Make TTL optional in FeatureViews

Is your feature request related to a problem? Please describe.

Some of our data are features likes created_at of a shop. This data can be very old even though the shop is still active. Therefore we would rather want to not have a TTL rather than specifying a crazy value like 9999999

There are definitely performance implications that we are aware of 👌 If we take the above example, created_at of a given shop, the dataset grows only when there is a new shop created. Therefore the “lifetime” dataset size remains just all unique shops. So here the performance implications are pretty limited. So the need is there, we just need to use TLL properly for each FeatureView 👌

Describe the solution you’d like A clear and concise description of what you want to happen. Make TTL mandatory but add the possibility to pass None or 0 to explain that there is no TTL during historical retrieval

Describe alternatives you’ve considered A clear and concise description of any alternative solutions or features you’ve considered. Add a very large value like 99999999 as a TTL

Additional context Add any other context or screenshots about the feature request here.

[jparthasarthy]

@woop I just really dislike the idea that someone could miss the TTL parameter and instead of throwing an error we would set it to infinity.

I think setting it to infinity should be something that has to be done explicitly by the user. Is there an API that makes sense with that in mind? Do you agree?

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.