Memorized Components Being Unnecessarily Re-rendered

[mwskwong]

Describe the bug
When using useTranslation in a memorized component (either memo or PureComponent), the component is being unnecessarily re-rendered. For me, it seems that this only happens if i18next-xhr-backend is used. In the codesandbox sample, 1 extra render is fired. While the worst I have met (in rare situation) is 7 extra renders.

Occurs in react-i18next version
[email protected]

To Reproduce
Just use useTranslation in a memorized component
https://codesandbox.io/s/react-i18next-test-wzj08
https://codesandbox.io/s/react-i18next-test-z4jx0 (extended by @jamuhl)

Expected behaviour
A memorized component should behave the same as a normal component in terms of utilizing react-i18next.

Additional context
Original discussion: #710 (comment)

@jamuhl Please read the console output for details (and forget whether the backend locates the translation.json or not).
** ONLY causes unnecessary re-renders when using i18next-xhr-backend**

https://codesandbox.io/s/react-i18next-test-wzj08

If you believe it is easier for you guys to trace and maintain, I can open a new issue instead of continuing the discussion on this issue.

Wow...that is so weird - I extended your sample a little: https://codesandbox.io/s/react-i18next-test-z4jx0

(making sure no saveMissing, ... gets in the way, also adding a counter to the MemorizedButton to see it gets rendered to DOM only once - while called twice)

I moved the sample to local so I could add some more logs to the useTranslation:

// ...
  if (!ready && !useSuspense) return ret; // not yet loaded namespaces -> load them -> and trigger suspense

  throw new Promise(function(resolve) {
    console.warn(`${props.name} throw Promise (trigger Suspense)`);
    loadNamespaces(i18n, namespaces, function() {
      console.warn(`${props.name} set state (loaded namespace -> sets new t function)`);
      setT(getT());
      console.warn(`${props.name} resolves Promise (resume Suspense)`);
      resolve();
    });
  });

The SimpleButton gets triggered it's rerender by the setT. Immediately after the MemorizedButton also renders - honestly I have no idea why. Then Suspense get resolved by the SimpleButton - removing the fallbackUI and showing the real content again. Then setT gets called for the MemorizedButton - and it rerenders (calls the function internally) while the rendered result stays the memoized memorizedButton1 (counter says 2)...

Change the order of the 2 Buttons MemorizedButton before SimpleButton -> only one render as SimpleButton gives a shit about MemorizedButton gets setT called.

Move every Component in own Suspense -> both render only once

So...if this is related to react-i18next - feel free to provide a PR...as I don't know which magic triggers a rerender on the memorized component...I have no clue how to fix this...

To make it more clear - I got no idea what is triggering this render:

I just know it's not useTranslation (neither the loading done as it comes later nor the bound events)

[jamuhl]

Fun fact...removing memo:

So with memo we get a rerender on Suspense triggered in the component, without not. The cause seems to be within memo...which asserts a Component using useState gets triggered a rerender on state change...

[mwskwong]

That is kind of a workaround and I actually have noticed that before. The problem is, sometimes I have to use memo + useTranslation to improve the performance of loading a heavy component.
e.g. an SVG chart with localized title, axes and legend labels.

[jamuhl]

I think there is no fix for this...we use Suspense in a very experimental way - outside of the only use case of react.lazy...

a) there is a workaround moving memo in an own Suspense
b) pass in t as props from the outer Component...
c) setting useSuspense: false and handle the ready flag yourself with rendering alternative fallback if ready is false

[mwskwong]

@jamuhl Ok, definitely a problem with useTranslation, have no such problem while using withTranslation
https://codesandbox.io/s/react-i18next-test-wzj08

[jamuhl]

funfact -> withTranslation uses useTranslation

[mwskwong]

Then it is really weird. As you can see in codesanbox, withTranslation indeed does not causes any extra rendering.

[jamuhl]

The only way to find out what happens would be to debug the react-dom reconciler as memo HOC gets handled there... 🤷‍♂️but that is a rabbit hole too deep for me 🥺

[jamuhl]

Barebone sample to reproduce: https://codesandbox.io/s/great-grothendieck-oof2w

[takakobem]

@matthewkwong2 This issue might be resolved by #1174

[tigerabrodi]

@matthewkwong2 This issue might be resolved by #1174

@jamuhl @adrai It seems like this issue was resolved already. I'd propose we close it 🎉.

Side question: I was wondering if there are more things I can do to contribute to this project? Also, do you have any plans for next year and the future of react-i18next? I'd love to be apart of it 🥰

[jamuhl]

@tigerabrodi for react-i18next there are currently no bigger plans. For i18next we're somehow waiting for an outcome of https://github.com/unicode-org/message-format-wg . If there will be a new standard come out for i18n we might adapt that in future...

In general...like said...we're more than happy for any help