Streaming Templates in node and the browser

I’m currently building a simple web app and I needed a simple templating engine that can stream dynamically generated responses to the network. 

It’s hard to build sites that are fast. One known pattern for building sites that render quickly is to ensure that the browser gets the HTML as quickly as possible. Yet, many of the tools (middleware and templating engines) that developers use to build sites wait for the response from the server application to be completely created before they are put on to the wire and send to the client.

Streaming template engines are important because they don’t wait until the entire response is ready before sending the data on, but they are also able to wait for gathered data (say from a database request) before proceeding with the rest of the response.

It feels like there are scant few streaming templating engines in Node. There are fewer that work in Node, in the Browser and in a service worker.

I came across the awesome flora-tmpl project. It’s a streaming templating engine that uses template literals html`Hello ${world}`; to make it easy to author streaming templates, and if you just need them for Node JS, then you should totally use it.

The project I am building needs to work in Node, the Browser and a Service Worker (I am running my app in both Node and the SW) and whilst it would have been possible to browserify the library, it makes it a lot larger than I need it to be. I know enough about ReadableStreams in the browser to damage, so I wanted to see if I could port flora-tmpl to use the WhatWG streams API…

whatwg-flora-tmpl (I might have to change the name, it’s not affiliated with the whatwg or flora really) is a small library that does pretty much everything flora does, but using the WhatWG Streams API in the browser (and a polyfill in Node).

The template literal function takes your, well, template and returns a ReadableStream, the template function will then push string literals on the stream, and when it needs to evaluate a variable, it will do that and then enqueue that on to the stream too. This means, for example that you can generate responses in a service worker fetch event like: new Response(tmpl`Hello ${world}`);

It’s not just basic values that can be evaluated (such as strings, numbers and arrays), it can also evaluate ReadableStreams, which means you stream templates in your template.

The demo for the video at the start of this article is below (you wouldn’t do this in real life, it’s just to show the templating engine can also embed readable streams which means it can embed the templating engine and stream that response too)

import tmpl from '../lib/index.js';
import streams from "web-streams-polyfill";

const read = (stream) => {
  const reader = stream.getReader();
  const decoder = new TextDecoder();

  return process(result) {
    if (result.done) {


const encoder = new TextEncoder();
const title = 'Awesome';

${new streams.ReadableStream({
  start(controller) {
    let counter = 0;
    const interval = setInterval(async () => { 
      if (counter >= 10) {
    }, 1000);

A better (albeit a lot more complex) example is the one I am using in my project.

const head = (data, bodyTemplate) => template`<!DOCTYPE html>
    <title>Baby Logger</title>
    <script src="/client.js" type="module" defer></script>
    <link rel="stylesheet" href="/styles/main.css">
    <link rel="manifest" href="/manifest.json">
    <meta name="viewport" content="width=device-width">

const body = (data, items) => template`<header>
    <h1>Baby Log</h1>
    <div><a href="/">All</a>, <a href="/feeds">Feeds</a>, <a href="/sleeps">Sleeps</a>, <a href="/poops">Poops</a>,  <a href="/wees">Wees</a></div>
    <span>Add</span><a href="/feeds/new" title="Add a feed">🍼</a><a href="/sleeps/new" title="Add a Sleep">💤</a><a href="/poops/new" title="Add a Poop">💩</a><a href="/wees/new" title="Add a Wee">⛲️</a>

const aggregate = (data) => {
  // Do a lot of work in IndexedDB and other data munging

new Response(template`${head(data,

We have many templates, head generates the skeleton of the HTML, body is the content of the page based on the output of aggregate. The latter of those functions does a lot of asynchronous work which takes some time, however the response can start rendering pretty much straight away because the head and body functions mostly output text.

I think it’s pretty neat, and hopefully I can write up a lot more about the architecture of the app I am building that uses this across the Server, Client and Service Worker.

Finally, a big shout out to Matthew Phillips who wrote the awesome original version of flora-tmpl.

Picture of me smiling.

Paul Kinlan

I lead the Chrome Developer Relations team at Google.

We want people to have the best experience possible on the web without having to install a native app or produce content in a walled garden.

Our team tries to make it easier for developers to build on the web by supporting every Chrome release, creating great content to support developers on, contributing to MDN, helping to improve browser compatibility, and some of the best developer tools like Lighthouse, Workbox, Squoosh to name just a few.

RSS Github Medium