React v16.2.0: Improved Support for Fragments
reactjs.orgSeveral people had already noted [0] that you could implement equivalent behavior by using a component that simply returns its own children:
const Fragment = ({children}) => children;
// later
return <Fragment><Component1 /><Component2 /></Fragment>;
Nice to see this added as both a built-in component type and a useful syntax extension. (Now if they'd just modify JSX syntax so that we can pass props by matching local variable names similar to how ES6 object literals work, like `<SomeComponent a b c />`, rather than having that become a defaulted-true boolean, I'd be happy :) )[0] https://medium.com/@gajus/using-react-v16-to-create-self-des...
To be pedantic, it's not 100% equivalent. You can find the nitty-gritty details about the (intentional) differences here: https://github.com/facebook/react/pull/10783.
We did mention the author of react-aux in the blog post though :-)
>Thanks to Gajus Kuizinas and other contributors who prototyped the Fragment component in open source.
That works in ReasonML :)
Ive started with React recently and strangely the lack of fragments were one of the things that caught my attention in the first days.
Another thing I'm looking for in React as I get started is better styling support. I'm resourcing to React JSS for that, but I feel it should be part of the language. The web trinity if you may is HTML-CSS-JS and React basically left CSS out half-baked into almost unusable style attributes, so I still have to resource to className and CSS. Or am I just being noob-anal?
I've found styled components to be a nice attempt at bundling styles with the component. Haven't tried that many options though.
It has out of the box support in Atom and scss (like?) syntax.
I agree with you that CSS support is a bit lacking. I think they did this on purpose though, because many people prefer CSS preprocessors like SCSS/Sass or Less. However, with Create React App, you get nice CSS support via webpack and PostCSS. I recommend you check out styled-components; getting used to the back tick syntax takes some time, but after you do get used to it, working with them is really nice. Also, Babel syntax highlighting (in Atom at least) properly highlights the CSS used in styled-component.
Another vote for styled-components. I tried everything I could (model and library wise). That's the one that felt the most natural to me.
I appreciate the addition of Fragments, but I'm not sure if the addition to the JSX syntax was really a good decision.
While it initially took some time to wrap my head around JSX (especially with a lot of JS mixed into it), after that I really started to enjoy JSX for its simplicity. The Fragment syntax adds some additional magic to JSX (which is rarely a good thing), and will also very likely break all the syntax highlighting out there again when used.
> The Fragment syntax adds some additional magic to JSX (which is rarely a good thing)
In: <></>
Out: <React.Fragment></React.Fragment>
WITCHCRAFT!! SORCERY!!
Teasing aside, I don't get what's magical about this. It's a straightforward transform. You can call that "magic," I'll call it "syntax."
Sure, we can call it syntax. Of course, for anyone who already knows what it does, it's just a transform, but the problem with extra syntax is that it makes it harder for newcomers to pick it up.
Taking your example when putting it in front of someone seeing it for the first time:
In: <></>
Out (Newcomer): WTF does that mean? Did someone forget to put something in those tags?
Same newcomer, in a couple of weeks: "Don't know how I lived without it".
I understand your point, but in practice people love tiny affordances like this, especially when they are needed as often as fragments are.
This is a balance, but it's important to consider the needs of power users too. Beginners don't stay beginners forever. We've thought about this for a few months, and decided that adding syntax would be better. I understand some people will always disagree though, and that's fine :-)
Yeah, it seems like the decision to make the fragment a first-class JSX concept has a lot to do with portability between renderers. The post makes it clear you all have given this a lot of thought and care, and I appreciate that (even if it will take me awhile to get used to seeing empty tags, haha).
Sure it's straightforward enough and much better than what we had before.
But for a newcomer to JSX it looks odd.
I'm probably just missing the reason for the syntax, we're already compiling - why can't fragments be added automatically as required?
I'm guessing it's because JSX is XML-like and so parsers may require a single root element for each JSX "document". It's a good question though, hopefully someone from the React team might be able to comment.
Because it is ambiguous. :-)
https://news.ycombinator.com/item?id=15808311
If you have a specific proposal for how it should work, I'm happy to look at it and try to poke holes in it.
It would be, wouldn't it. Silly I didn't realize that right off the bat.
Can't imagine putting forward a proposal that wouldn't in essence be what you're doing, i.e. wrapping in JSX delimiters. Because attempting anything fancier than that would probably have edge cases for days, plenty of material for you to poke holes in.
These will be quite nice when wrapping things like css grid components which must be on the same DOM level.
Man, although fragments are a pretty useful concept, the implementation is really terrible. It feels so uninspired and without thought of the design, in my opinion. They could have chosen some character instead of an empty tag and it would have been much better, like an asterisk[1], because it would be clear that the tag serves a defined purpose. I would almost argue that if this weird shorthand is necessary to accomplish this task (something that irked me when creating text heavy pages) then JSX should be redesigned to handle it natively without the weird fragment syntax.
1: Like this:
<*> ... </*>>feels so uninspired and without thought of the design, in my opinion.
Thanks for feedback! I assure there wasn't a lack of thought, in fact we discussed this for several weeks before even beginning to implement this. Your proposal was also considered but it doesn't bring anything over <>, whereas * already has a meaning in JS related to generators. It would be confusing to use it here for a different purpose.
FYI, <> is not original. As mentioned in the post, it has prior art. It already went through standardization once (in E4X). It has also been used for months (maybe over a year) in ReasonML, and its users reported really enjoying this syntax.
