This is an automated email from the ASF dual-hosted git repository.
github-bot pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/datafusion.git
The following commit(s) were added to refs/heads/main by this push:
new 4ecccde087 feat: Add `output_bytes` to baseline metrics (#18268)
4ecccde087 is described below
commit 4ecccde0877ad859451e791bdc47f928e0783057
Author: Yongting You <[email protected]>
AuthorDate: Mon Oct 27 12:45:21 2025 +0800
feat: Add `output_bytes` to baseline metrics (#18268)
## Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and
enhancements and this helps us generate change logs for our releases.
You can link an issue to this PR using the GitHub syntax. For example
`Closes #123` indicates that this PR will close issue #123.
-->
- Closes https://github.com/apache/datafusion/issues/16244
## Rationale for this change
<!--
Why are you proposing this change? If this is already explained clearly
in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand
your changes and offer better suggestions for fixes.
-->
Support `output_bytes` in `BaselineMetrics` (a common metrics set for
almost all operators)
```
DataFusion CLI v50.3.0
> explain analyze select * from generate_series(1, 1000000) as t1(v1) order
by v1 desc;
+-------------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| plan_type | plan
|
+-------------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Plan with Metrics | SortExec: expr=[v1@0 DESC],
preserve_partitioning=[false], metrics=[output_rows=1000000,
elapsed_compute=96.421534ms, output_bytes=7.6 MB, spill_count=0,
spilled_bytes=0.0 B, spilled_rows=0, batches_split=0] |
| | ProjectionExec: expr=[value@0 as v1],
metrics=[output_rows=1000000, elapsed_compute=34.125µs, output_bytes=7.7 MB]
|
| | LazyMemoryExec: partitions=1,
batch_generators=[generate_series: start=1, end=1000000, batch_size=8192],
metrics=[output_rows=1000000, elapsed_compute=2.262626ms, output_bytes=7.7 MB]
|
| |
|
+-------------------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
1 row(s) fetched.
Elapsed 0.080 seconds.
```
Note it might overestimate memory due to a well-known issue. See the PR
snippet for details
```rs
/// Memory usage of all output batches.
///
/// Note: This value may be overestimated. If multiple output
`RecordBatch`
/// instances share underlying memory buffers, their sizes will be
counted
/// multiple times.
/// Issue: <https://github.com/apache/datafusion/issues/16841>
output_bytes: Count,
```
I think this metric provides valuable insight, so it's better for it to
overestimate than not exist at all.
## What changes are included in this PR?
<!--
There is no need to duplicate the description in the issue here but it
is sometimes worth providing a summary of the individual changes in this
PR.
-->
1. Add `output_bytes` to `BaselineMetrics`, and it's set to `summary`
analyze level. (see config `datafusion.explain.analyze_level` for
details)
2. This metrics will be automatically tracked through `record_poll()`
API, which is a common interface most operators uses when a new output
batch is generated.
## Are these changes tested?
UT
<!--
We typically require tests for all PRs in order to:
1. Prevent the code from being accidentally broken by subsequent changes
3. Serve as another way to document the expected behavior of the code
If tests are not included in your PR, please explain why (for example,
are they covered by existing tests)?
-->
## Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be
updated before approving the PR.
-->
<!--
If there are any breaking changes to public APIs, please add the `api
change` label.
