So with these virtual DOM libraries, after seeing a lot of them try to fill in this supposed niche, and seeing two this week including this one, I'm of the opinion that it should either be completely JS looking, e.g. a simple wrapper around createElement, or React looking, e.g. with JSX.
The ones that are in between, while potentially having nice syntax or a nice looking homepage, aren't enough to convince me to use them over React, mainly because 1) if I'm doing any production code, I want other engineers to be familiar with it and its pitfalls; 2) I want to make sure it'll be supported 6 months down the road; 3) engineers are familiar enough with JSX or with createElement, but some other syntax may be confusing and have corner cases that a bigger project like React hasn't thought of... I don't want a lookalike React.createElement because I can just use React.createElement; 4) many of these libraries are built with "performance" in mind, but does it really matter how many more times it can process a ToDoApp if it doesn't really affect user facing performance?
While this isn't a jab at this library specifically (although it hasn't had much substantial development in a while...), but too many libraries are created to fill some supposed niche and then abandoned when they move to different pastures, and if you need to fix any of those corner cases you will inevitably run into, ideally you can submit a PR and get it fixed easily and with little delay.
I'm interested in tools like this because it allows me to build my front-end out of regular functions, which I can compose however I want using regular code. I never liked writing HTML or CSS in the first place.
> I never liked writing HTML or CSS in the first place
I also did not like the split between code and markup and or style.
> build my front-end out of regular functions, which I can compose however I want using regular code
Yes, me too! I want to share what I found.
PureScript. Has some nice libs like Thermite (React under the hood) and Halogen (no React, plain PureScript, cleaner). And much more then these two...
Elm. Has the batteries included. Nice lib in this language is elm-css, which makes CSS/HTML styling symbols typesafe!
GHCJS. The heavy weight. Compiles Haskell to JS (Elm and PureScript merely inspired by Haskell). Coming along quite well, may not be very ready for production. Lots of interesting libraries. Big lib here is Reflex.
Isn't that a lot like hyperscript? Maybe I like this or mercury or I one of the other varieties better for one reason or another. Is ReactDOM the best fit for everyone?
I've never used HyperScript, and I made no claims about anything being the ONE TRUE TOOL. I suggested that yladiz's recommendation would fit your described preferences better without injecting JSX.
I really dig Maquette. The library is easy to use, the documentation is good, and the interactive examples make it really easy to get running.
The Hyperscript notation Maquette uses is compatible with other vDOM implementations, so it's possible to use JSX [1], Jade/Pug [2], Handlebars, or any other templating solution that compiles to Hyperscript [3].
I personally prefer really "dumb" views and have my controllers attach events and link model data. Thanks to the modular design of Maquette, instead of the Maquette `createRender()` pattern I was able to use SpineJS for the Model and Controller portions of the app.
I used Maquette recently to create an API and Web UI for Bull job queue, which I'll be releasing this week as open-source. I did not run into any bugs or difficulties in Maquette, it was as efficient and straightforward as advertised.
Previously I've used Templatizer (precompiled Jade templates), but Maquette's feature to map and track lists of objects/elements spurred me to change. I'm sold on it for now but really appreciate the thriving competition of vDOM renderers.
I find the virtual DOM is much easier to create using regular HTML syntax. I like Vue 2.0's approach where it creates a virtual DOM from the actual DOM itself.
Edit: I left a huge comment regarding POJOs and how the example on the homepage is pretty ugly, but I just found out that there's something called Hyperscript, and it's used to create a virtual DOM. So, while the example is in Hyperscript, Maquette also supports JSX.
I still don't understand the point of all of this I guess. Isn't it time we as JavaScript devs took a step back and thought "why do we need this?"
Obviously, because real DOM is (1) slow and prone to hard-to-predict reflows, (2) non-transactional, and (3) not going to change, due to backwards compatibility.
Thus the need to operate on a fast, cheap virtual DOM, and then apply the smallest sufficient change to the real DOM the users see.
The main conclusion from this benchmark was that all 'modern' virtual DOM implementations are all capable of achieving 60fps on large pages on most devices. There is no point in picking the 'fastest' anymore, because users simply will not see the difference.
The first line on the page reads: "This benchmark has some serious flaws, and the only thing that it tests is a children reconciliation algorithm."
Not exactly something to base an opinion on.
There is something to be said for using APIs whose total surface area is close to the surface area that you need. If I want to use different parts of three libraries, but those libraries have excess functionality with a lot of overlap, then I have to deal with potentially conflicting expectations. I also have to worry that the subset of functionality I do use will go away or change because it's not the central focus of the library.
The thing is, if you're using a VDOM, you'll probably want the other stuff Mithril's got anyways. And the tight integration with the vdom renderer helps make sure that rerenders happen only when something actually changes, and that partial states don't render. For 3 megs, I'll take it.
The ones that are in between, while potentially having nice syntax or a nice looking homepage, aren't enough to convince me to use them over React, mainly because 1) if I'm doing any production code, I want other engineers to be familiar with it and its pitfalls; 2) I want to make sure it'll be supported 6 months down the road; 3) engineers are familiar enough with JSX or with createElement, but some other syntax may be confusing and have corner cases that a bigger project like React hasn't thought of... I don't want a lookalike React.createElement because I can just use React.createElement; 4) many of these libraries are built with "performance" in mind, but does it really matter how many more times it can process a ToDoApp if it doesn't really affect user facing performance?
While this isn't a jab at this library specifically (although it hasn't had much substantial development in a while...), but too many libraries are created to fill some supposed niche and then abandoned when they move to different pastures, and if you need to fix any of those corner cases you will inevitably run into, ideally you can submit a PR and get it fixed easily and with little delay.