Replace Create React App recommendation with Vite #5487
Conversation
Size Changes📦 Next.js Bundle AnalysisThis analysis was generated by the next.js bundle analysis action 🤖 This PR introduced no changes to the javascript bundle 🙌 |
|
Totally agree that there are good alternatives to CRA now and React should more prominently feature them. However, as maintainer of Parcel, an alternative to Vite, I would appreciate if React did not play favorites (also with Next.js vs other frameworks). Since CRA was maintained by the React team it kinda made sense, but harder when it's not. Maybe a small description of each option with more details about what it's best for and the tradeoffs vs other options would be good? Like "if you're building an SPA, Vite and Parcel are good options. Vite is a bundle-less dev server that works with React and other frameworks ... Parcel is an easy to use bundler that works great with React ... If you're building a server rendered app, Next.js and Remix are good options. Next.js is a React framework with file system routing, server components, ... Remix is a React framework focused on web standards ... If you're building a static website, Gatsby is a good option. Gatsby is ..." These are really rough descriptions. Might wanna contact the authors of these tools for a short summary. |
|
Most would agree that the recommendation of CRA needs to be removed, considering it's a dead repo. But recommending a less optimal solution might be about as bad for a beginner as multiple recommendations. My completely subjective, gut feel is that vite should be that recommendation. But any active project would be fine. Multiple will just result in analysis paralysis, which kills beginners. More advanced users will find other projects on their own. At the very minimum, a single recommendation should be made for beginners and any other recommendations segmented away for more advanced devs to investigate for their specific need. |
|
@devongovett I agree. I would like to see a brief description of each option and avoid playing favorites, except, of course, for the inevitable placement of the options in a list which (despite using the But I want to point out an issue of almost equal importance, which I believe @t3dotgg can amend in this very PR by swapping just two lines. That is: Let's put Parcel above CRA. (If not outright removing CRA, as @david0178418 bravely suggest) As it stands, Parcel's DX is leagues ahead of CRA:
|
|
Quick preface: I am not associated with the React / CRA projects, just offering my two cents.
That's a lot to digest for someone who is just learning what React is. Many of the "new React devs" @t3dotgg mentions aren't going to know what most of those words mean, or if they need them. That's why CRA is useful, it's a one-stop "here's a React app, no other stuff" tool. That being said - @t3dotgg I'd suggest expanding on what the "countless issues" you are experiencing with CRA are. A Tweet with a bunch of engagement isn't really a justification, and this PR just comes off as inflammatory in it's current state. I'd also suggest people stop "Approving" this if you aren't part of the React team, you aren't helping anyone by doing so. |
|
@mattwilkinsonn the issue is that there are a myriad of issues that can't be resolved when a project is dead. Check the last year of activity. The last commit was in Sept last year. But there are 1.5k filed issues and over 400 prs lying idle that could be resolving problems. That not ideal for beginners who would have pain trying to debug these. |
I agree with that, but I also think React is used for a lot of different use cases, and there isn't a one-size-fits-all tool. It's important for beginners to know what the tradeoffs are so they can make an informed choice. A lot of current CRA users should really be using a server rendered framework, and conversely, some apps using server rendering might be better off with an SPA or static site (this is not intended to spur a debate 😉). As shown by CRA's popularity, React's recommendations have a lot of weight. The "Getting started" section in the current docs doesn't really show much other than an install command and linking to the CRA docs, and the next.js section only links to the next.js docs. The other tools only have links and no explanation about when you should use them. I think it would be helpful to explain what each tool does and why you should use it. One of the points about a tool could be "it's good if you're just getting started with React", so there could still be a recommendation for that as well. |
I don't think this is the case. It's not important to understand the intricacies when you're just getting started. Once you have a foundational understanding of the technology, it's great to be able to go and learn about the different directions/branches you can go down with the technology but that's only once you have an idea of what you're doing. These things are not necessary for a beginner to understand. And if you're talking about someone who is new to React but experienced in web development, then I wouldn't be calling them a beginner. |
|
This page is not only for beginners. In fact, if you read the whole thing, it already covers many aspects of this. There are descriptions of what a bundler does, what a minifier does, etc. There is also already a distinction between "minimal toolchains" and "frameworks". All I am suggesting is that there are similar descriptions for each tool listed with the use cases that it is good for. That doesn't preclude having a recommendation for "learning React" at all. At the same time, it helps beginners understand the confusing landscape of tools out there. |
I'm inclined to agree with @mattwilkinsonn's point about this being far too wordy for a new developer. However, I do agree that the docs shouldn't necessarily be opinionated about any one approach. |
| * [Parcel](https://parceljs.org/getting-started/webapp/) | ||
|
|
||
| > Vite's default templates don't handle backend logic or databases. You can use it with any backend. When you build a project, you'll get a folder with static HTML, CSS and JS. Because this template can't take advantage of the server, it doesn't provide the best performance. If you're looking for faster loading times and built-in features like routing and server-side logic, we recommend using a framework instead. |
There was a problem hiding this comment.
you could add on frameworks the usage of Nx Framework which includes Vite by default in React applications
There was a problem hiding this comment.
you could add on frameworks the usage of Nx Framework which includes Vite by default in React applications
Nx for a beginner???
There was a problem hiding this comment.
Actually you can see Nx as starter up as standalone app in version 15. You folks tied Nx to monorepo angular projects, but now Nx is much more than that.
There was a problem hiding this comment.
@nexxeln yep. "Standalone apps" support doesn't involve monorepos (which Nx might be mostly known for).
I published an article on that topic two weeks ago that addresses exactly this issue: https://dev.to/nx/react-vite-and-typescript-get-started-in-under-2-minutes-56f 😃
There was a problem hiding this comment.
