sql: use new batch limit to limit all scans #5214
Conversation
|
Why does the Reviewed 1 of 1 files at r1. Comments from the review on Reviewable.io |
|
We will retrieve 50 rows but that will not be enough to return 50 results because of the Review status: 1 of 3 files reviewed at latest revision, all discussions resolved, all commit checks successful. Comments from the review on Reviewable.io |
The benchmarks use a table with 10 * 1000 rows. The query involves 4 spans of 1000 rows each, with the filter passing on only the first 10,30,50,70 rows respectively.
We change the kvFetcher code to use the new per-batch limit instead of limiting only single-span scans. The scan_test code is modified to work with multiple spans (previous code assumed OR branches would result in spans but that's not the case) ``` name old time/op new time/op delta Scan10000FilterLimit1_Cockroach-4 10.1ms ±51% 0.3ms ±23% -97.43% (p=0.000 n=10+9) Scan10000FilterLimit10_Cockroach-4 9.07ms ±17% 0.27ms ± 2% -97.05% (p=0.000 n=9+9) Scan10000FilterLimit50_Cockroach-4 10.5ms ± 8% 10.8ms ± 9% ~ (p=0.211 n=10+9) ```
|
Ah, that makes sense. Review status: 1 of 3 files reviewed at latest revision, all discussions resolved, all commit checks successful. Comments from the review on Reviewable.io |
RaduBerinde
force-pushed
the
limit-multi-span
branch
from
d4c776a to
22e9b7d
Compare
sql: use new batch limit to limit all scans
We change the kvFetcher code to use the new per-batch limit instead of limiting
only single-span scans.
The scan_test code is modified to work with multiple spans (previous code
assumed OR branches would result in spans but that's not the case)
The benchmarks use a table with 10 * 1000 rows. The query involves 4 spans of
1000 rows each, with the filter passing on only the first 10,30,50,70 rows
respectively.
This change is