Comment by Richard Kho on "Would you use create-react-app for a new project?"

I’m fundamentally opposed to the idea of surrendering your configuration to something that may be too broad (installing dependencies that you may not need) and not fully understanding everything that’s going on under the hood.

Learning how to set up your own Webpack configuration as well as architect your own React application are things that are very useful to know.

If you’re trying to churn out a quick prototype, I could understand why you would want to scaffold it. Otherwise, why would you ever push code you don’t fully understand into a production environment?

While I generally agree that it's good to understand how tools work, I do wonder how far you think it is worth going. For example, Webpack uses a JavaScript parser called Acorn. Does one need to understand Acorn (and how parsers work in general) to build a React app? I would say no, although it is useful to some degree.

Acorn is on a lower level of abstraction than Webpack. Similarly, Webpack is on a lower level of abstraction than Create React App.

The way I see it, you can start from any level of abstraction that helps you be productive. Then you can start peeling back layers, and decide if understanding those is valuable to you or not. For example, understanding Webpack will help you create much more customized setups. But you can go deeper, and understand Acorn to work on support for new syntax proposals in Webpack itself. And so on.

I think it's important we don't insist some tools are "real" and some aren't. These are all different levels of abstraction, and as long as they help you solve the problem, they're all good. But I agree there is value in looking at what's beneath them when you have the time and need for that.

Hi Dan!

Thanks for your reply. I think I really poorly articulated my position earlier and wish to retract.

I agree with you that I don't believe understanding Acorn is important in building a React app and that CRA is a great way to be instantly productive in React.

My feelings towards not using it comes from an experience I had with a different generator (Yeoman).

The first time I ever tried to scaffold an Angular 1.x app with Yeoman, I completely destroyed my productivity. For some reason, the recipe I used would not play well with both ng-router and ui-router, and I ended up spending hours debugging something only to throw it all out and start from scratch.

It was a tremendously valuable experience for me: I put my trust in something I didn't fully understand and learned that I couldn't figure out how to fix it. Once I decided to restart with a blank slate and be careful about the choices that I was making in the project, I had a much easier time and things just worked. This left me with the impression that struggling through setting something up at least that first time is super valuable.

Thinking about this topic a lot more, I think I can find more scenarios that make create-react-app useful to more than just prototypes.

That's a fair point, and I agree with it! We've tried to avoid this kind of problem by:

Only focusing on build tooling and not generating any code. So you can drop any router (or any other runtime library) into the project without worrying that it would not match the conventions.
Not generating configs. Since configs are hidden, you can’t mess them up.

I agree that there are corner cases where the abstraction is leaky, and it's definitely useful to know what underlying tools are used.

I like the level of discussion in this thread.

There is another point of view. If you know webpack (or name the abstraction level you want here) THAT good you want to avoid "copy-pasting" all the time. So you either build something like CRA or use it happily and you can easily debug it!

Search Hashnode