Skip to content

Minivera/lander

Repository files navigation

Lander

Lander is yet-another-frontend-framework aimed at exploring performance optimizations through the use of localised components. It is currently very experimental and only provides the bare minimum to get an app going. It also has some technical issues and limitations that make it very hard to recommend, even in development projects.

Docs

If you are interested to learn more, check out our docs for the individual packages. The rest of this README will cover the reasoning and ideas behind this framework.

Why Lander?

There are no shortages of frontend frameworks and most projects are likely to end up using one of the big ones such as React, Vue, Svelte, etc. Lander was not created to replace those frameworks, I could never hope to have the capacity to build the kind of products these teams do. Rather, it was created for research purposes. While I do hope Lander can one day become mature enough to be used in production projects, its main objective at the moment is to serve as a tool for testing and experimenting with algorithms.

Lander was initially created as a learning project for myself and went through, at least, 4 versions. It was shown off at ForwardJS 2019 and iterated over throughout the years. Since then, the repository was used for researching performance and testing ways of improving the diffing algorithms for university research. The current version of the library is an updated version of these experiments.

Features experiments

The main goal of the current version of Lander is to try to solve two challenges I had identified with the existing algorithms and implementations for the modern, VDOM based, frontend frameworks. The VDOM is an added overhead and literature I found on the subject suggested replacing it as a solution to its inherent drawbacks. My experiments were rather exploring if we could reduce the load on a single VDOM algorithm as a way to increase performance and maintain the great DX offered by these algorithms.

In addition, the versions of React, Vue, and a few others frameworks I explored at the time often suffered from a lack of portability. In the case of React 16, for example, the understanding is that there is only one single "instance" of React running as your application, you should not split it up. This assumption might be wrong in the recent versions, but it becomes very much true in all versions when factoring it some of the global state libraries. In all the React codebases I worked on, there was only a single entry point for an entire React application. You couldn't render it in multiple places in the application or combine it with static HTML, React is the application.

To solve these challenges, Lander uses web component and a localized version of the (fairly slow) VDOM algorithm we purpose built for the needs of this library. These optimizations work in the following ways:

All components built with Lander are rendered in the actual DOM tree. They are not kept as virtual components only used to render other elements, but rather as elements in the tree. Whenever a rerender is requested, only the closest latest component will rerender. Updates do not cause the entire tree to diff, but rather only the closest tree to the update. The diffing algorithm walks the tree of the updated component and, whenever it hits another component in its children, it requests an update for that component. Diffing is essentially isolated within a component. The intent was to avoid diffing, even if optimized, an entire complex tree if only a few leaf elements change.

Since components are actual elements in the tree, we are also able to store context and state in that element's memory. While this was not an optimization I ended up going for, we could technically not keep a virtual DOM tree an only diff the actual DOM, since all parts of the virtual tree exists within the real DOM.

Finally, using web components means that we can make components portable and render a component anywhere in an application, even in an application using other frontend frameworks. There is no global memory storing the tree and components states, which means that multiple instances of Lander can live alongside one another. One goal of the library that I could not implement was to allow adding components by name in HTML files and have Lander pick them up for render when the application loads. The current version still requires an index file to start any root component you want to render, like with ReactDOM.render.

What is the future for this library?

The academic portion of Lander has been completed and is unlikely to continue as my own academic interests don't align with this type of research. Ultimately, I would hope to complete the library and release it to the public in a state where interested people can use it in production environments. It is not there yet, unfortunately.

At the moment, the library is pretty-much in archive mode and was only worked on recently to write this README and solve any critical security issues. I do not plan to archive this repository, but I strongly recommend NOT using it in any kind of production environment. Pull requests are unlikely to be reviewed, though I will answer any questions or comments posted as issues. If this library looks interesting to you, please reach out!

Is lander related to go-lander?

You may have landed here (pun totally intended) through my other library of a similar name, go-lander.

Go-Lander initially started as an implementation of an earlier version of Lander in Golang, compiled in WASM to use as an alternative frontend framework. The early algorithm for Go-Lander was a carbon copy of the ForwardJS 2019 version of lander. This older version did not work and was full of bugs. The plan of Go-Lander was to be the server side implementation of Lander, which would compile to JavaScript Lander components augmented with WASM code.

Go-lander has since been fully rewritten and optimized, and it has lost any link to Lander in the process. The two libraries are not intended to become compatible or use one another. If anything, I would be more likely to turn lander into a Web Component utility for creating Go-Lander components rather than having the two work together.

If you want a better and actually finished version of Lander, please check out go-lander. Go-Lander is not production ready however, WASM support in Golang is still experimental and unstable. Use at your own risk.

Why "Lander"?

Because it makes landing pages.

Releases

No releases published

Packages

No packages published