Fixed #34580 -- Avoided unnecessary computation of selected expressions in SQLCompiler.

[smithdc1]

This is @charettes's commit not mine, so commit author will need changing if this works. See Ticket #34580

I'm just opening this to see if the benchmark script can see a performance gain (it's about a 20%-30% improvement locally).

       before           after         ratio
     [ca5d3c99]       [2aa6996a]
     <form_docs>       <SQLCompiler>
-     1.81±0.01ms         1.43±0ms     0.79  query_benchmarks.query_exists.benchmark.QueryExists.time_query_exists
-         876±7μs          625±8μs     0.71  query_benchmarks.query_aggregate.benchmark.QueryAggr.time_aggregate
-     1.30±0.01ms          914±9μs     0.70  query_benchmarks.query_count.benchmark.QueryCount.time_query_count

[charettes]

Thanks David, didn't know we had that benchmark job!

[smithdc1]

No problem -- I'm not sure it's been used all that much so this was an interesting case to see if it is useful in practice.

It seems that running the workflow on github's infrastructure led to no significant results being found. However, running locally again I see some quite large benefits from this.

       before           after         ratio
     [ca5d3c99]       [2aa6996a]
     <form_docs>       <SQLCompiler>
-     1.83±0.01ms      1.41±0.01ms     0.77  query_benchmarks.query_exists.benchmark.QueryExists.time_query_exists
-         874±2μs          621±8μs     0.71  query_benchmarks.query_aggregate.benchmark.QueryAggr.time_aggregate
-     1.35±0.02ms          905±4μs     0.67  query_benchmarks.query_count.benchmark.QueryCount.time_query_count

[charettes]

It seems like a valuable change so I'll accept the ticket but I'm not surprised that we get weird results on Github CI; any form of virtualized / containerized process is doomed to provide inaccurate results.

[felixxm]

Thanks both 👍

[felixxm]

I added a release note.