This is an automated email from the ASF dual-hosted git repository.
github-merge-queue[bot] pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/iceberg-rust.git
The following commit(s) were added to refs/heads/main by this push:
new 582b9e103 feat(delete-vector): make the delete_vector module public
(#3461)
582b9e103 is described below
commit 582b9e103513f21094f61cd1f09fdbf39dccf13e
Author: Giladi <[email protected]>
AuthorDate: Thu Oct 15 05:44:23 2026 +0000
feat(delete-vector): make the delete_vector module public (#3461)
An engine that uses the crate for metadata but reads Parquet with a reader
of its own cannot go through ArrowReader, and so far had to reimplement the
deletion-vector-v1 decoding to apply deletion vectors. Expose the module,
the way puffin already is, with the docs deny(missing_docs) then requires,
an is_empty next to len, and the public API snapshot regenerated.
Co-authored-by: Claude <[email protected]>
---
crates/iceberg/public-api.txt | 22 ++++++++++++++++++++++
crates/iceberg/src/delete_vector.rs | 20 +++++++++++++++++++-
crates/iceberg/src/lib.rs | 2 +-
3 files changed, 42 insertions(+), 2 deletions(-)
diff --git a/crates/iceberg/public-api.txt b/crates/iceberg/public-api.txt
index 466614757..ff417ddb5 100644
--- a/crates/iceberg/public-api.txt
+++ b/crates/iceberg/public-api.txt
@@ -229,6 +229,28 @@ impl core::marker::Copy for
iceberg::cow_rewrite::CowRewriteStats
impl core::marker::StructuralPartialEq for
iceberg::cow_rewrite::CowRewriteStats
pub trait iceberg::cow_rewrite::CowBatchRewriter: core::marker::Send +
core::marker::Sync
pub fn iceberg::cow_rewrite::CowBatchRewriter::rewrite_batch(&self, batch:
arrow_array::record_batch::RecordBatch) ->
iceberg::Result<iceberg::cow_rewrite::CowBatchRewrite>
+pub mod iceberg::delete_vector
+pub struct iceberg::delete_vector::DeleteVector
+impl iceberg::delete_vector::DeleteVector
+pub fn iceberg::delete_vector::DeleteVector::deserialize(blob: &[u8]) ->
iceberg::Result<Self>
+pub fn iceberg::delete_vector::DeleteVector::insert(&mut self, pos: u64) ->
bool
+pub fn iceberg::delete_vector::DeleteVector::insert_positions(&mut self,
positions: &[u64]) -> iceberg::Result<usize>
+pub fn iceberg::delete_vector::DeleteVector::is_empty(&self) -> bool
+pub fn iceberg::delete_vector::DeleteVector::iter(&self) ->
iceberg::delete_vector::DeleteVectorIterator<'_>
+pub fn iceberg::delete_vector::DeleteVector::len(&self) -> u64
+pub fn iceberg::delete_vector::DeleteVector::new(roaring_treemap:
roaring::treemap::RoaringTreemap) -> iceberg::delete_vector::DeleteVector
+impl core::default::Default for iceberg::delete_vector::DeleteVector
+pub fn iceberg::delete_vector::DeleteVector::default() ->
iceberg::delete_vector::DeleteVector
+impl core::fmt::Debug for iceberg::delete_vector::DeleteVector
+pub fn iceberg::delete_vector::DeleteVector::fmt(&self, f: &mut
core::fmt::Formatter<'_>) -> core::fmt::Result
+impl core::ops::bit::BitOrAssign for iceberg::delete_vector::DeleteVector
+pub fn iceberg::delete_vector::DeleteVector::bitor_assign(&mut self, other:
Self)
+pub struct iceberg::delete_vector::DeleteVectorIterator<'a>
+impl iceberg::delete_vector::DeleteVectorIterator<'_>
+pub fn iceberg::delete_vector::DeleteVectorIterator<'_>::advance_to(&mut self,
pos: u64)
+impl core::iter::traits::iterator::Iterator for
iceberg::delete_vector::DeleteVectorIterator<'_>
+pub type iceberg::delete_vector::DeleteVectorIterator<'_>::Item = u64
+pub fn iceberg::delete_vector::DeleteVectorIterator<'_>::next(&mut self) ->
core::option::Option<Self::Item>
pub mod iceberg::encryption
pub mod iceberg::encryption::kms
pub struct iceberg::encryption::kms::GeneratedKey
diff --git a/crates/iceberg/src/delete_vector.rs
b/crates/iceberg/src/delete_vector.rs
index 9b2d48d91..dc76eeadb 100644
--- a/crates/iceberg/src/delete_vector.rs
+++ b/crates/iceberg/src/delete_vector.rs
@@ -15,6 +15,11 @@
// specific language governing permissions and limitations
// under the License.
