject will generate the project in the apps/myproject folder with the name myproject. Example 2: nx generate @nxlv/python:poetry-project myproject --directory=api will generate the project in the apps/api/myproject folder with the name api-myproject.

Nx documentation reference: https://nx.dev/deprecated/as-provided-vs-derived#project-generators

Options

rootPyprojectDependencyGroup

When the workspace is configured to use a shared virtual environment (see below), the rootPyprojectDependencyGroup option specifies the dependency group to be used in the root pyproject.toml file, by default, the main dependency group is used.

devDependenciesProject

This approach consists of moving all the dev dependencies from the projects to separate projects, this project is referenced in the root pyproject.toml and all the local projects as a dev dependency.

Benefits:

• Centralize the dev dependencies in a single project

templateDir

The templateDir option specifies a custom directory to be used as a template for the project, by default, the @nxlv/python has a built-in template folder that is used to generate the project.

However, there are some cases where the developer wants to use different files or customize the way the project is generated without the need to create a custom generator based on the @nxlv/python generator.

The files in the templateDir needs to follow the Nx generator convention by using EJS to customize the files based on the options provided by the @nxlv/python.

Template variables

All the options listed above are available as variables in the template files, for example, to use the packageName option in the template file, use <%= packageName %>.

Additional variables are available in the template files:

• offsetFromRoot: This variable contains the relative path from the project to the root directory. (e.g. for the project folder apps/my-project the offsetFromRoot value will be ../../)
• projectRoot: This variable contains the relative path of the project.
• individualPackage: This variable is a boolean that indicates if the workspace is using a shared virtual environment or not.
• dot: This variable is a string that contains a dot (.), it is used to create files that start with a dot (e.g. __dot__gitignore.template).
• template: This variable is an empty string, this variable is usually combined with the dot variable, because the Nx generator file function only generates files with extensions, however, there are some cases where the file doesn't have an extension (e.g. .gitignore), so, the template variable is used to create files without an extension (e.g. __dot__gitignore.template).

Global Default Options

By default, @nxlv/python:poetry-project generator defines linter and unit test runner with all reports enabled by default, however, those default options can be globally changed by using the following configuration in the nx.json config file.

{
  ...
  "generators": {
    "@nxlv/python:poetry-project": {
      "unitTestHtmlReport": false,
      "codeCoverageThreshold": 100,
      "devDependenciesProject": "shared-development"
    }
  }
  ...
}

The property names are the same as the options listed in the table above.

Add a new dependency to a project

nx run {project}:add --name {projectName} --local

Add an external dependency to the project

To add a new dependency to the project use the nx run {project}:add command detailed below. This ensures that any dependent projects are updated.

nx run {project}:add --name {dependencyName}

Uv

nx generate @nxlv/python:uv-project myproject

Options

Shared Virtual Environment

By default, the @nxlv/python manages the projects individually, so, all the projects have their one set of dependencies and virtual environments.

However, In some cases, we want to use a shared virtual environment for the entire workspace to save some installation time in your local environment and CI tool, we use this mode when the workspace contains many projects with the same dependencies and versions that don't conflict in the workspace level.

To migrate to this mode, run the following command:

npx nx generate @nxlv/python:migrate-to-shared-venv

Options:

After the migration is completed, the workspace now has a pyproject.toml in the root directory, and all the local projects are referencing the root pyproject.toml file.

The projects still have their own pyproject.toml file to manage each project's dependencies, however, the package versions cannot conflict because the root pyproject.toml file is referencing all the dependencies.

Benefits:

• Save time in the local environment and CI tool
• Reduce the size of the workspace
• Reduce the number of dependencies installed in the local environment and CI tool
• Single-version policy (recommended by Nx)
• Better VSCode integration (currently, the VSCode Python extension doesn't support multiple virtual environments in the same workspace, it needs to switch between them manually)

Cons:

• Package versions cannot conflict at the workspace level
• Local packages with the same module name don't work properly in the VSCode, because when the VSCode Python extension is activated, it uses the root pyproject.toml file to resolve the packages, so, it will use the first module found in the pyproject.toml file.

Executors

add

The @nxlv/python:add executor handles add command to provide a level of abstraction and control in the monorepo projects.

• poetry: poetry add {args}
• uv: uv add {args}

Features

• Add new external dependencies
• Add local dependencies

Both features updates the local workspace dependency tree to keep the lock/venv updated.

Options

update

The @nxlv/python:update executor handles update command to provide a level of abstraction and control in the monorepo projects.

• poetry: poetry update {args}
• uv: Uv doesn't have a native update command, so, the executor runs the following commands:
  • uv lock --upgrade-package {name}
  • uv sync

Features

• Update external dependencies
• Update local dependencies

Both features updates the local workspace dependency tree to keep the lock/venv updated.