So lets, make NX the recommended approach for react, it also handles migrations etc which is more than others offer. Plus thanks to community work you not limiting to a set bundler. You have SWC, vite, parcel and webpack. NX also working on remix if im not mistake @juristr...
PLUS nx moving from Angular to React so makes sense 😂
There was a problem hiding this comment.
I would say we want avoid throwing these concepts at beginners:
monorepos(not required but still very present in docs)- choosing a bundler (sws, vite, parcel, webpack: they don't feel confortable to choose)
- migrations
- generators
- targets/executors (let them learn package.json scripts first)
I concur that Nx is powerful and may seem simple for us that have more experience, but Vite is simpler and has less friction for beginners.
I would add that we should recomend the simplest way for them and Vite fits for being simple in the sense it doesn't try to have all these abstractions baked in.
Having both recommendations would not be advised as well.
There was a problem hiding this comment.
Nx doesn't require using monorepos anymore
There was a problem hiding this comment.
@wesleycoder yeah as @nickmccurdy said. Not Nx in the monorepo setup. This is targeted towards new learners as far ad I understood.
Nx's new "standalone app" setup is almost identical to what CRA produces, and allows ppl to choose whether they want Vite or Wepback
I wrote it up here: #5487 (comment)
Create React App has stagnated and is no longer a "recommended" way to develop a React SPA. See reactjs/react.dev#5487. Create React App also has a lot of dependencies, and at the moment, building the app gives a warning about babel-preset-react-app. While there are several possible options, the simplest one is to migrate to Vite. The alternative is to adpot a framework like Next.js, which is more work for an existing app. The transition from CRA + Jest to Vite + Vitest is quite painless. We also get some benefits such as reduced build time. We also upgrade to Node 20, and upgrade several yarn dependencies. Signed-off-by: Antonin Bas <abas@vmware.com>
Create React App has stagnated and is no longer a "recommended" way to develop a React SPA. See reactjs/react.dev#5487. Create React App also has a lot of dependencies, and at the moment, building the app gives a warning about babel-preset-react-app. While there are several possible options, the simplest one is to migrate to Vite. The alternative is to adpot a framework like Next.js, which is more work for an existing app. The transition from CRA + Jest to Vite + Vitest is quite painless. We also get some benefits such as reduced build time. We also upgrade to Node 20, and upgrade several yarn dependencies. Signed-off-by: Antonin Bas <abas@vmware.com>
Create React App has stagnated and is no longer a "recommended" way to develop a React SPA. See reactjs/react.dev#5487. Create React App also has a lot of dependencies, and at the moment, building the app gives a warning about babel-preset-react-app. While there are several possible options, the simplest one is to migrate to Vite. The alternative is to adpot a framework like Next.js, which is more work for an existing app. The transition from CRA + Jest to Vite + Vitest is quite painless. We also get some benefits such as reduced build time. We also upgrade to Node 20, and upgrade several yarn dependencies. Signed-off-by: Antonin Bas <abas@vmware.com>
create-react-app seems to struggle a bit [1] and many folks are recommending vite as an alternative, as that project is more active. For me the reason to switch to vite is to reduce dependencies and speed up the build process. Replacing create-react-app with vite reduce the number of dependencies from 1762 to 444. I was also quite surprised, by how small the actual diff in this commit is. From that perspective there is almost no discernible difference between the two tools. [1]: reactjs/react.dev#5487 (comment)
To future prove our identifier webapp, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ Related: reactjs/react.dev#5487
To future prove our identifier webapp, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ Related: reactjs/react.dev#5487
To future prove our identifier webapp, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ Related: reactjs/react.dev#5487
To future prove React Codespaces, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ and https://vitest.dev/ Related: reactjs/react.dev#5487
To future prove React Codespaces, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ and https://vitest.dev/ Related: reactjs/react.dev#5487
To future prove our identifier webapp, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ Related: reactjs/react.dev#5487
To future prove our identifier webapp, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ Related: reactjs/react.dev#5487
To future prove our identifier webapp, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ Related: reactjs/react.dev#5487
To future prove our identifier webapp, this change abandons react-scripts and moves everything to vite as a near seamless transition from the create-react-app based approach when this app was created. Reference: https://vitejs.dev/ Related: reactjs/react.dev#5487
The thing prompting this now is that our deployments are broken because CRA is giving us obscure build errors for some of our newer dependencies, like @radix-ui/themes. This is apparently caused by outdated dependencies within CRA's toolchain. (radix-ui/themes#104, facebook/create-react-app#11767, and similar.) It's unclear if or when CRA is going to update those dependencies. From reactjs/react.dev#5487 (comment) and the lack of CRA releases since then, it seems like people generally aren't holding their breath. I can't discern whether Vite is just JS framework hype, but for now at least, it seems easier to port to it than to keep limping along with CRA. So, to do the port: * Replace CRA's config files with the defaults from `npm create vite@latest`. * Delete all the Progressive Web App stuff from CRA. We're not worrying about that for now. * Update package.json... * Add `"type": "module"` per https://vitejs.dev/guide/troubleshooting.html#cjs. * Update scripts and dependencies per `npm create vite@latest`. Drop various testing-library dependencies added by CRA, since we weren't using them. * Separate dependencies from devDependencies while we're here, just for human organization. * Replace CRA's default readme with Vite's default readme. * Port unit tests from Jest to Vitest. This just involves using explicit imports for the globals like `it` and `expect`.
Create React App is not a great recommendation to be making, especially for newer developers.
As an educator, I run into countless issues w/ new React devs running into unnecessary issues due to the continued recommendation of CRA. I made a tweet about this and was amazed by the amount of devs who agree
I'm in no way attached to the specific changes I made here, mostly opening this in hopes of a discussion around the technologies recommended in the new Beta docs.