Improve performance of filtered distinct symbol query

Is your feature request related to a problem?

A query like

SELECT DISTINCT node_uid FROM request_logs

runs fast because it uses the symbol table (metadata) and doesn't have to scan the whole node_id table. However, if we add a filter:

SELECT DISTINCT node_uid FROM request_logs WHERE node_id != NULL

the query enters a full table scan, filtering out NULLs.

Describe the solution you'd like.

We can rewrite the query to

SELECT * FROM (SELECT DISTINCT node_uid FROM request_logs) WHERE node_uid != NULL

This would do the same work as the first query above, and then filter out a possible single NULL row.

Describe alternatives you've considered.

No response

Full Name:

Marko Topolnik

Affiliation:

QuestDB

Additional context

No response

No files or tests are named. Start by locating the planner or execution path for SELECT DISTINCT with a WHERE filter, then compare it with the symbol-table path used by the unfiltered query. Reproduce both SQL examples and verify that the filtered query avoids a full node_id table scan while excluding NULL results.