Table functions are run before errors could be found during query analysis

[alexey-milovidov]

Use case

milovidov-pc :) SELECT trace, count() FROM url('https://s3.amazonaws.com/clickhouse-test-reports/60468/92ebf3d5becdf7af2bc17e3afb479fd456f008c1/stress_test__debug_/trace_log.tsv.zst') WHERE trace_type = 'Real'

SELECT
    trace,
    count()
FROM url('https://s3.amazonaws.com/clickhouse-test-reports/60468/92ebf3d5becdf7af2bc17e3afb479fd456f008c1/stress_test__debug_/trace_log.tsv.zst')
WHERE trace_type = 'Real'

Query id: eb093bad-9439-425e-949c-6e74f74e40db


Elapsed: 1.039 sec. 

Received exception:
Code: 215. DB::Exception: Column trace is not under aggregate function and not in GROUP BY keys. In query SELECT trace, count() FROM url('https://s3.amazonaws.com/clickhouse-test-reports/60468/92ebf3d5becdf7af2bc17e3afb479fd456f008c1/stress_test__debug_/trace_log.tsv.zst') WHERE trace_type = 'Real'. (NOT_AN_AGGREGATE)

Wasted one second.

[JackyWoo]

hi alexey, I wandor whether the elapsed 1 second is caused by reading schema of the tsv file but not real execution.

[alexey-milovidov]

Yes, very likely.

[JackyWoo]

I debugged the sql and found that there is only schema reading (actually 2 rows ) but no execution in query analysis stage.