3 releases

0.1.2 Mar 21, 2024
0.1.1 Dec 1, 2023
0.1.0 Aug 17, 2023

#1727 in Web programming

Download history 17/week @ 2024-02-19 26/week @ 2024-02-26 4/week @ 2024-03-04 16/week @ 2024-03-11 144/week @ 2024-03-18 19/week @ 2024-03-25 64/week @ 2024-04-01

244 downloads per month
Used in 2 crates (via html-compare)

MIT license

14KB
372 lines

MRML

Crates.io Crates.io FOSSA Status

.github/workflows/main.yml codecov

Average time to resolve an issue Percentage of issues still open Maintainability

Introduction

This project is a reimplementation of the nice MJML markup language in Rust.

How to use it in my code

Update your cargo.toml:

[dependencies]
mrml = "3"
serde = { version = "1.0", features = ["derive"] }

Create your main.rs:

use mrml::prelude::parser::http_loader::{HttpIncludeLoader, BlockingReqwestFetcher};
use mrml::prelude::parser::ParserOptions;
use mrml::prelude::render::RenderOptions;
use std::collections::HashSet;


fn main() {
  let resolver = HttpIncludeLoader::<BlockingReqwestFetcher>::new_allow(HashSet::from(["http://localhost".to_string()]));
  let parser_options = ParserOptions {
      include_loader: Box::new(resolver),
  };
  let render_options = RenderOptions::default();
  let template = r#"<mjml>
  <mj-body>
    <mj-include path="http://localhost/partials/mj-body.mjml" />
  </mj-body>
</mjml>"#;
  match mrml::parse_with_options(template, &parser_options) {
      Ok(mjml) => match mjml.render(&render_options) {
        Ok(html) => println!("{html}"),
        Err(err) => eprintln!("Couldn't render template: {err:?}"),
      },
      Err(err) => eprintln!("Couldn't parse template: {err:?}"),
  }
}

It's also possible to use an async include loader

use mrml::mj_include::body::MjIncludeBodyKind;
use mrml::prelude::parser::http_loader::{AsyncReqwestFetcher, HttpIncludeLoader};
use mrml::prelude::parser::local_loader::LocalIncludeLoader;
use mrml::prelude::parser::memory_loader::MemoryIncludeLoader;
use mrml::prelude::parser::multi_loader::{MultiIncludeLoader, MultiIncludeLoaderItem, MultiIncludeLoaderFilter};
use mrml::prelude::parser::noop_loader::NoopIncludeLoader;
use mrml::prelude::parser::loader::AsyncIncludeLoader;
use mrml::prelude::parser::AsyncParserOptions;
use mrml::prelude::render::RenderOptions;

#[tokio::main]
async fn main() {
  let resolver = MultiIncludeLoader::<Box<dyn AsyncIncludeLoader + Send + Sync + 'static>>::new()
      .with_starts_with("file://", Box::new(LocalIncludeLoader::new(PathBuf::default().join("resources").join("compare").join("success"))))
      .with_starts_with("https://", Box::new(HttpIncludeLoader::<AsyncReqwestFetcher>::allow_all()))
      .with_any(Box::<NoopIncludeLoader>::default());
  let parser_options = AsyncParserOptions {
      include_loader: Box::new(resolver),
  };
  let render_options = RenderOptions::default();
  let json = r#"<mjml>
  <mj-body>
    <mj-include path="file://basic.mjml" />
  </mj-body>
</mjml>"#;
  match mrml::async_parse_with_options(json, std::sync::Arc::new(parser_options)).await {
      Ok(mjml) => match mjml.render(&render_options) {
        Ok(html) => println!("{html}"),
        Err(err) => eprintln!("Couldn't render template: {err:?}"),
      },
      Err(err) => eprintln!("Couldn't parse template: {err:?}"),
  }
}

Why?

  • A Node.js server rendering an MJML template takes around 20 MB of RAM at startup and 130 MB under stress test. In Rust, less than 1.7 MB at startup and a bit less that 3 MB under stress test.
  • The JavaScript implementation cannot be run in the browser; the Rust one (and WebAssembly one) can be.

You want to contribute?

Feel free to read our contributing section and the code of conduct.

Performance

With the same Linux amd64 machine, to render the amario template using hyperfine (see the script in the benchmarks folder).

Benchmark 1: mjml /amario.mjml
  Time (mean ± σ):     634.1 ms ±   5.2 ms    [User: 669.3 ms, System: 168.2 ms]
  Range (min … max):   625.8 ms … 642.3 ms    10 runs

Benchmark 2: /usr/bin/mrml /amario.mjml render
  Time (mean ± σ):       5.6 ms ±   0.1 ms    [User: 2.8 ms, System: 2.9 ms]
  Range (min … max):     5.5 ms …   7.1 ms    494 runs

Summary
  /usr/bin/mrml /amario.mjml render ran
  112.83 ± 2.12 times faster than mjml /amario.mjml

From this, you can see that mrml is more than 110 faster than mjml.

Missing implementations

  • mj-style[inline]: not yet implemented. It requires parsing the generated html to apply the inline styles afterward (that's how it's done in mjml) which would kill the performances. Applying it at render time would improve the performance but it would still require to parse the CSS.

Who is using MRML?

If you are using MRML and want to be added to this list, don't hesitate to create an issue or open a pull request.

What is using MRML?

mjml_nif - Elixir library

mrml-ruby - Ruby library

mjml-python - Python library

If you are using MRML and want to be added to this list, don't hesitate to create an issue or open a pull request.

License

FOSSA Status

Dependencies

~11MB
~150K SLoC