You may not like it but it doesn't mean we haven't given this any thought. :-)
Yes, technically if the empty tags do suffice, then the asterisk brings no additional value, and it would be potentially confusing to use a character that already serves a purpose in Javascript, although I doubt it since JSX tags don't support generators as an attribute or name. You could replace the asterisk with a different character in that case. My main point is that the empty tag feels, well, empty and made without a real purpose, in part because it's just so foreign compared to HTML (which is what JSX is based on?) as this kind of thing is never found in HTML. While I'm sure that there wasn't a lot of discussion within the React team about this, the inclusion of at least a single character, whatever that would be, would make it feel like it was actually designed properly, and I would also argue that adding something to make the tag feel nonempty would actually have a value (to make the tag explicit and not confusing).
I don't necessarily buy the prior art arguments too because E4X was barely used even though it was a standard and is now deprecated, and ReasonML is pretty specialized and much less popular, as well as has a different set of users, than React. If you had told me a product in the same category as React, like Angular, Vue, Ember, Ractive, etc., used that syntax, I would be much more receptive.
>I doubt it since JSX tags don't support generators as an attribute or name
Not currently, but we do have some plans regarding do-expressions, generators, and JSX. So * might get a meaning, even though not in attributes but children.
>My main point is that the empty tag feels, well, empty and made without a real purpose, in part because it's just so foreign compared to HTML (which is what JSX is based on?) as this kind of thing is never found in HTML.
I think most React users in practice don't touch HTML that often. JSX has quite a few differences with HTML, and we don't really plan to bring it closer to HTML in the future. It is more important to us to make JSX pleasant and convenient to everyday React users than it is to have 1:1 mapping to every HTML concept.
>ReasonML is pretty specialized and much less popular, as well as has a different set of users, than React.
I would argue that people using ReasonML today are the same kind of early adopters (and often even the same people) that started using React four years ago when the ecosystem didn't exist (and shaped that ecosystem). I trust their judgment.
>If you had told me a product in the same category as React, like Angular, Vue, Ember, Ractive, etc., used that syntax, I would be much more receptive
By that logic, we could dismiss React when it came out, because it was not like anything else :-) Give this syntax some time. It felt odd to me at first, after living with it for a few months it feels very natural. I'm sure you'll get to like it eventually too. At that point it might feel odd that other libraries don't provide this syntax.
I love to see these small improvements added to React.
Everything is not perfect though, some issues have been here for a looong time. At the moment, my biggest problem is the non-support of passive events [0]. And today, some browsers are starting to make some events passive by default (`touchmove` on Chrome for example), it is quietly starting to become a serious problem for developers to get things working fine cross-browser without anti-patterns.
I was already thinking "but what about Typescript..." then saw it already supports it :)
Well done react team and contributors.
Why can't they automatically wrap every bit of JSX with a <Fragment> tag? That way users would need to include an empty tag when they're returning multiple elements, and they're not it'll be a fragment tag that just contains a single element.
Because you don't know where JSX ends and code begins. Don't forget that fragments can include text nodes too.
return <div />
// am I a comment? Or text after div?
Ah, hmm... good catch.
This is a bit off topic, but does anyone reading this happen to know the status of async rendering in react?
There was chatter about it for version 16, and then nothing really else. I'm curious about what they are testing or at least planning in this area.
Unfortunately i've found it really hard to search for anything related to async rendering for react, as there is just so much other crap about react and "async" it just gets lost.
You can track it here. Still plenty of work to do but we are making progress.
Thank you! This is exactly what I was looking for and it seems all search engines have failed me!
Fibers? I thought they were implemented and that was what React 16 was all about.
From React 16 blog post (https://reactjs.org/blog/2017/09/26/react-v16.0.html):
>We think async rendering is a big deal, and represents the future of React. To make migration to v16.0 as smooth as possible, we’re not enabling any async features yet, but we’re excited to start rolling them out in the coming months. Stay tuned!
The rewrite was about adding the foundation for making this even possible. But there's still work to do before we can enable async rendering by default.
Ah ok thanks, not super familiar with react. Had seen a video about fibers/async and saw something about 16 implementing them. Put 2 and 2 together and got 5.
No mention of prior work? This is part of web platform since forever. See document.createDocumentFragment().
We did mention some prior art in the section discussing the new syntax:
> Fragment syntax in JSX was inspired by prior art such as the XMLList() <></> constructor in E4X. Using a pair of empty tags is meant to represent the idea it won’t add an actual element to the DOM.
You're right that we could have referenced document.createDocumentFragment(), too. Perhaps we'll add that to the documentation. Thanks for the suggestion!
Thanks, it's a useful part of the DOM, perhaps somewhat overlooked.
I just can't help but sneer. Opinionated document.write() using "jsx".
Working with react has been the worst experience of my life. I had to call a meeting and laid in down in October. If 2018 is React, I am leaving on Dec 31.
So soon I am going swiftly back into the arms of Angular and I can't wait. Honestly, my eyes are blurring as I type this with pure happiness.
Enjoy the fragments guys, it will be good practice for picking up what little is left of your happiness and transferable skill set in years to come.
>back into the arms of Angular
Curious: which one are you referring to?
Not to feed a troll but how is doubling down on _Angular_ better for a ”transferable skill set” than React?
Each to his own.
Kinda weird that it upset you that much.
Next time post this kind of stuff on your blog, not HN comments.
you are totally right