Adrian Kosowski

Hey @ilyanoskov, thanks for the additional context! Please disregard the discussion of "volatile" functions and triggers from my previous reply in that case. Please note that `AsyncTransformer` is called row...

@ilyanoskov Yes do keep us posted! I missed one point / question - whether, say, in year 7, you would just want data from year 7 to go into the...

> do you by any chance know, how to prevent the sliding window from sliding beyond available data? I did the duration for 10 years, and for 2017 the window...

Defining each window as a separate table (with groupby) is probably a good idea. As for joining it all back, it is convenient to use `ix_ref` - here is an...

@KamilPiechowiak Thanks for this fix. I'd say if we are down to ~25s single-threaded (and then ideally to under 5s on 8 worker threads), this can be considered reasonable timing...

> Unless there is a good way to pre-compute the large dataset of features (with Clickhouse or Polars) and then update it with Pathway incrementally in a streaming fashion? But...

This is a great point. Actually, `diff` is a particularly unpleasant (slow and memory-inefficient) operation to perform in a streaming or distributed system in its full generality - and may...

Comment: I'm bookmarking https://cloud.google.com/bigquery/docs/reference/standard-sql/navigation_functions to have a terminology reference for Topic for future participants of the discussion (taken from the non-streaming window world).

Dear @ilyanoskov, glad you are enjoying it! We are working to add additional distribution channels, @janchorowski ptal regarding the inclusion of conda-forge among them.

Thanks for bringing this up. Providing package extras that would allow e.g. `pip install pathway[no-xpacks]` with a minimum set of dependencies is on the roadmap. Keeping this one open until...