Skip to content

perf(scan): prune index manifest entries #703

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
JingsongLi merged 1 commit into from
Aug 10, 2026
+130 −17
Merged

perf(scan): prune index manifest entries #703

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
147 changes: 130 additions & 17 deletions crates/paimon/src/table/table_scan.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@ use super::stats_filter::{
data_file_matches_predicates_with_key_stats, group_by_overlapping_row_id, FileStatsRows,
ResolvedStatsSchema,
};
use super::Table;
use super::{find_field_id_by_name, Table};
use crate::io::FileIO;
use crate::spec::{
avro::SharedSchemaCache, bucket_dir_name, BinaryRow, BucketFunctionType, CoreOptions,
Expand Down Expand Up @@ -394,33 +394,56 @@ fn split_row_ranges_for_files(
Ok(Some(split_ranges))
}

fn scan_predicate_field_ids(predicates: &[Predicate], schema_fields: &[DataField]) -> HashSet<i32> {
let mut leaf_refs = Vec::new();
for predicate in predicates {
crate::arrow::residual::collect_predicate_leaf_refs(predicate, &mut leaf_refs);
}
leaf_refs
.into_iter()
// `_ROW_ID` uses a placeholder leaf index, so match the scanner's name-based resolution.
.filter_map(|(column, _)| find_field_id_by_name(schema_fields, column))
.collect()
}

fn retain_index_manifest_entry(
entry: &IndexManifestEntry,
global_index_needed: bool,
deletion_vectors_needed: bool,
) -> bool {
(deletion_vectors_needed && entry.index_file.index_type == DELETION_VECTORS_INDEX_TYPE)
|| (global_index_needed
&& normalize_sorted_global_index_type(&entry.index_file.index_type).is_some())
entry.kind == FileKind::Add
&& ((deletion_vectors_needed && entry.index_file.index_type == DELETION_VECTORS_INDEX_TYPE)
|| (global_index_needed
&& normalize_sorted_global_index_type(&entry.index_file.index_type).is_some()))
}

fn retain_index_manifest_entry_for_scan(
entry: &IndexManifestEntry,
global_index_needed: bool,
deletion_vectors_needed: bool,
partition_filter: Option<&PartitionFilter>,
predicate_field_ids: &HashSet<i32>,
) -> crate::Result<bool> {
if !retain_index_manifest_entry(entry, global_index_needed, deletion_vectors_needed) {
return Ok(false);
}
// Deletion vectors are selected by partition and bucket when splits are built.
if let Some(filter) = partition_filter {
if !filter.matches_entry(&entry.partition)? {
return Ok(false);
}
}
if normalize_sorted_global_index_type(&entry.index_file.index_type).is_none() {
return Ok(true);
}
partition_filter
.map(|filter| filter.matches_entry(&entry.partition))
.transpose()
.map(|matched| matched.unwrap_or(true))
let Some(global_index) = entry.index_file.global_index_meta.as_ref() else {
// Keep malformed entries so the scanner can preserve fail-loud validation.
return Ok(true);
};
Ok(predicate_field_ids.contains(&global_index.index_field_id)

@XiaoHongbo-Hope XiaoHongbo-Hope Aug 10, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Malformed index_meta for an unrelated sorted index is pruned before validation, so it no longer fails loudly as described in the PR.

@QuakeWang QuakeWang Aug 10, 2026

Copy link
Copy Markdown
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@XiaoHongbo-Hope Good catch. This is intentional: unrelated indexes are pruned before index_meta validation, matching Java. I've updated the PR description to clarify the fail-loud scope.

|| global_index.extra_field_ids.as_ref().is_some_and(|ids| {
ids.iter()
.any(|field_id| predicate_field_ids.contains(field_id))
}))
}

