Cache "how it works"

[piotrblasiak]

It is quite important to me to know how the cache works in order to be able to stop the client from evicting data when in offline mode. I´m using react-apollo but I assume this will be important for all clients. Is there such a document/guide/thread already that I have missed?

[stubailo]

Currently no data is evicted, but we're working on some refactors to make the implementation more flexible and swappable.

For example, you'll be able to replace the redux cache with a different backend, etc.

Do you have any particular questions I can answer? We have a bit of info about how stuff works and there is also the devtools, but I'm not sure exactly what you're asking about.

[piotrblasiak]

Oh, so once an entity is loaded it stays in the cache forever? My use case, which I think should be pretty common is that I load something like this:

@graphql(gql`{
query($projectId: ID!) {
  project(id: $projectId) {
    childEntities {...}
  }
}`

And I want the data that is referenced by that component to be available to it for as long as it needs it. But once the $projectId changes the previously loaded data may not be needed again for quite some time if ever. I would expect some garbage collection to happen, but you´re saying there is nothing like that currently? If so, how do you stop an application using apollo from using up a ton of memory in the end?

[stubailo]

If so, how do you stop an application using apollo from using up a ton of memory in the end?

You can call resetStore periodically to avoid this, but we're working on cache improvements that will make this no longer the case. First step is decoupling the store from the core so that it's easy to plug in new cache implementations with different features, and offline is the primary motivating use case for this new approach.

See here: apollographql/apollo-client#1432 (comment)