Awesome

The best-of-generator is a CLI tool to generate a markdown page of ranked open-source projects based on a list of projects defined in a yaml file. It is integrated with different package managers - such as PyPI, NPM, Conda, and Docker Hub - to automatically collect a variety of project metadata and calculate project-quality scores. It also comes with a GitHub Action workflow for a fully automized update process.

🧙‍♂️ Create your own best-of list in just 3 minutes with this guide.

Highlights

• 📇  Generates a beautiful markdown page from a yaml list.
• 🔌  Integrates various package managers (npm, pypi, conda ...).
• 🥇  Calculates a project-quality score based on a variety of metrics.
• 📈  Identifies trending projects based on collected metrics.
• 🔄  GitHub Action workflow for automated weekly updates.

Getting Started

🧙‍♂️ If you want to create your own best-of list, we strongly recommend to follow this guide instead of setting up best-of manually. With the guide, it will only take about 3 minutes to get you started. It is already set-up to automatically run the best-of generator via our GitHub Action and includes other useful template files. Installing the best-of CLI tool is not required.

1. Install best-of generator via pip:

   pip install best-of
   

2. Create a projects.yaml file based on the documented structure. This file should contain at least one project. For example:

   projects:
      - name: "best-of-ml-python"
        github_id: "ml-tooling/best-of-ml-python"
   

3. Run best-of generator via command-line:

   best-of generate -g <GITHUB_API_TOKEN> ./projects.yaml
   

You can find further information on how to configure the projects.yaml file and additional features in the documentation section below.

Support & Feedback

This project is maintained by Benjamin Räthlein, Lukas Masuch, Jan Kalkan, and Johannes Rieke. Please understand that we won't be able to provide individual support via email. We also believe that help is much more valuable if it's shared publicly so that more people can benefit from it.

Documentation

The best-of generator is a CLI tool to generate a markdown page from a list of projects configured in a yaml file. The documentation sections below will provide information on the projects.yaml structure, on its different sections (projects, labels, categories & configuration), on some of the best-of features (e.g. project-quality score & trending projects), and instructions on how to run the markdown generation via the command-line interface or via GitHub Actions.

projects.yaml Structure

The projects.yaml file has the following structure:

• configurations (optional): Can be used to overwrite the default configuration of the best-of list. More information in the configuration section.
• categories (required): All used categories should be listed here with at least a descriptive title. More information in the categories section.
• labels (optional): Used labels can be added here to extend the label with additional aspects (e.g. URL, image, description). More information in the labels section.
• projects (required): All projects that are supposed to be shown in the generated markdown page should be listed here. More information in the projects section.

The following yaml shows a small example:

# Optional: change the default configuration
configuration:
    markdown_header_file: "config/header.md"
    markdown_footer_file: "config/footer.md"

# Optional: add categories
categories:
  - category: "data-engineering"
    title: "Machine Learning & Data Engineering"
    subtitle: "Best-of lists about machine learning, data engineering, data science, or other topics related to big data."

# Optional: add labels
labels:
  - label: "python"
    image: "https://www.python.org/static/favicon.ico"
    description: "Best-of list with Python projects"

# Required: list of all projects
projects:
  - name: "best-of-ml-python"
    github_id: "ml-tooling/best-of-ml-python"
    labels: ["python"]
    category: "data-engineering"

Projects

A project is the main component of a best-of list. In most cases, a project is hosted on GitHub and released on different package managers. Such a project should be added with the github_id and the IDs of all the package managers it is released to. However, it is also possible to add projects which are not hosted on GitHub or released on a package manager, as shown in the example below.

