2 phase comparison query

[pjain1]

Split comparison time range query into two phase -

1. Run a groupby query on base time range
2. Run another query which joins the inline results from first query with groupby comparison query having added filter from base query

• This optimization is only possible when sorting by base column. (can extend to sort by comparison column but UI does not seem to use that anywhere)
• CTE rewrite is disabled by default now as it cause double scan of CTE in Druid, Duckdb and CH. It can be enabled but does not take effect if 2 phase optimization is done. It can be enabled to run queries on very high cardinality (~1M or greater) when sorting by other than base column which may fail in engines like Druid with ResourceLimitException.

Side note

• May be I can reuse the inline select concept for the null filling PR, also figured that druid support inline datasource with alternative syntax so that can be used and partial support with context keys can be removed

[pjain1]

Another approach is to do two separate topn queries for base and comparison time range and then a 3rd query with results from both queries to calculate measures and return the results, have a working code for this approach as well in a branch. However, this seems more cleaner and does only 2 queries.