Options

remove

The @nxlv/python:remove executor handles remove command to provide a level of abstraction and control in the monorepo projects.

• poetry: poetry remove {args}
• uv: uv remove {args}

Features

• Remove external dependencies
• Remove local dependencies

Both features updates the local workspace dependency tree to keep the lock/venv updated.

Options

build

The @nxlv/python:build command handles the sdist and wheel build generation. When the project has local dependencies the executor copies the package/dependencies recursively.

Options

Locked Versions Build

Using the default (lockedVersions and bundleLocalDependencies) options, the executor uses the locked versions across all the dependencies and bundles the local dependencies in the same wheel file.

Poetry Example

packages/proj1/pyproject.toml

[tool.poetry]
name = "pymonorepo-proj1"

  [[tool.poetry.packages]]
  include = "pymonorepo_proj1"

  [tool.poetry.dependencies]
  python = ">=3.8,<3.10"
  pendulum = "^2.1.2"

    [tool.poetry.dependencies.pymonorepo-lib1]
    path = "../lib1"
    develop = true

packages/lib1/pyproject.toml

[tool.poetry]
name = "pymonorepo-lib1"
version = "1.0.0"

  [[tool.poetry.packages]]
  include = "pymonorepo_lib1"

  [tool.poetry.dependencies]
  python = ">=3.8,<3.10"
  numpy = "^1.24.1"

When the build is executed in the proj1 package, the dist tar/whl file will contain the lib1 package and all dependencies in the poetry.lock file.

packages/proj1/dist/pymonorepo-proj1-1.0.0.tar.gz/pyproject.toml

[tool.poetry]
name = "pymonorepo-proj1"
version = "1.0.0"

  [[tool.poetry.packages]]
  include = "pymonorepo_proj1"

  [[tool.poetry.packages]]
  include = "pymonorepo_lib1"

  [tool.poetry.dependencies]
  python = ">=3.8,<3.10"

    [tool.poetry.dependencies.numpy]
    version = "1.24.1 "
    markers = 'python_version >= "3.8" and python_version < "3.10"'
    optional = false

    [tool.poetry.dependencies.pendulum]
    version = "2.1.2 "
    markers = 'python_version >= "3.8" and python_version < "3.10"'
    optional = false

    [tool.poetry.dependencies.python-dateutil]
    version = "2.8.2 "
    markers = 'python_version >= "3.8" and python_version < "3.10"'
    optional = false

    [tool.poetry.dependencies.pytzdata]
    version = "2020.1 "
    markers = 'python_version >= "3.8" and python_version < "3.10"'
    optional = false

    [tool.poetry.dependencies.six]
    version = "1.16.0 "
    markers = 'python_version >= "3.8" and python_version < "3.10"'
    optional = false

Note, that python-dateutil is a dependency of pendulum, and the pymonorepo_lib1 is now part of the project instead of a dependency.

Non-Locked Versions Build

Using the --lockedVersions=false option, the executor uses the versions from the pyproject.toml file across all the dependencies and bundles the local dependencies in the same wheel file.

Poetry Example

packages/proj1/dist/pymonorepo-proj1-1.0.0.tar.gz/pyproject.toml

[tool.poetry]
name = "pymonorepo-proj1"
version = "1.0.0"

  [[tool.poetry.packages]]
  include = "pymonorepo_proj1"

  [[tool.poetry.packages]]
  include = "pymonorepo_lib1"

  [tool.poetry.dependencies]
  python = ">=3.8,<3.10"
  numpy = "^1.24.1"
  pendulum = "^2.1.2"

Note, the pymonorepo_lib1 still bundled in the project but the dependencies are listed in the same way it is on the pyproject (using ^).

Non-Bundled Local Dependencies Build

Using the --bundleLocalDependencies=false and --lockedVersions=false options, the executor checks if the local dependency is publishable and uses the version from the pyproject.toml file, instead of bundling the package.

packages/proj1/dist/pymonorepo-proj1-1.0.0.tar.gz/pyproject.toml

[tool.poetry]
name = "pymonorepo-proj1"
version = "1.0.0"

  [[tool.poetry.packages]]
  include = "pymonorepo_proj1"

  [tool.poetry.dependencies]
  python = ">=3.8,<3.10"
  pendulum = "^2.1.2"
  pymonorepo-lib1 = "1.0.0"

To identify if the package is publishable, the executor checks project.json file, property targets.build.options.publish.

If the publish option is set to false and the --bundleLocalDependencies=false option is used, the executor will bundle the package.

