Itsy-Gitsy

A static website generator for Git repositories.

What

Itsy-Gitsy spiders across a collection of Git repositories, passes a subset of their contents through user-defined input templates, and generates a set of output files. In standard usage, the input templates describe a website layout, and the output is a static website suitable for browsing locally, or hosting on a web server.

See a demo here.

Since the output is static text files, Itsy-Gitsy can generate more than just websites. With custom templates, it can render your Git repositories as plain text, CSV, LaTeX, org-mode, Markdown, TOML, or whatever format you desire.

More generically, Itsy-Gitsy is a utility to generate text-based document trees from Git repositories based on descriptive templates.

Why

The primary motivation is for self-hosting simple Git repository frontends without a dynamic web application. For self-hosters, there can be many motivations for not wanting dynamically generated content: particularly system resource, security, or maintenance requirements on the web server. Most existing Git repository frontends are dynamic "forges" or "hubs", where even the lightweight ones are large, complex applications with significant resource requirements and large attack surfaces.

Side benefits include offline, local browsing without a web host, and generation of non-website content. For instance, Itsy-Gitsy is also suitable for generating Git repositories as Gopher or Gemini sites, or even e-mail newsletters.

It is written in Rust and compiles in a single native executable, making it possible to (cross-)compile in a development environment, and run on a low-resource server with few dependencies.

How

Behavior is dictated by a single TOML configuration file, which specifies the Git repositories to parse and various global and per-repository settings. Itsy-Gitsy has a high degree of configurability, which allows for great flexibility of the output format.

Itsy-Gitsy uses git2, which in turn uses libgit2, to parse a list of local Git repositories into an internal representation. Remote repositories available over non-authenticated HTTPS connections can also be specified, in which case they are cloned locally first.

The internal representation of repositories are passed through a configurable set of templates using the Tera template engine, which is itself based on the Jijna2 templating language. After template substitution, the generated documents are written to user-configurable file paths.

If a file content template is provided, the contents of files stored in the repository can also be rendered. These can optionally be rendered into HTML from Markdown with pulldown-cmark, and optionally be rendered as syntax highlighted HTML with syntect.

Itsy-Gitsy currently only supports indexing a single branch for each repository. The branch that is to be indexed is configurable.

Project Status

Itsy-Gitsy is young and experimental. It is under-tested and missing probably-important features. Git supports extremely many combinations, and Itsy-Gitsy has only been tested with a tiny, tiny fraction of them.

The template "API" is not stable, and shouldn't be expected to be before a version 1.0.

Feedback, bug reports, and feature requests are welcome.

Features

• Generate static, templated, multi-page output from Git repositories
• Index any number of Git repositories
  • local directories (bare or full)
  • remote repositories (only unauthenticated HTTPS)
• Configurable name, path, branch, description, website, etc.
• Site-wide and per-repository settings
• Multiple supported templates
  • List of all repositories
  • Summary of each repository
  • List of historical commits, branches, tags
  • Per-commit, per-tag, per-branch
  • Per file, with content
    • optional Markdown rendering
    • optional Syntax highlighting
  • Error page
• Configurable output
  • configurable file names, with variable substitution
  • configurable directory structure
  • unlimited template-to-output mappings
• Paginated output
• Configurable limits for RAM and disk space usage
• Site-wide and per-repository asset files
• Basic "permalink" support

Getting Started

You can quickly try out Itsy-Gitsy by using the default configuration to render and open a locally browseable copy of a remote repository:

$ git clone https://github.com/mrmekon/itsy-gitsy
$ cd itsy-gitsy
$ cargo build --release
$ ./target/release/itsy-gitsy --clean --local --open --repo "https://github.com/mrmekon/itsy-gitsy"

Next, you will want to edit config.toml and point it to the Git repositories you want to index. The default configuration file (config.toml) includes documentation to help you get started.

Now you can regenerate and view your repositories locally with:

$ ./target/release/itsy-gitsy --clean --local --open

If and when you are ready to move the site to a web server, regenerate it with:

$ ./target/release/itsy-gitsy --clean

Command-line Arguments

Configuration

Itsy-Gitsy is configured via a single TOML configuration file. By default, it looks for a file named config.toml in the working directory, but you can point it to another file with the --config command-line option.

See the included config.toml for full documentation of the available settings.

The top of the file contains global, site-wide settings like the site name, description, base URL, pagination rules, and memory limits. Here you can also specify directories that contain Git repositories in subdirectories, which is used for bulk-import of many repositories.

This should be followed by the [gitsy_outputs] section, which defines which input templates to use, and which files to output. Input templates that are not specified will not be generated, so you can disable any output types that you don't need. Templates can be used as many times as desired, generating arbitrarily many outputs.

A few special variables can be used in output filenames:

An optional [gitsy_extra] section can be used to provide global, user-defined key/value pairs to all of the templates. Use this if you want to add custom site-wide variables for use in your templates.

Finally, zero or more sections with arbitrary names define individual Git repositories to index. Here, you can override most of the global settings at a per-repository level. This is more powerful and allows specifying more metadata than bulk-import.

Templates

Templates are defined using the Tera template engine. Tera is a powerful templating language which allows variables, conditionals, loops, filtering, includes, hierarchical inheritance, and more. Read its official documentation for details.