-->
---
datafusion/core/tests/sql/explain_analyze.rs | 23 ++++++++++++++
datafusion/physical-plan/src/metrics/baseline.rs | 20 +++++++++++++
datafusion/physical-plan/src/metrics/builder.rs | 8 +++++
datafusion/physical-plan/src/metrics/mod.rs | 1 +
datafusion/physical-plan/src/metrics/value.rs | 38 ++++++++++++++++--------
docs/source/user-guide/metrics.md | 9 +++---
6 files changed, 82 insertions(+), 17 deletions(-)
diff --git a/datafusion/core/tests/sql/explain_analyze.rs
b/datafusion/core/tests/sql/explain_analyze.rs
index 6d386cc456..43f79ead02 100644
--- a/datafusion/core/tests/sql/explain_analyze.rs
+++ b/datafusion/core/tests/sql/explain_analyze.rs
@@ -63,36 +63,59 @@ async fn explain_analyze_baseline_metrics() {
"AggregateExec: mode=Partial, gby=[]",
"metrics=[output_rows=3, elapsed_compute="
);
+ assert_metrics!(
+ &formatted,
+ "AggregateExec: mode=Partial, gby=[]",
+ "output_bytes="
+ );
assert_metrics!(
&formatted,
"AggregateExec: mode=FinalPartitioned, gby=[c1@0 as c1]",
"metrics=[output_rows=5, elapsed_compute="
);
+ assert_metrics!(
+ &formatted,
+ "AggregateExec: mode=FinalPartitioned, gby=[c1@0 as c1]",
+ "output_bytes="
+ );
assert_metrics!(
&formatted,
"FilterExec: c13@1 != C2GT5KVyOPZpgKVl110TyZO0NcJ434",
"metrics=[output_rows=99, elapsed_compute="
);
+ assert_metrics!(
+ &formatted,
+ "FilterExec: c13@1 != C2GT5KVyOPZpgKVl110TyZO0NcJ434",
+ "output_bytes="
+ );
assert_metrics!(
&formatted,
"ProjectionExec: expr=[]",
"metrics=[output_rows=5, elapsed_compute="
);
+ assert_metrics!(&formatted, "ProjectionExec: expr=[]", "output_bytes=");
assert_metrics!(
&formatted,
"CoalesceBatchesExec: target_batch_size=4096",
"metrics=[output_rows=5, elapsed_compute"
);
+ assert_metrics!(
+ &formatted,
+ "CoalesceBatchesExec: target_batch_size=4096",
+ "output_bytes="
+ );
assert_metrics!(
&formatted,
"UnionExec",
"metrics=[output_rows=3, elapsed_compute="
);
+ assert_metrics!(&formatted, "UnionExec", "output_bytes=");
assert_metrics!(
&formatted,
"WindowAggExec",
"metrics=[output_rows=1, elapsed_compute="
);
+ assert_metrics!(&formatted, "WindowAggExec", "output_bytes=");
fn expected_to_have_metrics(plan: &dyn ExecutionPlan) -> bool {
use datafusion::physical_plan;
diff --git a/datafusion/physical-plan/src/metrics/baseline.rs
b/datafusion/physical-plan/src/metrics/baseline.rs
index 45cef58b5d..858773b946 100644
--- a/datafusion/physical-plan/src/metrics/baseline.rs
+++ b/datafusion/physical-plan/src/metrics/baseline.rs
@@ -21,6 +21,8 @@ use std::task::Poll;
use arrow::record_batch::RecordBatch;
+use crate::spill::get_record_batch_memory_size;
+
use super::{Count, ExecutionPlanMetricsSet, MetricBuilder, Time, Timestamp};
use datafusion_common::Result;
@@ -53,6 +55,16 @@ pub struct BaselineMetrics {
/// output rows: the total output rows
output_rows: Count,
+
+ /// Memory usage of all output batches.
+ ///
+ /// Note: This value may be overestimated. If multiple output `RecordBatch`
+ /// instances share underlying memory buffers, their sizes will be counted
+ /// multiple times.
