subscribers aren't triggered when subscriptions are reloaded via figwheel

[aaronjensen]

If I make a change to a sub, I have to reload the page or otherwise change the state in order for the sub to get triggered. Ideally, the subs should fire if they're reloaded.

[mike-thompson-day8]

I might need a bit more information. Does this reloading involve figwheel? If it does, what are doing in your on-jsload fucntion?

[mike-thompson-day8]

BTW, both re-frame examples now better demo a figwheel workflow. But that cleanup is only in the develop branch for the moment.

[aaronjensen]

Indeed.

I had this, which may be old config:

(figwheel/watch-and-reload
  :websocket-url "ws://localhost:3449/figwheel-ws"
  :jsload-callback core/mount-root)

[mike-thompson-day8]

I still don't know what core/mount-root does. If there is a problem it will be what you have done in there.

Please have a careful look at the various examples. Like this one:
https://github.com/reagent-project/reagent-template/blob/master/src/leiningen/new/reagent/src/cljs/reagent/core.cljs#L47-L48

Notice how mount-root completely re-renders from the top down.
(reagent/render [current-page] (.getElementById js/document "app")

I'm going to close this, because it is not really not a re-frame issue.

[aaronjensen]

Ah, sorry, I thought that was from a re-frame template.

My mount-root is

(defn mount-root
  []
  (reagent/render [my.views/my-app] (.getElementById js/document "app")))

So yes, it rerenders the entire page, at least in theory. In practice, it doesn't even call my.views/my-app, likely because its result is cached. Note that everything works properly if the views.cljs file is the one changing, but if it's a file w/o the views, then the render is effectively a no-op.

[aaronjensen]

After digging in though, you're right, it's not a reframe issue. I thought it had to do w/ the subscriptions, but it actually has to do w/ reagent/render doing its job. I'll check w/ reagent to see if there's a way to blow the cache.

[aaronjensen]

For posterity, I tossed ^:figwheel-always on the views file namespace and that takes care of things.

[mike-thompson-day8]

It is puzzling that was needed. As I understand it, you are modifying a subscription handler, right, so that subs.cljs file will be reloaded, and then mount-root is being called by figwheel.

[aaronjensen]

So here's the real fix:

(defn mount-root
  []
  (reagent/render [#'my.views/my-app] (.getElementById js/document "app")))

Note the #'

See: https://github.com/reagent-project/reagent/blob/ff10d235911a60135d52c18fced9f071a7d8e366/src/reagent/core.cljs#L100-L104

[mike-thompson-day8]

Again, I'm puzzled as to why that would be a real fix. You aren't calling force-update-all you are force rerendering the entire DOM - a sledge hammer (which is good).

For further background:
reagent-project/reagent#94
reagent-project/reagent#112

[aaronjensen]

Here's a repro. You can ignore the piggieback stuff, just check the links. I'm guessing the problem has to do with memoized ratoms, but i haven't dug too deep.

https://github.com/aaronjensen/figwheel-plus-vim-fireplace

[vganshin]

I ran into the issue with shadow-cljs, and while googling the solution I found this issue. Then I found re-frame.core/clear-subscription-cache! function in the source code, which should be on every code reloading while development before React re-rendering. Maybe it will save time for someone.

P.S. See also Why clear sub cache note.

[mike-thompson-day8]

@vganshin I also would suggest creating an app using re-frame-template and noticing the way that it structures things.