Project Examples

projects:
  # Projects with different package managers:
  - name: "Tensorflow"
    github_id: "tensorflow/tensorflow"
    pypi_id: "tensorflow"
    conda_id: "conda-forge/tensorflow"
    dockerhub_id: "tensorflow/tensorflow"
  - name: "Best-of Generator"
    pypi_id: "best-of"
    github_id: "best-of-lists/best-of-generator"
  # Link to another project collection:
  - name: "Best-of Overview"
    homepage: "https://best-of.org"
    resource: True
  # Project that is not on GitHub:
  - name: "Quart"
    pypi_id: "quart"
    homepage: "https://gitlab.com/pgjones/quart"
    description: "Quart is a Python ASGI web microframework with the same API as Flask."
    license: "MIT"
    star_count: 772
    show: True

The example above will be rendered as shown below:

Every project can also be expanded to show additional project information (by clicking on the project), for example:

Project Properties

While you can theoretically overwrite all project metadata, we suggest to only set the properties which the best-of generator is not able to find on GitHub or the configured package managers. There are also other undocumented properties, but for most projects those properties should not be overwritten.

• created_at
• update_at
• github_url
• github_release_downloads
• github_dependent_project_count
• last_commit_pushed_at
• star_count
• commit_count
• dependent_project_count
• contributor_count
• fork_count
• monthly_downloads
• open_issue_count
• closed_issue_count
• release_count
• latest_stable_release_published_at
• latest_stable_release_number
• trending
• helm_id
• brew_id
• apt_id
• yum_id
• snap_id
• maven_id
• dnf_id
• yay_id
• <PACKAGE_MANAGER>_url
• <PACKAGE_MANAGER>_latest_release_published_at
• <PACKAGE_MANAGER>_dependent_project_count

Categories

A category allows to add additional structure to the best-of list by grouping related projects into a shared category. Thereby, every project is grouped into exactly one category. If no category is provided with the project metadata, the project will be categorized into Others.

Category Example

categories:
  - category: "data-engineering"
    title: "Machine Learning & Data Engineering"
    subtitle: "Best-of lists about machine learning, data engineering, data science, or other topics related to big data."

projects:
  - name: "best-of-ml-python"
    github_id: "ml-tooling/best-of-ml-python"
    category: "data-engineering"

The example above will be rendered as shown below:

Category Properties

Labels

A label allows to highlight similarities or special features shared between projects. Compared to categories, a project can have any number of labels. The labels are shown as badges attached to the project description. It can have only an image (favicons are recommended), only a name, or both. We recommend to use image labels (or only very short labels) since the usage of labels will shorten the visible description text of a project.

Label Example

labels:
  - label: "python"
    image: "https://www.python.org/static/favicon.ico"
    description: "Best-of list with Python projects"
  - label: "libraries"
    name: "libraries"

projects:
  - name: "best-of-ml-python"
    github_id: "ml-tooling/best-of-ml-python"
    labels: ["libraries", "python"]
    category: "data-engineering"

The example above will be rendered as shown below:

Label Properties

Configuration

Many aspects of the best-of list can be configured. Since most default values are selected to support the widest range of different lists, changing the default configuration is not required for most cases.

Configuration Example

configuration:
  min_stars: 0
  min_projectrank: 0
  allowed_licenses: ["all"]
  markdown_header_file: "config/header.md"
  markdown_footer_file: "config/footer.md"

The configuration example above changes the default configuration to show all projects regardless of star count (via min_stars), projectrank (via min_projectrank), or license (via allows_licenses). It also configures a header (via markdown_header_file) and footer (via markdown_footer_file) markdown files that will be attached to the generated content.

Configuration Options

Project Quality Score

All projects in a best-of list are ranked and sorted by a project-quality score (also called projectrank). The score is calculated based on various metrics automatically collected from GitHub and different package managers. The score is just a sum of points which a project collects for various aspects and metrics. The score only has a meaning when it is compared to the project-quality score of other projects. We currently use the following aspects to calculate the score:

This calculation is just chosen by experience. There is no scientific proof that this really reflects the quality of a project.