+ /// Issue: <https://github.com/apache/datafusion/issues/16841>
+ output_bytes: Count,
+ // Remember to update `docs/source/user-guide/metrics.md` when updating
comments
+ // or adding new metrics
}
impl BaselineMetrics {
@@ -71,6 +83,9 @@ impl BaselineMetrics {
output_rows: MetricBuilder::new(metrics)
.with_type(super::MetricType::SUMMARY)
.output_rows(partition),
+ output_bytes: MetricBuilder::new(metrics)
+ .with_type(super::MetricType::SUMMARY)
+ .output_bytes(partition),
}
}
@@ -84,6 +99,7 @@ impl BaselineMetrics {
end_time: Default::default(),
elapsed_compute: self.elapsed_compute.clone(),
output_rows: Default::default(),
+ output_bytes: Default::default(),
}
}
@@ -211,6 +227,8 @@ impl RecordOutput for usize {
impl RecordOutput for RecordBatch {
fn record_output(self, bm: &BaselineMetrics) -> Self {
bm.record_output(self.num_rows());
+ let n_bytes = get_record_batch_memory_size(&self);
+ bm.output_bytes.add(n_bytes);
self
}
}
@@ -218,6 +236,8 @@ impl RecordOutput for RecordBatch {
impl RecordOutput for &RecordBatch {
fn record_output(self, bm: &BaselineMetrics) -> Self {
bm.record_output(self.num_rows());
+ let n_bytes = get_record_batch_memory_size(self);
+ bm.output_bytes.add(n_bytes);
self
}
}
diff --git a/datafusion/physical-plan/src/metrics/builder.rs
b/datafusion/physical-plan/src/metrics/builder.rs
index 74ba5a2a18..88ec1a3f67 100644
--- a/datafusion/physical-plan/src/metrics/builder.rs
+++ b/datafusion/physical-plan/src/metrics/builder.rs
@@ -151,6 +151,14 @@ impl<'a> MetricBuilder<'a> {
count
}
+ /// Consume self and create a new counter for recording total output bytes
+ pub fn output_bytes(self, partition: usize) -> Count {
+ let count = Count::new();
+ self.with_partition(partition)
+ .build(MetricValue::OutputBytes(count.clone()));
+ count
+ }
+
/// Consume self and create a new gauge for reporting current memory usage
pub fn mem_used(self, partition: usize) -> Gauge {
let gauge = Gauge::new();
diff --git a/datafusion/physical-plan/src/metrics/mod.rs
b/datafusion/physical-plan/src/metrics/mod.rs
index 0fd7bfb8c8..02aad6eb60 100644
--- a/datafusion/physical-plan/src/metrics/mod.rs
+++ b/datafusion/physical-plan/src/metrics/mod.rs
@@ -296,6 +296,7 @@ impl MetricsSet {
MetricValue::ElapsedCompute(_) => false,
MetricValue::SpillCount(_) => false,
MetricValue::SpilledBytes(_) => false,
+ MetricValue::OutputBytes(_) => false,
MetricValue::SpilledRows(_) => false,
MetricValue::CurrentMemoryUsage(_) => false,
MetricValue::Gauge { name, .. } => name == metric_name,
diff --git a/datafusion/physical-plan/src/metrics/value.rs
b/datafusion/physical-plan/src/metrics/value.rs
index 3149fca95b..fc94793550 100644
--- a/datafusion/physical-plan/src/metrics/value.rs
+++ b/datafusion/physical-plan/src/metrics/value.rs
@@ -395,6 +395,8 @@ pub enum MetricValue {
SpillCount(Count),
/// Total size of spilled bytes produced: "spilled_bytes" metric
SpilledBytes(Count),
+ /// Total size of output bytes produced: "output_bytes" metric
+ OutputBytes(Count),
/// Total size of spilled rows produced: "spilled_rows" metric
SpilledRows(Count),
/// Current memory used
@@ -449,6 +451,9 @@ impl PartialEq for MetricValue {
(MetricValue::SpilledBytes(count),
MetricValue::SpilledBytes(other)) => {
count == other
}
+ (MetricValue::OutputBytes(count), MetricValue::OutputBytes(other))
=> {
+ count == other
+ }
(MetricValue::SpilledRows(count), MetricValue::SpilledRows(other))
=> {
count == other
}
@@ -505,6 +510,7 @@ impl MetricValue {
Self::OutputRows(_) => "output_rows",
Self::SpillCount(_) => "spill_count",
Self::SpilledBytes(_) => "spilled_bytes",
+ Self::OutputBytes(_) => "output_bytes",
Self::SpilledRows(_) => "spilled_rows",
Self::CurrentMemoryUsage(_) => "mem_used",
