For those who use Vue regularly, are there any breaking changes in Vue3? How easy it is to upgrade from Vue2 to Vue3? Can one just upgrade the version in package.json and start using vue3 for new components?
Vue 3 introduces breaking changes, e.g. filters and event bus ($emit, $on, $off) are removed as far as I know. The Vue documentation team will release a migration guide when Vue 3 is released.
Another comment says something similar but I'll restate it. `$emit()` is still available on components and parents can still listen for them, this is just removing it from the `Vue` instance so that people don't create global (or scoped) event busses (which thankfully I don't do).
They are even adding the ability for the child to define what types of events they emit (I'm super excited for the autocomplete possibilities for this and how people implement it in TS) and add validation logic to the events they define (like not emitting if the data isn't valid/in the right state).
I got REALLY worried/confused when I saw the parent's comment because I was so confused at how to alert the parent of something in the child but it looks like everything is fine.
So the way I’ve been using it is pretty much for toaster notifications and asynchronous updates that are sent from the server to the client. Any component can subscribe to async updates on the shared bus and it will know when a particular object has been updated and can choose what to do with it. We don’t use the global store for all the objects because we deal with a lot of large page-specific state. Toaster notifications are also really nice this way: a lowly modal can send a top page “an error happened” notification without having to do the plumbing for this through every page and component. I don’t see any of that as an anti pattern.
And FB suffered from this due to undebuggable double notification bug because it could be really hard to debug when any component can send any event to other without any restriction.
They eventually moved to flux pattern due to this.
That said. It is possible to avoid such situation with strong lint rule. Like you did here. But you can avoid it completely with ~~JUST DON'T~~ .
Why should a framework encourage people to do bad things?
Um, no. That’s the whole thing is that the components don’t have to know how it is implemented. They do this.$bus.showError(‘Oops’). The bus implementation under the hood sends a message to the top level App component which in turn has the toaster notification implementation. I don’t need a linter to make this work.
You do need a guide to prevent people from sending random unrelated message to random handler hide deep inside some random component and emit again to other random component.
It `is` fine to use it and only use on notification purpose, but there is no way to enforce everyone to do so.
More of time it `will` be used in a terrible way that I just described (something like .emit('login') and trigger some random function causing xhr and emit notification everywhere).
You can't really assume everyone will use them nicely instead of resulting in spaghetti code. It is not the case most of time.
The experience I learned from I start coding is. `Anything designed to be easy to mess up will eventually mess up`
Maybe this is what happens when you’ve been coding for years and years but the thing is I was using this particular foot gun and it was working well. I don’t need a framework developer to babysit me and prevent me from misusing a feature. I know when it is a foot gun and I can treat it as such. It sounds like the Vue devs simply removed this feature that was in some ways useful and didn’t replace it with anything better. They removed it because it enabled people to do something they didn’t like, which yes as a maintainer of a popular framework is kind of your responsibility to be help people figure out. But this is like if C removed the ability to dereference pointers because it’s dangerous to do. Of course it’s dangerous, that’s why it’s so powerful. If you can’t use it responsibly, that’s on you, not on the framework developers.
Basically, it’s not like they added a formalized way to create notification buses. They simply removed a useful and widely used feature because they didn’t like how people were using it. This is exactly the kind of shit that killed Angular and honestly makes me reevaluate the trust I’ve placed in Vue. Improvements are always welcome. Wholesale removing useful features without providing an alternative, not so much.
Well, unlike angular, the event bus thing isn't integrated into vue self in any way (and it don't need to). It didn't prevent you from use it if you must need to either.
Moreover, You can polyfill it simply in one line. And it actually exist long before the vue ever exist.
const { EventEmitter } = require('events').
It always make me wonder why was vue ever want to pack this into itself given it did not use it as all.
It's mostly backward compatible, it depends on how complex your project is and what dependencies you rely on. Simple projects should be fairly straightforward - If you rely on the EventBus install mitt instead, will take you maybe 10 mins to replace. Rewriting your filters so they are plain methods instead should also be straightforward.
I tried to upgrade from vue 2 which using the minimized cdn js but couldn’t get it to work. Is there a minimized version of the vue 3 code somewhere that I could try?
The real advantage of using a compiler is that more complicated state changes can be instrumented without hacky introspection, which is probably why Svelte feels cleaner.
The big tradeoff is that it forces the use of the Svelte compiler. Which doesn't sound all that bad until you want to use something like TypeScript and realize that it's maybe a bit odd that a framework now has a tight grip over what language/tooling is used.
It seems like Svelte just recently finished their TS support, but I tried Svelte and while I liked the concept, the overreaching felt like at that rate I'd rather use ReasonML or something because Svelte JS is technically not JS, just looks a lot like it.
> We have started several branches (webpack5, next, vue3) and so far Webpack SSR support for Vue3 landed but not a precise timeline for beta testing at the moment. As many internals including vue-app and builder is to be rewritten from scratch to align with vue3. (of course with less breaking changes as much as possible)