Itsy-Gitsy has a predetermined set of known template types, designed with a multi-page static Git browser in mind. A default set of templates are provided, which demonstrate the generation of a full, multi-page Git repository browser. You are very much encouraged to write your own, or modify the provided ones to suit your needs.

The templates absolutely do not need to be used for their "intended" purposes. Modify their meaning to suit your needs! You could, for instance, ignore the intended meaning of repo_list and instead use that template to generate an e-mail newsletter with the most recent 10 commits messages from all indexed repositories.

Any templates that are not specified in the configuration file are not evaluated, and their matching output files are not generated. Use this to disable any features your site does not require.

Filters

Tera templates support custom functions and filters, and Itsy-Gitsy defines a few for convenience:

url_string can be used in conjunction with the %PATH% and %REF% filename variables. Both use a very primitive form of "slugifying" the strings into a format that can be used in a URL. This allows for basic permalinks.

Security

Security is, for the most part, outsourced to the libraries Itsy-Gitsy depends on. git2 handles the security of Git repo access, Tera and the templates files themselves handle sanitizing HTML outputs, pulldown-cmark handles sanitizing Markdown output, and syntect handles sanitizing syntax highlighted file contents. If any of these libraries contain security issues, then so does Itsy-Gitsy.

The main thing Itsy-Gitsy itself is responsible for is ensuring it only writes files to its output directory. It has some basic protections against obvious attempts to write outside of the output subdirectory, but nothing invincible.

As always, if security is a concern, best practice is to follow the rules of least-privilege. Run Itsy-Gitsy under a dedicated, low-privilege user account against Git repositories with read-only or no upstream access. For maximum paranoia, disable syntax highlighting and Markdown rendering, and use filesystem namespaces to restrict it to read-only access of Git repositories and read-write access to the output directory.

Performance

High performance is not a primary goal of Itsy-Gitsy, since it is mostly intended for indexing small personal projects, but various settings are provided to allow it to handle large repositories. Most operations take advantage of parallelism. By default, Itsy-Gitsy splits the tasks across all cores, but the number of parallel threads can be reduced in the configuration file.

Syntax highlighting uses syntect's pure-Rust implementation by default, to avoid an extra dependency. This implementation is quite slow, and performance can be greatly improved by using syntect's onig mode, which uses the faster Oniguruma C library for highlighting. This can be enabled at build time with cargo build --features highlight_fast.

All metadata of all repositories, except for file contents, is held in memory. Large repositories can easily exhaust memory, and disk usage can also get quite high. There are several limit_* settings available in the configuration for restricting the amount of data held in memory, with the tradeoff of reducing the amount of data available for the generated output. limit_context and limit_diffs are particularly important restrictions to set on repositories with thousands of commits.

Small repositories with dozens to hundreds of commits can be generated on the order of a few seconds or less. Large repositories take considerably longer and significant amounts of RAM, but aggressively applying limits in the configuration can help; parsing 1,115,000 commits from the Linux kernel repository with syntax highlighting disabled, limit_commit_ids_to_related = true, limit_tree_depth = 3, limit_context = 100 and limit_diffs = 0 took ~7 minutes on a fast laptop, and produced a ~5.7GB website. Peak memory usage was ~8GB.

Other Considerations

The default templates are provided as a starting point, and demonstrate most features. It is fully expected that you will customize or replace them.

The default templates use the %ID% variable for outputting files, directories, branches, and tags. These are guaranteed to be unique and URL-friendly. Links to files and directories, however, are invalidated when changes are made to the repository. To get "permalinks", you can change the output variables to use %PATH% instead, and replace {{file.id}} and {{dir.id}} with {{file.path | url_string}} and {{dir.path | url_string}} in all of the template files.

Limitations

• Only indexes history of one branch.
• High memory usage for large repositories.
• Limited to the pre-defined set of input templates.
• Leaves output in unknown, partial state in case of errors.

Main Dependencies

• git2 -- Rust wrapper for libgit2 library.
• libgit2 -- C library for accessing Git repositories.
• Tera -- Rust templating engine based on Jijna2.
• pulldown-cmark -- Rust library for rendering Markdown to HTML.
• syntect -- Rust library for rendering file content into syntax highlighted HTML.
• onig (optional) -- Rust wrapper for Oniguruma syntax highlighting library.
• oniguruma (optional) -- C library for syntax highlighting.

Similar Software

Git Static Generators

I haven't used these, but they exist. I'm sure they are good, too. Itsy-Gitsy mostly differs in its configurability and focus on user-defined templates.

• git-arr -- Another Git static site generator, in Python.
• stagit -- Another Git static site generator, in C.

Git Forges

Forges typically include project management features, and particularly issue trackers.

• Gitea -- An open-source, self-hostable, dynamic forge.
• Kallithea -- An open-source, self-hostable, dynamic forge.
• GitLab -- An "open-core", self-hostable, dynamic forge.

Git Browsers

• gitweb -- Git's built-in webserver plus dynamic repo browser.
• cgit -- A dynamic repo browser.
• shithub -- A hyper-minimalistic, open-source, self-hostable, dynamic repo browser.