oso
oso copied to clipboard
Try Dagster Multi-assets to handle daily/weekly/monthly aggregations
What is it?
https://docs.dagster.io/concepts/assets/multi-assets
Let's remember to version any marts generated by this as _v1
Here's the current thought on what we want to do here:
- Right now we aggregate and count different event-types on different bucket sizes.
- In reality, we want to have different metric_types. For example, instead of having star event aggregations, we have a metric called
star_count_1_dayorstar_count_1_weeketc. This is in effect the star count from the last 1 day/week/month - We want to be able to have time-series data for any metric. So the table schema would look like
timestamp,project_id,metric_type,amount. - We can then use Dagster multi-assets to auto-generate multiple related metrics.
This isn't a requirement for RF4 so going to push this till after RF4.
Deprecating in favor of using either Cube.dev or sqlmesh