• Has homepage link & description: + 1
• Has an existing GitHub repository: + 1
• Has a license: + 1
• Has a commonly used license (e.g. MIT): + 1
• Has multiple releases: + 1
• Has stable releases based on semantic version: + 1
• Has a release that is less than 6 months old: + 1
• Repo was update in the last 3 months: + 1
• Is older than 6 months: + 1
• Metrics from GitHub & package mangers:
  • Number of stars: + log(COUNT / 2)
  • Number of contributors: + log(COUNT / 2) - 1
  • Number of commits: + log(COUNT / 2) - 1
  • Number of forks: + log(COUNT / 2)
  • Number of monthly downloads: + log(COUNT / 2) - 1
  • Number of dependent projects: + log(COUNT / 1.5)
  • Number of watchers: + log(COUNT / 2) - 1
  • Number of closed issues: + log(COUNT / 2) - 1

Trending Projects

The best-of list is able to automatically identify trending projects by comparing project-quality scores between the metadata of the current generation with the latest history file. If the history is activated (projects_history_folder is not set to null), the best-of generation will automatically create a <YYYY-MM-dd>_changes.md file in the configured history folder for every update and a latest-changes.md file in the folder of the generated markdown page. These files contain a list of projects that are trending up (higher quality score since last update) and down (lower quality score since last update) as well as a list of all added projects since the last update, as shown in the following example:

The GitHub Action workflow uses these markdown files to automatically create releases for every update. This allows to persist a useful changelog over many updates and enables readers to get valuable email updates whenever the list is updated (by watching for release events).

Generation via CLI

To use the CLI, you need to have the best-of generator installed via pip: pip install best-of

best-of generate [OPTIONS] PATH

Generates a best-of markdown page from a yaml file.

Arguments:

• PATH: Path to the yaml file containing the best-of metadata (e.g. ./projects.yaml).

Options:

• -g, --github-key TEXT: GitHub API Token (from https://github.com/settings/tokens).
• -l, --libraries-key TEXT: Libraries.io API Key (from https://libraries.io/api).
• --help: Show this message and exit.

Generation via GitHub Action

🧙‍♂️ If you want to create your own best-of list, we strongly recommend to follow this guide. With the guide, it will only take about 3 minutes to get you started. It already includes this GitHub Action and some other useful template files. Further manual steps for setting up the GitHub Action are not required.

The best-of-update-action makes it very easy to set-up automated scheduled updates for your best-of markdown page. Please refer to the best-of-update-action documentation for more detailed information about the GitHub Action and the workflow.

Generation via Python API

Usage of the Python API is not well documented yet and currently not recommended.

The best-of generator can also be used and integrated via its Python API. The full Python API documentation can be found here.

Updating Best-of Generator

Known Issues

GitHub only renders the first 512 kb of the main README.md file and will cut off the rendered version as soon as it has processed the first 512 kb of the raw markdown content. The rendering is only cut off when viewing the readme on the main repo page. If you directly select the README.md file, it will render in its entirety. To mitigate this issue, we optimized the markdown generation to require the minimum amount of characters. However, if you have a very large list of projects (more than 800), you might reach the 512 kb limit (check the file size of the generated README.md file). In this case, we suggest to extract some of the categories or projects into smaller best-of lists.

Contribution

• Pull requests are encouraged and always welcome. Read our contribution guidelines and check out help-wanted issues.
• Submit GitHub issues for any feature request and enhancement, bugs, or documentation problems.
• By participating in this project, you agree to abide by its Code of Conduct.
• The development section below contains information on how to build and test the project after you have implemented some changes.

Development

Requirements: Docker and Act are required to be installed on your machine to execute the containerized build process.

To simplify the process of building this project from scratch, we provide build-scripts - based on universal-build - that run all necessary steps (build, check, test, and release) within a containerized environment. To build and test your changes, execute the following command in the project root folder:

act -b -j build

Refer to our contribution guides for more detailed information on our build scripts and development process.

Licensed MIT. Created and maintained with ❤️  by developers from Berlin.