Self::ElapsedCompute(_) => "elapsed_compute",
@@ -523,6 +529,7 @@ impl MetricValue {
Self::OutputRows(count) => count.value(),
Self::SpillCount(count) => count.value(),
Self::SpilledBytes(bytes) => bytes.value(),
+ Self::OutputBytes(bytes) => bytes.value(),
Self::SpilledRows(count) => count.value(),
Self::CurrentMemoryUsage(used) => used.value(),
Self::ElapsedCompute(time) => time.value(),
@@ -550,6 +557,7 @@ impl MetricValue {
Self::OutputRows(_) => Self::OutputRows(Count::new()),
Self::SpillCount(_) => Self::SpillCount(Count::new()),
Self::SpilledBytes(_) => Self::SpilledBytes(Count::new()),
+ Self::OutputBytes(_) => Self::OutputBytes(Count::new()),
Self::SpilledRows(_) => Self::SpilledRows(Count::new()),
Self::CurrentMemoryUsage(_) =>
Self::CurrentMemoryUsage(Gauge::new()),
Self::ElapsedCompute(_) => Self::ElapsedCompute(Time::new()),
@@ -588,6 +596,7 @@ impl MetricValue {
(Self::OutputRows(count), Self::OutputRows(other_count))
| (Self::SpillCount(count), Self::SpillCount(other_count))
| (Self::SpilledBytes(count), Self::SpilledBytes(other_count))
+ | (Self::OutputBytes(count), Self::OutputBytes(other_count))
| (Self::SpilledRows(count), Self::SpilledRows(other_count))
| (
Self::Count { count, .. },
@@ -638,18 +647,21 @@ impl MetricValue {
/// numbers are "more useful" (and displayed first)
pub fn display_sort_key(&self) -> u8 {
match self {
- Self::OutputRows(_) => 0, // show first
- Self::ElapsedCompute(_) => 1, // show second
- Self::SpillCount(_) => 2,
- Self::SpilledBytes(_) => 3,
- Self::SpilledRows(_) => 4,
- Self::CurrentMemoryUsage(_) => 5,
- Self::Count { .. } => 6,
- Self::Gauge { .. } => 7,
- Self::Time { .. } => 8,
- Self::StartTimestamp(_) => 9, // show timestamps last
- Self::EndTimestamp(_) => 10,
- Self::Custom { .. } => 11,
+ // `BaselineMetrics` that is common for most operators
+ Self::OutputRows(_) => 0,
+ Self::ElapsedCompute(_) => 1,
+ Self::OutputBytes(_) => 2,
+ // Other metrics
+ Self::SpillCount(_) => 3,
+ Self::SpilledBytes(_) => 4,
+ Self::SpilledRows(_) => 5,
+ Self::CurrentMemoryUsage(_) => 6,
+ Self::Count { .. } => 7,
+ Self::Gauge { .. } => 8,
+ Self::Time { .. } => 9,
+ Self::StartTimestamp(_) => 10, // show timestamps last
+ Self::EndTimestamp(_) => 11,
+ Self::Custom { .. } => 12,
}
}
@@ -669,7 +681,7 @@ impl Display for MetricValue {
| Self::Count { count, .. } => {
write!(f, "{count}")
}
- Self::SpilledBytes(count) => {
+ Self::SpilledBytes(count) | Self::OutputBytes(count) => {
let readable_count = human_readable_size(count.value());
write!(f, "{readable_count}")
}
diff --git a/docs/source/user-guide/metrics.md
b/docs/source/user-guide/metrics.md
index f2634b9015..1fb2f4a5c7 100644
--- a/docs/source/user-guide/metrics.md
+++ b/docs/source/user-guide/metrics.md
@@ -27,10 +27,11 @@ DataFusion operators expose runtime metrics so you can
understand where time is
`BaselineMetrics` are available in most physical operators to capture common
measurements.
-| Metric | Description |
-| --------------- | ------------------------------------------------------ |
-| elapsed_compute | CPU time the operator actively spends processing work. |
-| output_rows | Total number of rows the operator produces. |
+| Metric | Description
|
+| --------------- |
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
|
+| elapsed_compute | CPU time the operator actively spends processing work.
|
+| output_rows | Total number of rows the operator produces.
|
+| output_bytes | Memory usage of all output batches. Note: This value may
be overestimated. If multiple output `RecordBatch` instances share underlying
memory buffers, their sizes will be counted multiple times. |
## Operator-specific Metrics
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
