Skip to content

perf(reader): Pass data file size and delete file size to reader to avoid stat() calls #2175

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
blackmwk merged 5 commits into from
Feb 26, 2026
Merged

perf(reader): Pass data file size and delete file size to reader to avoid stat() calls #2175

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
de3e543
Extract changse from reader_perf.
mbutrovich Feb 24, 2026
981ce7c
Restore accidentally removed comments.
mbutrovich Feb 24, 2026
f558510
Clean up comments.
mbutrovich Feb 24, 2026
422c9d2
cargo fmt
mbutrovich Feb 25, 2026
9ac2382
update tests to use real file sizes
mbutrovich Feb 25, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 15 additions & 6 deletions crates/iceberg/src/arrow/caching_delete_file_loader.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -235,7 +235,7 @@ impl CachingDeleteFileLoader {
PosDelLoadAction::Load => Ok(DeleteFileContext::PosDels {
file_path: task.file_path.clone(),
stream: basic_delete_file_loader
.parquet_to_batch_stream(&task.file_path)
.parquet_to_batch_stream(&task.file_path, task.file_size_in_bytes)
.await?,
}),
}
Expand All @@ -254,7 +254,7 @@ impl CachingDeleteFileLoader {
let equality_ids_vec = task.equality_ids.clone().unwrap();
let evolved_stream = BasicDeleteFileLoader::evolve_schema(
basic_delete_file_loader
.parquet_to_batch_stream(&task.file_path)
.parquet_to_batch_stream(&task.file_path, task.file_size_in_bytes)
.await?,
schema,
&equality_ids_vec,
Expand Down Expand Up @@ -614,7 +614,10 @@ mod tests {

let basic_delete_file_loader = BasicDeleteFileLoader::new(file_io.clone());
let record_batch_stream = basic_delete_file_loader
.parquet_to_batch_stream(&eq_delete_file_path)
.parquet_to_batch_stream(
&eq_delete_file_path,
std::fs::metadata(&eq_delete_file_path).unwrap().len(),
)
.await
.expect("could not get batch stream");

Expand Down Expand Up @@ -811,7 +814,10 @@ mod tests {
let basic_delete_file_loader = BasicDeleteFileLoader::new(file_io.clone());

let batch_stream = basic_delete_file_loader
.parquet_to_batch_stream(&delete_file_path)
.parquet_to_batch_stream(
&delete_file_path,
std::fs::metadata(&delete_file_path).unwrap().len(),
)
.await
.unwrap();

Expand Down Expand Up @@ -913,20 +919,23 @@ mod tests {

// Create FileScanTask with BOTH positional and equality deletes
let pos_del = FileScanTaskDeleteFile {
file_path: pos_del_path,
file_path: pos_del_path.clone(),
file_size_in_bytes: std::fs::metadata(&pos_del_path).unwrap().len(),
file_type: DataContentType::PositionDeletes,
partition_spec_id: 0,
equality_ids: None,
};

let eq_del = FileScanTaskDeleteFile {
file_path: eq_delete_path.clone(),
file_size_in_bytes: std::fs::metadata(&eq_delete_path).unwrap().len(),
file_type: DataContentType::EqualityDeletes,
partition_spec_id: 0,
equality_ids: Some(vec![2, 3]), // Only use field IDs that exist in both schemas
};

let file_scan_task = FileScanTask {
file_size_in_bytes: 0,
start: 0,
length: 0,
record_count: None,
Expand Down Expand Up @@ -993,7 +1002,7 @@ mod tests {

let basic_delete_file_loader = BasicDeleteFileLoader::new(file_io.clone());
let record_batch_stream = basic_delete_file_loader
.parquet_to_batch_stream(&path)
.parquet_to_batch_stream(&path, std::fs::metadata(&path).unwrap().len())
.await
.expect("could not get batch stream");

Expand Down
6 changes: 5 additions & 1 deletion crates/iceberg/src/arrow/delete_file_loader.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,7 @@ impl BasicDeleteFileLoader {
pub(crate) async fn parquet_to_batch_stream(
&self,
data_file_path: &str,
file_size_in_bytes: u64,
) -> Result<ArrowRecordBatchStream> {
/*
Essentially a super-cut-down ArrowReader. We can't use ArrowReader directly
Expand All @@ -65,6 +66,7 @@ impl BasicDeleteFileLoader {
false,
None,
None,
file_size_in_bytes,
)
.await?
.build()?
Expand Down Expand Up @@ -102,7 +104,9 @@ impl DeleteFileLoader for BasicDeleteFileLoader {
task: &FileScanTaskDeleteFile,
schema: SchemaRef,
) -> Result<ArrowRecordBatchStream> {
let raw_batch_stream = self.parquet_to_batch_stream(&task.file_path).await?;
let raw_batch_stream = self
.parquet_to_batch_stream(&task.file_path, task.file_size_in_bytes)
.await?;

// For equality deletes, only evolve the equality_ids columns.
// For positional deletes (equality_ids is None), use all field IDs.
Expand Down
22 changes: 22 additions & 0 deletions crates/iceberg/src/arrow/delete_filter.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -378,27 +378,46 @@ pub(crate) mod tests {

let pos_del_1 = FileScanTaskDeleteFile {
file_path: format!("{}/pos-del-1.parquet", table_location.to_str().unwrap()),
file_size_in_bytes: std::fs::metadata(format!(
"{}/pos-del-1.parquet",
table_location.to_str().unwrap()
))
.unwrap()
.len(),
file_type: DataContentType::PositionDeletes,
partition_spec_id: 0,
equality_ids: None,
};

let pos_del_2 = FileScanTaskDeleteFile {
file_path: format!("{}/pos-del-2.parquet", table_location.to_str().unwrap()),
file_size_in_bytes: std::fs::metadata(format!(
"{}/pos-del-2.parquet",
table_location.to_str().unwrap()
))
.unwrap()
.len(),
file_type: DataContentType::PositionDeletes,
partition_spec_id: 0,
equality_ids: None,
};

let pos_del_3 = FileScanTaskDeleteFile {
file_path: format!("{}/pos-del-3.parquet", table_location.to_str().unwrap()),
file_size_in_bytes: std::fs::metadata(format!(
"{}/pos-del-3.parquet",
table_location.to_str().unwrap()
))
.unwrap()
.len(),
file_type: DataContentType::PositionDeletes,
partition_spec_id: 0,
equality_ids: None,
};

let file_scan_tasks = vec![
FileScanTask {
file_size_in_bytes: 0,
start: 0,
length: 0,
record_count: None,
Expand All @@ -414,6 +433,7 @@ pub(crate) mod tests {
case_sensitive: false,
},
FileScanTask {
file_size_in_bytes: 0,
start: 0,
length: 0,
record_count: None,
Expand Down Expand Up @@ -464,6 +484,7 @@ pub(crate) mod tests {

// ---------- fake FileScanTask ----------
let task = FileScanTask {
file_size_in_bytes: 0,
start: 0,
length: 0,
record_count: None,
Expand All @@ -474,6 +495,7 @@ pub(crate) mod tests {
predicate: None,
deletes: vec![FileScanTaskDeleteFile {
file_path: "eq-del.parquet".to_string(),
file_size_in_bytes: 1, // never read; this test fails before opening the file
file_type: DataContentType::EqualityDeletes,
partition_spec_id: 0,
equality_ids: None,
Expand Down
Loading
Loading