This document records interesting code that we've deleted for the sake of discoverability for the future.
Interesting bits: the Heavy Users job used a custom Paritioner called ConsistentPartitioner
that optimized for copartitioning the same client_ids together even as the client_ids grow and wane.
This dataset was created as a one-off for the purposes of the Online News pioneer study. It created sessions that measured dwell time on a tld based on logs sent from users. It used a state machine to create the sessions, which is mildly interesting.
This dataset was created to count crashes on a daily basis, before we introduced error aggregates.
This dataset was created to monitor that various metrics conformed to the Quantum release criteria expectations.
This dataset was used to access crash pings, before we introduced the crash ping table in BigQuery.
These datasets were created to count clients on a daily and monthly basis, before we introduced clients last seen.
This dataset was used to access all histograms for a 1% sample of clients, before we introduced the main ping table in BigQuery.
This dataset was used for experiment analysis, before it was deprecated in Bug 1515134.
These jobs were reimplemented as BigQuery SQL in bigquery-etl/sql/telemetry_derived/.
This job was reimplemented as BigQuery SQL in bigquery-etl/sql/telemetry_derived/experiments_v1/query.sql.