You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We are diagnosing performance issues with the ClickBench benchmark and identified this inefficiency. It is unlikely to be material whilst on a single thread executor, but when multithreaded, the cost of the additional call will increase.
This plan has two redundant projections, after the scan we only have the URL column so the two projections aren't doing anything
└─ EXIT
└─ PROJECT (URL)
└─ SUBQUERY AS a
└─ LIMIT (1000000)
└─ PROJECT (URL)
└─ FILTER (URL LIKE '%google%')
└─ SCAN (hits) [URL]
The text was updated successfully, but these errors were encountered:
In trying to address this a bug was found in the projection pushdown strategy - it changes the type of the columns attribute in the Scan, SubQuery and Union operators to be lists of FlatColumns rather than LogicalColumns.
Fixing this is likely to have knock-on impacts to these operators and potentially to other optimization strategies.
joocer
changed the title
✨ Remove redundant projections
✨ [ClickBench] Remove redundant projections
May 29, 2024
We are diagnosing performance issues with the ClickBench benchmark and identified this inefficiency. It is unlikely to be material whilst on a single thread executor, but when multithreaded, the cost of the additional call will increase.
This plan has two redundant projections, after the scan we only have the URL column so the two projections aren't doing anything
The text was updated successfully, but these errors were encountered: