fix(connectInfiniteHits): fix #2928 #2939
Conversation
| @@ -115,7 +118,10 @@ export default function connectInfiniteHits(renderFn, unmountFn) { | |||
| results.hits = escapeHits(results.hits); | |||
| } | |||
|
|
|||
| hitsCache = [...hitsCache, ...results.hits]; | |||
| if (lastReceivedPage < state.page) { | |||
There was a problem hiding this comment.
we'll need to be careful with this line if we ever allow "scrolling up" as in #1972
There was a problem hiding this comment.
Do you think it's blocking for now?
There was a problem hiding this comment.
I think for now it's fine, we just need to keep it in mind if we ever allow scrolling to go the other direction
|
Deploy preview for algolia-instantsearch ready! Built with commit 5581069 https://deploy-preview-2939--algolia-instantsearch.netlify.com |
| @@ -54,22 +59,28 @@ describe('connectInfiniteHits', () => { | |||
| createURL: () => '#', | |||
| }); | |||
|
|
|||
| // test that rendering has been called during init with isFirstRendering = false | |||
| expect(rendering.callCount).toBe(2); | |||
There was a problem hiding this comment.
Is it intended to remove this assertion?
| expect(renderingOptions.hits).toEqual([ | ||
| { objectID: 'a' }, | ||
| { objectID: 'b' }, | ||
| ]); |
There was a problem hiding this comment.
We can use the same assertion rather than a different one.
expect(rendering).toHaveBeenLastCalledWith(
expect.objectContaining({
hits: [{ objectID: "a" }, { objectID: "b" }],
}),
false
);| expect(rendering).toHaveBeenLastCalledWith( | ||
| expect.objectContaining({ | ||
| hits: [{ objectID: 'a' }, { objectID: 'b' }], | ||
| results: expect.any(Object), |
There was a problem hiding this comment.
We don't need to assert this property it's not relevant to the test.
There was a problem hiding this comment.
I wasn't sure what was the good strategy, that's why it's a bit inconsistent 😅
|
Did the changes @samouss :) |
When there is a slow network, the stalled search mechanics will trigger a new render with the same content. This has generally no impact. However the infinite builds up a list of all the results, and there was no mechanics in place to check if the same page has been already received.
To test, you can use dev-novel and with the performance set to slow 3g, you can see that there are no duplicates.
Fix #2928
Might also be a patch for #2938 (need confirmation)