The Future Of Frontend Build Tools
About The Author
Frontend Engineer passionate about creating scalable and maintainable applications, spends his free time watching food and hunting shows.
Quick summary ↬
This article explores the concept of tooling for frontend development on the web. You will learn why we need frontend tooling, the various stages of evolution it has gone through, and the new developments that will shape the frontend build tools of the future. To follow along with this article, a general understanding of modern frontend development on the web is necessary.
As a result, users can enjoy more capable and feature-rich applications that remain performant through techniques like code-splitting, caching, prefetching, and other resource optimization techniques — with some applications that are even able to work offline.
Frontend tooling gives us so much today that it is hard to imagine that there was a time when it was not even needed at all. A trip down memory lane could help us understand how we got here.
one file to be served in the browser
As our applications got even more complex, matters like lack of dead code elimination, full rebuilds for small changes, and other performance issues made us realize that we needed something more than just concatenation. That gave rise to the more modern bundlers like Webpack, Parcel, and others.
With the pace of advancement in the frontend space not slowing down, we have started to observe gaps and issues with the modern build tools.
Some of the major limitations include:
The limitations of the mainstream build tools of the day have led to several attempts to reimagine what a frontend build tool should be and do, and there are quite a few new build tools in the wild today.
Closer inspection will reveal that these new tools seem to be taking two major approaches to solving the problem (not necessarily mutually exclusive): a change in paradigm and a change in platform — both powered by new advancements in the web development ecosystem.
More after jump! Continue reading below ↓
“Touch Design for Mobile Interfaces”, our brand-new Smashing Book
on designing for mobile with proven, universal, human-centric guidelines. 400 pages, jam-packed with in-depth user research and guidelines
you can apply immediately. Shipping starts early January 2022.
Jump to table of contents ↬
Meet “Touch Design for Mobile Interfaces”, our brand-new Smashing Book on designing for mobile with proven, universal, human-centric guidelines. 400 pages, jam-packed with in-depth user research and guidelines you can apply immediately. Shipping starts early January 2022.
Languages like Rust and Go have become popular options for authoring the next generation of build tools with a strong emphasis on performance. Rust, in particular, is popular not only for its performance but also for its impressive developer experience — voted the “most-loved” programming language six years in a row in the Stack Overflow Developer Survey.
In speaking about the decision to build Rome (the build tool and not the city) with Rust, Jamie Kyle says:
The project SWC is at the forefront of this idea of using Rust for frontend build tools. It is now powering projects like Next.js’s new compiler, Deno, Parcel, and others — with a performance that is many orders of magnitude above other existing build tools.
Projects like SWC prove that with a change of the underlying platform, the performance of build tools can be significantly improved.
A paradigm shift
To improve the performance in development mode, a lot of the optimizations that would take more time to complete are left out and are, instead run when we are bundling our production application. This ensures that it takes as little time as possible to spin up a dev server, run our application in development mode and get productive.
The bundling process still takes quite some time, though and as a project grows, build times (even in development) only get longer and longer. Wouldn’t it be great if we could somehow skip bundling altogether while still being able to write modules as usual and have the browser understand how to work with them? A new set of build tools is taking this approach, known as Unbundled Development.
Unbundled development is great. It solves a major issue with existing build tools: they often need to rebuild entire sections of your application for even trivial code changes, and build times get longer as the application grows. We lose the rapid feedback — essential for a pleasant development experience.
One might wonder, if unbundled development is so great, why isn’t it the norm today? There are two major reasons why unbundled development is only starting to gain traction: browser compatibility for cutting edge features and processing node module imports.
1. Browser compatibility for cutting edge features
export keywords that we are all familiar with;
ES modules have been around for quite some time. Still, we are only able to start taking advantage of it for things like unbundled development, mostly because of how long its standardization took across all the players in the web ecosystem.
In her article about ES Modules, on Mozilla Hacks, Lin Clark says:
The issue of browser support or lack thereof has plagued frontend development for a long time. This is why we have vendor prefixing our CSS, sometimes the reason for polyfills, why we spend time ensuring cross-platform support for our web applications, and why it sometimes takes quite some time before we can take advantage of the latest and greatest web features in our day to day work.
Try to visit a StackBlitz project using Safari, and you will be greeted with the following screen communicating the lack of support for WebContainers in non-Chromium-based browsers.
The pace of feature adoption is not the same across browser providers, and there are often variations in how different vendors implement certain features. However, the future looks bright with initiatives like Interop 2022.
2. Processing node module imports
Most, if not all, frontend applications we write today depend on external libraries from NPM. For a typical react application, We would import react at the top of our component files like so:
Trying to load this directly in the browser, as we would need to do for unbundled development, will lead to two issues. First, the browser does not know how to resolve the path to find
react and secondly, the react library is published as a Common JS (CJS) module — which cannot run natively in the browser without some pre-processing.
The latter is the bigger issue here, as it is possible, and even trivial, to simply replace our node module imports with relative paths to specific files. Still, the fact that most NPM packages are written in a module format more suitable for Node JS than the browser requires that our NPM dependencies are treated specially to facilitate unbundled development.
With ES Modules now being mainstream in most modern browsers, and clever workarounds for NPM dependencies, build tools like Vite and Snowpack can offer the option of unbundled development with drastically improved performance, snappy builds, besides super fast HMR.
Frontend development has come a long way, and our needs are constantly evolving and increasing in complexity. Build tools are an essential part of how we build frontend applications, and existing tools are falling short of the mark sparking the development of new tools that reimagine how build tools should be designed.
With a huge focus on performance, ease of use, and less complex configuration, the next generation of build tools are poised to power ambitious frontend applications for some time to come.
Are you excited about the recent developments in this space? Let me know in the comments what you think about the upcoming innovations and the current landscape.
This content was originally published here.