🐛 correct tracking of producer views

[lucasrodes]

Fixes #3855

Producer analytics does not accurately count the views of data from a given producer. It might be over estimating these.

This PR drops the usage of VersionTracker for regular queries to our database. There are two benefits:

• Numbers will be more accurate
• Performance is better. Loading the steps dataframe is significantly slower.

[owidbot]

Quick links (staging server):

Site Dev	Site Preview	Admin	Wizard	Docs

Login: ssh owid@staging-site-producer-analytics-bug-views

chart-diff: ✅

No charts for review.

data-diff: ✅ No differences found

Legend: +New  ~Modified  -Removed  =Identical  Details
Hint: Run this locally with etl diff REMOTE data/ --include yourdataset --verbose --snippet

Automatically updated datasets matching weekly_wildfires|excess_mortality|covid|fluid|flunet|country_profile|garden/ihme_gbd/2019/gbd_risk are not included

Edited: 2025-01-16 17:21:06 UTC
Execution time: 14.56 seconds

[lucasrodes]

Update: the bug might be coming from VersionTracker.steps_df. In particular, from the list "all_chart_slugs".

Example:

df = VersionTracker(exclude_steps=[]).steps_df
df.loc[df.step.str.contains("grapher/gcp/2024-11-21/global_carbon_budget"), "all_chart_slugs"].item()

[lucasrodes]

On my computer, using regular queries to DB improves performance significantly.

Old code (with VersionTracker): Took ~8 seconds.
New code (regular queries): Takes ~1 second.