/// Builds a map from (partition, bucket) to (data_file_name -> DeletionFile) from index manifest entries.
Expand Down Expand Up @@ -1256,13 +1279,19 @@ impl<'a> PaimonTableScan<'a> {
};
let table_path = self.table.location().trim_end_matches('/');
let path = format!("{table_path}/{MANIFEST_DIR}/{index_manifest_name}");
let predicate_field_ids = if global_index_needed {
scan_predicate_field_ids(&self.data_predicates, self.table.schema().fields())
} else {
HashSet::new()
};
let mut entries = Vec::new();
for entry in IndexManifest::read(self.table.file_io(), &path).await? {
if retain_index_manifest_entry_for_scan(
&entry,
global_index_needed,
deletion_vectors_needed,
self.partition_filter.as_ref(),
&predicate_field_ids,
)? {
entries.push(entry);
}
Expand Down Expand Up @@ -2078,17 +2107,18 @@ mod tests {
group_data_files_by_partition_bucket, manifest_file_overlaps_row_range_index,
prune_data_evolution_group_by_read_fields, retain_index_manifest_entry,
retain_index_manifest_entry_for_scan, retain_manifest_entry_row_ranges,
retain_manifest_row_ranges, should_skip_level_zero_for_scan, split_row_ranges_for_files,
LimitPushdownAccumulator, PaimonTableScan, RowRangeIndex, TableScan,
retain_manifest_row_ranges, scan_predicate_field_ids, should_skip_level_zero_for_scan,
split_row_ranges_for_files, LimitPushdownAccumulator, PaimonTableScan, RowRangeIndex,
TableScan,
};
use crate::catalog::Identifier;
use crate::io::FileIOBuilder;
use crate::spec::{
stats::BinaryTableStats, ArrayType, BinaryRow, BinaryRowBuilder, BucketFunctionType,
ColumnMove, CommitKind, DataField, DataFileMeta, DataType, Datum, DeletionVectorMeta,
FileKind, IndexFileMeta, IndexManifestEntry, IntType, ManifestEntry, ManifestFileMeta,
Predicate, PredicateBuilder, PredicateOperator, Schema as PaimonSchema, SchemaChange,
Snapshot, TableSchema, VarCharType,
FileKind, GlobalIndexMeta, IndexFileMeta, IndexManifestEntry, IntType, ManifestEntry,
ManifestFileMeta, Predicate, PredicateBuilder, PredicateOperator, Schema as PaimonSchema,
SchemaChange, Snapshot, TableSchema, VarCharType,
};
use crate::table::bucket_filter::{compute_target_buckets, extract_predicate_for_keys};
use crate::table::partition_filter::PartitionFilter;
Expand Down Expand Up @@ -4183,6 +4213,10 @@ mod tests {
retained(true, true),
vec!["DELETION_VECTORS", "btree", "bitmap"]
);

let mut deleted = entry("btree");
deleted.kind = FileKind::Delete;
assert!(!retain_index_manifest_entry(&deleted, true, false));
}

#[test]
Expand All @@ -4202,6 +4236,7 @@ mod tests {
)],
)
.unwrap();
let predicate_field_ids = HashSet::from([1]);
let entry = |partition, index_type: &str| IndexManifestEntry {
version: 1,
kind: FileKind::Add,
Expand All @@ -4213,33 +4248,111 @@ mod tests {
file_size: 1,
row_count: 1,
deletion_vectors_ranges: None,
global_index_meta: None,
global_index_meta: (index_type == "btree").then_some(GlobalIndexMeta {
row_range_start: 0,
row_range_end: 0,
index_field_id: 1,
extra_field_ids: None,
index_meta: None,
source_meta: None,
}),
},
};

assert!(retain_index_manifest_entry_for_scan(
&entry(matching_partition, "btree"),
&entry(matching_partition.clone(), "btree"),
true,
false,
Some(&filter),
&predicate_field_ids,
)
.unwrap());
assert!(!retain_index_manifest_entry_for_scan(
&entry(partition(8), "btree"),
true,
false,
Some(&filter),
&predicate_field_ids,
)
.unwrap());
assert!(retain_index_manifest_entry_for_scan(
assert!(!retain_index_manifest_entry_for_scan(
&entry(partition(8), "DELETION_VECTORS"),
true,
true,
Some(&filter),
&predicate_field_ids,
)
.unwrap());
assert!(retain_index_manifest_entry_for_scan(
&entry(matching_partition, "DELETION_VECTORS"),
false,
true,
Some(&filter),
&predicate_field_ids,
)
.unwrap());
}

#[test]
fn test_retain_index_manifest_entries_for_predicate_fields() {
let entry = |index_field_id, extra_field_ids| IndexManifestEntry {
version: 1,
kind: FileKind::Add,
partition: Vec::new(),
bucket: 0,
index_file: IndexFileMeta {
index_type: "btree".to_string(),
file_name: format!("btree-{index_field_id}.idx"),
file_size: 1,
row_count: 1,
deletion_vectors_ranges: None,
global_index_meta: Some(GlobalIndexMeta {
row_range_start: 0,
row_range_end: 0,
index_field_id,
extra_field_ids,
index_meta: None,
source_meta: None,
}),
},
};
let predicate_field_ids = HashSet::from([2]);
let retained = |entry: &IndexManifestEntry| {
retain_index_manifest_entry_for_scan(entry, true, false, None, &predicate_field_ids)
.unwrap()
};

assert!(retained(&entry(2, None)));
assert!(retained(&entry(1, Some(vec![2]))));
assert!(!retained(&entry(1, Some(vec![3]))));

let mut missing_meta = entry(2, None);
missing_meta.index_file.global_index_meta = None;
assert!(
retained(&missing_meta),
"malformed sorted entries must reach scanner validation"
);
}

#[test]
fn test_scan_predicate_field_ids_resolve_by_name() {
let fields = vec![
DataField::new(10, "id".to_string(), DataType::Int(IntType::new())),
DataField::new(20, "value".to_string(), DataType::Int(IntType::new())),
];
let predicate = Predicate::and(vec![
PredicateBuilder::new(&fields)
.equal("value", Datum::Int(1))
.unwrap(),
crate::spec::row_id_leaf(PredicateOperator::Eq, vec![Datum::Long(7)]),
]);

assert_eq!(
scan_predicate_field_ids(&[predicate], &fields),
HashSet::from([20])
);
}

#[tokio::test]
async fn test_skip_index_manifest_without_active_consumer() {
let table = Table::new(
Expand Down
Loading