form-1 components, reg-sub-raw subscriptions and clear-subscription-cache! infinite render loop

[bwalex]

I'm using re-frame 0.10.2. Following what was discussed in #218, I have form-1 components using subscriptions. In particular, they are subscriptions to a reg-sub-raw sub.

I'm also using hot-reload using figwheel, where my reload handler is:

(defn mount-root []
  (re-frame/clear-subscription-cache!)
  (reagent/render [views/main-panel]
                  (.getElementById js/document "app")))

(defn on-js-reload []
  (mount-root))

When the page is loaded for the first time, everything's fine - render gets called once or twice, depending on what exactly the subscription does to the atom, but it stops after that and everything's ok.

However, when the hot-reload path gets triggered, it ends up in an infinite render loop; the render causes a new subscription to be created, the new subscription changes the value deref'd in the component, and that again causes a re-render with a new subscription being created.

If I comment out the clear-subscription-cache! call, things work as intended. That makes some sense based on the comment in #218 saying that it works because of the subscription cache (and sub dedup).

I'm not sure it matters ( I don't think it does) but in the reg-sub-raw-created subscription I'm creating a new reagent.core/atom instead of using the re-frame app-db. The subscription handler is a wrapper around the apollo graphql client, which implements its own cache, so using the app-db wouldn't really be productive here.

If I had to guess, what's happening is that the render method of the already-mounted component (as it's just reloading) gets called the second the subscription gets called, before it gets a chance to update the subscription cache. That render itself also causes a new subscription to be created, and, since the cache hasn't been updated yet, it just goes around in an infinite loop.

[danielcompton]

Can you post a project that reproduces this, and also try updating to the latest beta? There were some changes in #428 but that isn't in 0.10.2.

[bwalex]

0.10.3-beta1 indeed seems to fix the issue.