Merge #61948 #62166 #62287 #62883 #62922 #62925 #62926 #62962

61948: roachtest/cdc: update status when watching changefeed latency r=andreimatei a=andreimatei

When watching the latency, we were logging every second. This patch
reduces the frequency to 1/10s and makes the test also update its status
so that the current latency high-water mark is visible in the web UI.

Release note: None

62166: sql: add roach test / workload for testing connection latencies r=rafiss a=RichardJCai

Add roach test and workload test for testing connection latencies.

Release note: None

Making it a private test right now (not available through `cockroach workload`)

Need to also add it to teamcity and make sure the results are reported in roachperf, will open PR there.

Resolves #59394


62287: cdc: Improve avro encoder performance.  r=miretskiy a=miretskiy

Add avro encoder benchmarks.

Avoid expensive allocations (maps) when encoding datums.
Improve encoder performance by ~40%, and significantly reduce
memory allocations per op.

Release Notes: None

62883: sqlsmith: add support for ALTER TYPE ... DROP VALUE r=rafiss a=arulajmani

Now that `ALTER TYPE ... DROP VALUE` is a thing, let's add it to SQL
Smith.

Closes #62812

Release note: None

62922: roachtest/tpccbench: don't require workload binary r=andreimatei a=andreimatei

tpccbench was copying around the `workload` binary, but it then doesn't
use it; it uses `cockroach workload ...`. This patch removes the
copying.

Release note: None

62925: colexec: optimize CASE, AND, and OR projection operators r=yuzefovich a=yuzefovich

Previously, in CASE, AND, and OR projection operators we would always
allocate internal selection vectors of maximum capacity in the
constructor. This is inefficient when the query processes only small
data sets. This commit makes the allocations lazy and of tight sizes
based on the batch lengths.

Addresses: #62212

Release note: None

62926: bench,sql: add logScope to some benchmarks r=yuzefovich a=yuzefovich

This commits adds `defer log.Scope(b).Close(b)` to some benchmarks in
`bench`, `colexec`, and `rowexec` packages. Note that in some cases this
might not matter, but I didn't think through each carefully since it
shouldn't do any harm.

Release note: None

62962: sql: fix bug in close of ieSyncResultChannel r=yuzefovich a=ajwerner

When finishing sending on the ieSyncResultChannel we do not need to and should
not close the channel used to unblock the sender. Doing so can result in a
panic if the reader is concurrently trying to unblock the sender. We could
close that channel but there's no obvious reason to.

Fixes #62939

Release note: None

Co-authored-by: Andrei Matei <[email protected]>
Co-authored-by: richardjcai <[email protected]>
Co-authored-by: Yevgeniy Miretskiy <[email protected]>
Co-authored-by: arulajmani <[email protected]>
Co-authored-by: Yahor Yuzefovich <[email protected]>
Co-authored-by: Andrew Werner <[email protected]>

pkg/bench/bench_test.go

@@ -1119,6 +1119,7 @@ func runBenchmarkWideTable(b *testing.B, db *sqlutils.SQLRunner, count int, bigC
 // BenchmarkVecSkipScan benchmarks the vectorized engine's performance
 // when skipping unneeded key values in the decoding process.
 func BenchmarkVecSkipScan(b *testing.B) {
+	defer log.Scope(b).Close(b)
 	benchmarkCockroach(b, func(b *testing.B, db *sqlutils.SQLRunner) {
 		create := `
 CREATE TABLE bench.scan(
@@ -1183,6 +1184,7 @@ func BenchmarkWideTableIgnoreColumns(b *testing.B) {
 // benchmark (and get memory allocation statistics for) the planning process.
 func BenchmarkPlanning(b *testing.B) {
 	skip.UnderShort(b)
+	defer log.Scope(b).Close(b)
 	ForEachDB(b, func(b *testing.B, db *sqlutils.SQLRunner) {
 		db.Exec(b, `CREATE TABLE abc (a INT PRIMARY KEY, b INT, c INT, INDEX(b), UNIQUE INDEX(c))`)
 

pkg/ccl/changefeedccl/BUILD.bazel

@@ -176,6 +176,7 @@ go_test(
         "//pkg/util/span",
         "//pkg/util/syncutil",
         "//pkg/util/timeutil",
+        "//pkg/util/timeutil/pgdate",
         "//pkg/util/uuid",
         "//pkg/workload",
         "//pkg/workload/bank",