The only problem with using re-frame with Rum is that re-frame has a hard dependency on Reagent's atoms and reactions

You'd have to rewrite the internals of re-frame to use something else

There are a number of libraries already that you could use to replace Reagent if you want to do that

It seems most libraries that give your client code some structure depend on Reagent for the atoms/reactions. It's a shame, but understandable. I know only of @metametadata's excellent #carry library which doesn't rely on Reagent 🙂 https://github.com/metametadata/carry

All of these libraries are quite similar. One big difference, if we compare carry and re-frame, is that Carry doesn't have global state, instead all view-functions get their data as an argument. Re-frame's global state has the disadvantage of added "magic", but it can have better performance due to how React works.

If I understand React correctly, because re-frame components don't get arguments, we can completely skip the should-component-update calls, and we avoid situations where we re-calculate a component because its arguments change, but in fact it wouldn't have to change.

When we give the model as an argument, the least React will have to do is call should-component-update on our components. Needless re-calculations are avoidable, but due to programmer error we'll likely have some.

By re-calculation I mean that React calls the render functions to get the new hiccup, or whatever format we use 🙂

@metametadata, yes several instances of pure-frame that talk to each other sounds great

@metametadata, from your first comment, it is not clear but I think pure-frame has no global state, that’s the point

@metametadata thanks for Elm link, looks interesting

@kauko, carry looks great I will definitely study it

@kauko "re-frame components don't get arguments » ? I’m not sure about that, is it like reagent components? take props as 1st arguments and follow the react lifecycle if it changes? reagent is a bit messy for that I think, if props changes, lifecycles methods are called but not if a ratom is changing… can be troubling at first...

In re-frame you don't typically give arguments to your view functions, instead you use the subscriptions and whatnot.

oh I see, sorry

I don’t have much practice of re-frame

And about the "ratom is changing" thing: lifecycle methods are called even when your argument is an atom (or a reagent/atom), but atoms are not compared by their contents. Instead, you just check whether it's the same atom or not.

if the first arg to reagent component change it trigger lifecycle, as far as I know

with ratom magic, no lifecycle I think

but I may be wrong

Eh, I don't think it has anything to do with the order of the arguments 🙂

when you call reagent/props

you get back only the first argument of a reagent component

I have to test this further

Are you sure? Could it be just the arguments in a vector?

I’m probably missing something

yes i’m sure

It cause me trouble at work

🙂

the rerendering model of reagent is tricky I think

I have to go, see you later, thank you

@kauko here's what i tried to explain 🙂 from https://github.com/Day8/re-frame/wiki/When-do-components-update%3F Lifecycle Functions When props change, the entire underlying React machinery is engaged. React Components can have lifecycle methods like component-did-update and these functions will get called, just as they would if you were dealing with a React Component. But ... when the rerender is re-run because an input ratom changed, Lifecycle functions are not run. So, for example, component-did-update will not be called on the Component. Careful of this one. It trips people up.

maybe it is specific to re-frame, I believed it comes from reagent docs, not re-frame

Yeah that's exactly what I was talking about too

The thing is, the atom doesn't "change" when its contents change

yes

"props" is the first argument or "the entire args vector" in this case?

don't know

should experiment

"Props" in React are the arguments of the component, given by the parent.

ok so it is reagent/props + reagent/children ?

Hmm I don't know what reagent/children does actually. It sounds like it would return the child components, but the docstring says Returns the children passed to a component.

Why are we discussing this in #rum and not #reagent by the way? 😄

1st arg is props and others is children I think

lol you are right

Huh, I think you're right. It looks like with React.js you always give one props object to a component.

I didn't know that!

just ask for the lifecycle stuff on #reagent

asked

off topic police

🙂

@pbaille yeah, I wanted to say that pure-frame doesn't define a global model var, but it probably still gravitates towards the usage of a single model per the application.

@metametadata, I don’t know, I have to do experiments on combining “frames” together, sounds fun to try.

Did some work today consolidating global state in re-frame: https://github.com/martinklepsch/re-frame/pulls — I'd eventually like to be able to use all but the subscription stuff with Rum (and perhaps other React wrappers) but that will probably take quite a bit longer 🙂