topic Time travel on views in Data Engineering

Time travel on views

NathanE — Fri, 17 Nov 2023 09:00:47 GMT

Hello,

At my company, we design an application to analyze data, and we can do so on top of external databases such as Databricks. Our application cache some data in-memory and to avoid synchronization issues with the data on Databricks, we rely heavily on the time travel feature, which works very well !

However, time travel on views (and materialized views) is not an available feature, and I would like to know if it is something planned.

There are some technical implementation details that would make that feature perfect for our use case: we want to be able to time travel on views, either through the view version (just like for delta table currently), or through the versions of the tables that are part of the view.

Indeed, our application, can use a view just like a table, and so we would like to time travel on it like for tables. But sometimes, our application use a view and is aware it is a view of other tables (the application routes query to the view or the tables base on the granularity of the request). In that case, we would like to time travel to a specific version matching the specific versions of the underlying tables.

Regards,

Re: Time travel on views

karthik_p — Fri, 17 Nov 2023 15:00:45 GMT

@NathanE As you said, based on below article it may not support currenlty https://docs.databricks.com/en/sql/user/materialized-views.html, but at the same time looks as Materialized View is built on top of table and It is synchronous operation ( when backend table changes version then you can create refresh of u r MV, but direct way is not possible i think) , seems near future we need to see databricks may come up with different approach to handle time travel for views

Re: Time travel on views

robert1213 — Thu, 22 Jan 2026 18:40:41 GMT

Hi there,

Your use case for time travel on views is really interesting. I can see why being able to track historical versions of both views and their underlying tables would be crucial for an application that relies on caching and granular queries. Right now, as you mentioned, this feature isn’t natively supported in most systems, but it’s definitely an area where many analytics platforms are evolving.

On a slightly different note, I was recently exploring nz-tourism datasets for a project, and I found that having proper time-travel support could be a game-changer for analyzing trends over time—like visitor numbers or seasonal patterns—without worrying about the underlying changes in data sources. It really shows how valuable this feature could be, not just for enterprise applications but also for analytics on evolving datasets.

Hopefully, the platform roadmap includes this enhancement soon, as it would make historical analysis and reproducibility much more seamless.

Cheers!