Gilbert WebProducer is a data-driven tool for highly performant production of websites and web applications.
While Gilbert could be thought of as static site generator it differs in that it is:
-
Data-driven: Whereas most static site generators process a tree of page-centric markdown files in the filesystem, Gilbert takes a stream of data and generates content by merging it with in-memory "component" templates. This approach allows Gilbert to be extremely fast and efficient for building content with complex IA needs.
-
Streams based: It makes heavy use of readable, writeable and transformation streams allowing for large volumes of content to be processed at high speed with minimal memory requirements. This also makes Gilbert ideal for serverless deployments using AWS Lambda or Azure Cloud Functions.
Since Gilbert consumes and produces streams its behaviour can be easily adapted by transforming the inbound data with a pipe and transforming the outbound data with a pipe. For instance, data could be streamed from a REST API call to a transformation function that restructures the data schema before piping on to Gilbert.
-
Decoupled architecture: The core of Gilbert is implemented as a reusable NPM module. Since it only accepts and returns streams it can be leveraged for uses cases including:
- Local web development: It can build a 200+ page site with minified HTML, bundled and minified ES modules, and bundled and minified CSS in well under a second in a moderate workstation
- Serverless web publishing: Triggered by a webhook from a headless CMS a Gilbert serverless function can query a data endpoint (REST, GraphQL, etc), rebuild and redeploy pages to a production server in about a second.
-
Lean and Lightweight: Requiring just 6 dependencies Gilbert's at-rest and in-memory footprint is very small.
-
Uses Handlebars Templates: Although Handlebars has been around for a long time it is still significantly faster than <insert bloaty framework du jour here>. And, the Handlebars token replacement syntax means it can be used to generate files of many types including HTML, CSS, JS, text, XML, etc. This makes it easy to generate all files in a website including robots.txt, sitemap.xml, manifest.json, etc.
For a detailed technical explanation please see the Developer Guide.
- Clone this repo
git clone git@github.com:tforster/webproducer.git - CD into the project:
cd webproducer - Install dependencies:
npm i - CD into the examples/sample-site directory:
cd examples/getting-started - Run the CLI:
npx webproducer - Review the output in examples/getting-started/dist
Technically the Gilbert CLI requires no installation as it can be run using npx. However, installing it as a developer dependency will reduce the loading latency.
npm install @tforster/webproducer --save-devnpx webproducer [options]Gilbert can run nicely out-of-the-box against an appropriately structured source tree it also has lots of configuration options for fine tuning. These can be viewed anytime by typing npx webproducer -h which produces the following:
-V, --version output the version number
-r, --relative-root [root] The relative root of all the source files (default: "./src")
-d, --data [data] Path to the JSON data file. (default: "./src/data/data.json")
-t, --theme [theme] Glob to handlebars files (default: "./src/theme/**/*.hbs")
-s, --scripts [scripts] Comma separated list of ES Module entry points (default: "./src/scripts/main.js")
-c, --css [css] Comma separated list of stylesheet entry points (default: "./src/stylesheets/main.css")
-f, --files [files] Comma separated list of static file globs (default: "./src/images/**/*.*")
-o, --out [out] Output directory (default: "./dist")
-x, --prefix-css Enable autoprefixing of CSS (default: false)
--no-scripts Do not process scripts
--no-css Do not process stylesheets
--no-files Do not process static files
--no-uris Do not process uris
-h, --help display help for commandGilbert provides defaults for all paths that follow the Joy naming structure developers are free to override as required to suit their particular directory conventions.
- -r, --relative-root: The relative root of all the source files (default: "./src")
- -d, --data: Path to the JSON data file. (default: "./src/data/data.json")
- -t, --theme: Glob to handlebars files (default: "./src/theme/*/.hbs")
- -s, --scripts: Comma separated list of ES Module entry points -(default: "./src/scripts/main.js")
- -c, --css: Comma separated list of stylesheet entry points -(default: "./src/stylesheets/main.css")
- -f, --files: Comma separated list of static file globs (default: "./src/images/*/.*")
- -o, --out: Output directory (default: "./dist")
-
-x, --prefix-css: Enable vendor prefixes of CSS (default: false). If newer and/or experimental CSS features are used then enabling CSS prefixing with
--prefix-csswill use the data based on current browser popularity and property support to apply prefixes.Note: This is accomplished with PostCSS's Autoprefixer which calls out to the Can I Use website. The extra overhead does increase Gilbert execution time so you may wish to defer this until you are building a release version.
While Gilbert is already very fast, the various --no-* switches can be used to disale specific pipelines for even more speed. For instance, you may be focusing on template development and not actively editing JS and CSS. In which case adding --no-scripts --no-css would disable those pipelines, leaving the last state of generated script and CSS in place in the out directory.
- --no-scripts: Disables the scripts pipeline. No scripts will be parsed or written to the output stream.
- --no-css: Disables the stylesheets pipeline. No stylesheets will be parsed or written to the output stream.
- --no-files: Disables the static files pipeline that copies assets from source. No static files will be parsed or written to the output stream.
- --no-uris: Disables the templates pipeline. No template driven files will be parsed or written to the output stream.
A couple of options to help get the most out of Gilbert.
- -h, --help: Get a summary of all available options at any time
- -V, --version: Output the current version of the Gilbert CLI
WebProducer expects source files to be found in the following tree structure by default, but they can be overridden using various CLI flags.
. (your project root)
└── src
├── data
├── fonts
├── images
├── scripts
├── stylesheets
└── theme
├── common
└── templatesThere are several breaking changes to be aware of when upgrading to v1.0.0.
Pre 1.0.0 saw the URI keys at the top level of the data structure. E.g.
{
"/index": { ... },
"/about": { ... },
...
}For increased legibility these URIs should be moved to the uris property as in:
{
"uris": {
"/index": { ... },
"/about": { ... },
...
},
...
}Note that URI is favoured over URL since the output of Gilbert does not always have to be a web page.
The modelName property used to identify an .hbs template was named so as a direct result of using a DatoCMS GraphQL source in an early implementation of WebProducer. To be more datasource agnostic and reduce the likelihood of property collisions modelName has been changed to webProducerKey.
In addition, the new webProducerKey can accept path separators to align with richer component based front-end architectures.
For example, consider the following data snippet and related themes directory structure:
{
"uris": {
"/admin/reports": {
...
"webProducerKey": "/admin/report/report"
}
}
}. (your project root)
└── src
├── ...
└── pages
└── admin
├── ...
└── report
├── ...
└── report.hbsSee how easy Gilbert's simple streams API is with Use Remote REST API to Create XML files on S3. Or check out the examples directory for others.
Full API details are coming in the Developer Guide.
See CHANGELOG.md for more information.
See CODE_OF_CONDUCT.md for more information.
See CONTRIBUTING.md for more information.
Troy Forster – @tforster – troy.forster@gmail.com
See LICENSE for more information.