This is an automated email from the ASF dual-hosted git repository.

hanahmily pushed a commit to branch vectorized-query
in repository https://gitbox.apache.org/repos/asf/skywalking-banyandb.git

commit 0b079e752bfd07ca3bde2edc662c8027d12b5bf3
Author: Hongtao Gao <[email protected]>
AuthorDate: Wed May 13 23:23:00 2026 +0000

    feat(banyand/measure): plumb --measure-vectorized-aggregation-enabled CLI 
flag (G8d.2)
    
    Wire the VectorizedConfig.AggregationEnabled field (added in commit
    ad994165) through bindVectorizedFlags so the gate is reachable from
    the standalone and data-node command lines:
    
      --measure-vectorized-aggregation-enabled=true
    
    Default false keeps GroupBy+Agg routing through the row path in
    production; integration tests and operators flip it explicitly once
    the vec aggregation egress matches row-path output byte-for-byte.
    Help text references the egress-parity verification gate so the next
    operator sees the rationale without needing to read the commit log.
    
    No behavior change for clusters that don't pass the flag.
---
 banyand/measure/measure.go | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/banyand/measure/measure.go b/banyand/measure/measure.go
index a90df68ea..a797156cc 100644
--- a/banyand/measure/measure.go
+++ b/banyand/measure/measure.go
@@ -185,4 +185,8 @@ func bindVectorizedFlags(flagS *run.FlagSet, cfg 
*vmeasure.VectorizedConfig) {
                "row count per vectorized batch")
        flagS.IntVar(&cfg.QueryMemoryMiB, 
"measure-vectorized-query-memory-mib", defaults.QueryMemoryMiB,
                "per-query memory budget for the vectorized path, in MiB")
+       flagS.BoolVar(&cfg.AggregationEnabled, 
"measure-vectorized-aggregation-enabled",
+               defaults.AggregationEnabled,
+               "route GroupBy+Agg requests through the vectorized aggregation 
pipeline "+
+                       "instead of the row-path aggregator (default false; 
flip after egress parity is verified)")
 }

Reply via email to