User account reset when switching to history

[antho1404]

On the detail page of an NFT, if we switch between tabs bid/history, the information of the creator/owner are reset

Ex: https://dev.uaan.liteflow.com/tokens/97-0xcfbb45ff528c8ff6542887f631a1f704b92cf7db-34081265737503256337038727118962159868999502952850214258663667513124745213751?filter=history

Screen.Recording.2023-07-07.at.23.16.03.mov

[ismailToyran]

Just for context, this is Apollo cache issue.

    cache: new InMemoryCache({
      typePolicies: {
        Account: {
          keyFields: ['address'],
        },
        Asset: {
          keyFields: ['chainId', 'collectionAddress', 'tokenId'],
        },
        Collection: {
          keyFields: ['chainId', 'address'],
        },
      },
    }).restore(windowApolloState || {}),

[ismailToyran]

@antho1404 I remember having this issue on admin as well. And it's so hard to find an example that causes an issue. For example your account doesn't trigger this issue anymore. We could close this issue if you check and confirm?

But again, as this occurs very rarely, even if we close this, this might be an ongoing issue. On admin, to fix this, I had to remove Account from typePolicies as cache was causing this issue strangely.

[NicolasMahe]

Should we even use the typePolicies of the Apollo cache?
Apollo cache works out of the box with queries' result, no need for typePolicies.

This typePolices is to enable a level of cache where provided type are extracted from queries' result and are cached and shared across the app. This cache is useful to avoid executing a refresh query when a mutation occurs for example, but we build the starter-kit without this in mind.

[antho1404]

I would be food to close this issue as we can't reproduce anymore. If the apollo policies are an issue that could be done as a different issue

[EmmanuelDrouin]

@antho1404 @ismailToyran are you good to close this issue? The last comment from October stated that it is very hard to reproduce and you were ready to close the issue

[ismailToyran]

We don't have the cache implementation that was causing this issue anymore. So I'm pretty sure we can close this issue until we can produce it in the future somehow.