Skip to content

Files

Latest commit

 

History

History
170 lines (118 loc) · 4.67 KB

CONTRIBUTING.md

File metadata and controls

170 lines (118 loc) · 4.67 KB

Developers Guide

Thank you for your interest in contributing to the Optuna Dashboard project! This document will provide you with an overview of the repository structure and instructions on how to build optuna-dashboard.

Repository Structure

The repository is organized as follows:

.
├── optuna_dashboard/         # The Python package.
│   └── ts/                   # TypeScript code for the Python package.
│      ├── index.tsx          # Entry point for the Python package.
│      └── pkg_index.tsx      # Entry point for Jupyter Lab extension, output placed under `optuna_dashboard/pkg/`.
├── standalone_app/           # Standalone application that can be run in browser or within the WebView of the VS Code extension.
│   ├── browser_app_entry.tsx # Entry point for browser app, hosted on GitHub pages.
│   └── vscode_entry.tsx      # Entry point for VS Code extension, output placed under `vscode/assets`.
├── vscode/                   # The VS Code extension.
├── jupyterlab/               # The Jupyter Lab extension.
├── rustlib/                  # Rust library exporting Wasm functions.
│   └── pkg/                  # Output directory for rustlib, installed from package.json via `"./rustlib/pkg"`.
└── tslib/                    # TypeScript library shared for common use.
    ├── react/                # Common React components.
    ├── storage/              # Common code for handling storage.
    └── types/                # Common TypeScript types.

Python package

Building TypeScript packages

$ make tslib

Compiling TypeScript files

Node.js v16 is required to compile TypeScript files.

$ cd optuna_dashboard/
$ npm install
$ npm run build:dev
Watch for files changes
$ cd optuna_dashboard/
$ npm run watch
Production builds
$ cd optuna_dashboard/
$ npm run build:prd

Building a Docker image

$ docker build -t optuna-dashboard .

When failed above command due to the out of heap memory error (Exit code: 137), please check "Resources" tab on your Docker engine's preference since it requires a lot of memory to compile TypeScript files. You can use the Docker image like below:

# SQLite3
$ docker run -it --rm -p 8080:8080 -v `pwd`:/app -w /app optuna-dashboard sqlite:///db.sqlite3

Running dashboard server

$ pip install -e .
$ OPTUNA_DASHBOARD_DEBUG=1 optuna-dashboard sqlite:///db.sqlite3

Note that OPTUNA_DASHBOARD_DEBUG=1 makes the server will automatically restart when the source codes are changed.

Running tests, lint checks and formatters

Running unit tests for tslib/

$ make tslib-test

Running unit tests for python_tests/

$ pytest python_tests/

Running e2e tests using pytest playwright

$ pip install -r requirements.txt
$ playwright install
$ pytest e2e_tests

If you want to create a screenshot for each test, please run a following command, then check screenshots in tmp/ directory.

$ pytest e2e_tests --screenshot on --output tmp

If you want to generate a locator in each webpage, please use the playwright codegen. See this page for more details.

For more detail options, you can check this page.

Linters (flake8, black and mypy)

$ pip install -r requirements.txt
$ flake8
$ black --check .
$ isort . --check
$ mypy optuna_dashboard python_tests

Auto-formatting Python and TypeScript files

$ make fmt

Release the new version

The release process(compiling TypeScript files, packaging Python distributions and uploading to PyPI) is fully automated by GitHub Actions.

  1. Replace optuna_dashboard.__version__ to the next version (e.g. 0.8.0).
  2. Create a git tag (e.g. v0.8.0) and push it to GitHub. If succeeded, GitHub Action will build sdist/wheel packages and create a draft GitHub release.
  3. Edit a GitHub release, generate release note, write highlights of this release if needed, and mark "Create a discussion for this release" checkbox. Then make it publish. GitHub Action will release the new version to PyPI.

Standalone Single-page Application

Please install wasm-pack and execute the following command.

$ make serve-browser-app

Open http://localhost:5173/

VS Code Extension

$ npm i -g vsce
$ make vscode-extension

Jupyter Lab Extension

$ make jupyterlab-extension