Custom source specification

In addition when adding dependencies in this way its also possible to configure a custom source for a package. This works similar to the publish option in that its specified on the target dependencies build options. To use this set the customSourceName and customSourceUrl to valid values for the source to retrieve the package from for each package stored on a custom Pypi.

project.json example:

{
  ...
  "targets": {
    ...
    "build": {
      "executor": "@nxlv/python:build",
      "outputs": ["apps/myapp/dist"],
      "options": {
        "outputPath": "apps/myapp/dist",
        "publish": false,
        "customSourceName": "example",
        "customSourceUrl": "http://example.com/"
      }
    },
  }
}

Alternatively its also possible to configured it within the nx.json as targetDefaults across the whole repository.

flake8

The @nxlv/python:flake8 handles the flake8 linting tasks and reporting generator.

Options

install

The @nxlv/python:install handles the install command for a project.

• poetry: poetry install {args}
• uv: uv install {args}

Options

sync

The @nxlv/python:sync executor handles sync the locked dependencies with the python virtual environment.

• poetry:
  • v1: poetry install --sync
  • v2: poetry sync
• uv: uv sync

Options

lock

The @nxlv/python:lock executor handles the lock command for a project.

• poetry:
  • With default options:
    • v1: poetry lock --no-update
    • v2: poetry lock
  • With update option:
    • v1: poetry lock
    • v2: poetry lock --regenerate
• uv:
  • With default options:
    • uv lock
  • With update option:
    • uv lock --upgrade

Options

publish

The @nxlv/python:publish executor handles the publish command for a project.

• poetry: poetry publish {args}
• uv: uv publish {args}

Options

This executor first executes the build target to generate the tar/whl files and uses the --keepBuildFolder flag to keep the build folder after the build process.

For must scenarios, running the poetry/uv publish with @nxlv/python:run-commands executor is enough, however, when the project has local dependencies and the --bundleLocalDependencies=false option is used, the default poetry/uv publish command doesn't work properly, because the poetry/uv publish command uses the current pyproject.toml file, which doesn't have the local dependencies resolved, the @nxlv/python:publish executor solves this issue by running the poetry publish command inside the temporary build folder generated by the @nxlv/python:build executor, so, the pyproject.toml file has all the dependencies resolved.

run-commands (same as nx:run-commands)

The @nxlv/python:run-commands wraps the nx:run-commands default Nx executor and activates the virtual environment before running the command.

POETRY NOTE: This executor only changes the default nx:run-commands if the workspace is configured to use the Shared virtual environment mode and the autoActivate option is set to true in the root pyproject.toml file. POETRY NOTE: The autoActivate option is set to false by default.

POETRY ONLY:

root pyproject.toml

...
[tool.nx]
autoActivate = true

...

The options and behavior are the same as the nx:run-commands executor.

See the Nx documentation for more information

sls-deploy

The @nxlv/python:sls-deploy executor handles npx sls deploy command for serverless framework projects.

This executor uses the @nxlv/python:build artifacts to generate a requirements.txt and to be used with serverless-python-requirements plugin.

Serverless YAML example:

service: myservice

plugins:
  - serverless-python-requirements

custom:
  pythonRequirements:
    usePoetry: false

The property usePoetry must be false, so, the serverless-python-requirements uses the requirements.txt file generated by this executor, this is required when the project has more than 2 levels of local dependencies.

Example:

- root:
  - sls-app
   - local-lib1
    - local-lib2

Using the native serverless-python-requirements plugin with poetry the 2 levels of local dependencies are not supported.

project.json example:

{
  "$schema": "../../node_modules/nx/schemas/project-schema.json",
  "projectType": "application",
  "sourceRoot": "apps/myapp/lambda_functions",
  "targets": {
    "deploy": {
      "executor": "@nxlv/python:sls-deploy",
      "dependsOn": ["build"],
      "options": {}
    },
    "package": {
      "executor": "@nxlv/python:sls-package",
      "dependsOn": ["build"],
      "options": {}
    },
    ...
    "build": {
      "executor": "@nxlv/python:build",
      "outputs": ["apps/myapp/dist"],
      "options": {
        "outputPath": "apps/myapp/dist",
        "publish": false
      }
    },
  }
}

Options

Releases

This plugin supports the Nx releases feature.

If you are already using the @nxlv/python plugin and want to enable the releases feature, please run the following command:

nx generate @nxlv/python:enable-releases

Release Lock arguments

The @nxlv/python:release-version generator has a lockArgs option that allows you to pass additional arguments to the lock command.

To use it, you need to update the nx.json file to add the lockArgs option in the release.version.generatorOptions object.

Example:

{
  ...
  "release": {
    ...
    "version": {
      "generatorOptions": {
        ...
        "lockArgs": "--index-strategy unsafe-best-match"
      }
    }
  }
}