+//! Deletion vectors: the deleted row positions of one data file, as a Puffin
+//! `deletion-vector-v1` blob decodes to. Public so an engine that reads
+//! Parquet with a reader of its own can still decode the vectors Iceberg
+//! writes.
+
use std::ops::BitOrAssign;
use bytes::Buf;
@@ -33,24 +38,28 @@ const DV_MAGIC_BYTES: usize = 4;
const DV_CRC_BYTES: usize = 4;
const DV_MIN_BLOB_BYTES: usize = DV_LENGTH_PREFIX_BYTES + DV_MAGIC_BYTES +
DV_CRC_BYTES;
+/// The deleted row positions of one data file: a set of 64-bit positions,
+/// kept as a [`RoaringTreemap`].
#[derive(Debug, Default)]
pub struct DeleteVector {
inner: RoaringTreemap,
}
impl DeleteVector {
- #[allow(unused)]
+ /// Wraps `roaring_treemap`, whose members are the deleted positions.
pub fn new(roaring_treemap: RoaringTreemap) -> DeleteVector {
DeleteVector {
inner: roaring_treemap,
}
}
+ /// Iterates the deleted positions in ascending order.
pub fn iter(&self) -> DeleteVectorIterator<'_> {
let outer = self.inner.bitmaps();
DeleteVectorIterator { outer, inner: None }
}
+ /// Marks `pos` as deleted, returning whether it was not already.
pub fn insert(&mut self, pos: u64) -> bool {
self.inner.insert(pos)
}
@@ -73,10 +82,16 @@ impl DeleteVector {
Ok(positions.len())
}
+ /// How many positions are deleted.
pub fn len(&self) -> u64 {
self.inner.len()
}
+ /// Whether no position is deleted.
+ pub fn is_empty(&self) -> bool {
+ self.inner.is_empty()
+ }
+
/// Parses a `deletion-vector-v1` Puffin blob into a `DeleteVector`.
///
/// The layout, defined by the Iceberg Puffin spec and matching
Iceberg-Java's
@@ -216,6 +231,7 @@ fn decode_roaring_directory(mut reader: &[u8]) ->
Result<RoaringTreemap> {
// There is a PR open on roaring to add this
(https://github.com/RoaringBitmap/roaring-rs/pull/314)
// and if that gets merged then we can simplify `DeleteVectorIterator` here,
refactoring `advance_to`
// to just a wrapper around the underlying iterator's method.
+/// Iterates the deleted positions of a [`DeleteVector`] in ascending order.
pub struct DeleteVectorIterator<'a> {
// NB: `BitMapIter` was only exposed publicly in
https://github.com/RoaringBitmap/roaring-rs/pull/316
// which is not yet released. As a consequence our Cargo.toml temporarily
uses a git reference for
@@ -253,6 +269,8 @@ impl Iterator for DeleteVectorIterator<'_> {
}
impl DeleteVectorIterator<'_> {
+ /// Skips the positions below `pos`, so the next one yielded is the first
+ /// at or after it. Does nothing before the first position is yielded.
pub fn advance_to(&mut self, pos: u64) {
let hi = (pos >> 32) as u32;
let lo = pos as u32;
diff --git a/crates/iceberg/src/lib.rs b/crates/iceberg/src/lib.rs
index 9733fac32..72c2acd3e 100644
--- a/crates/iceberg/src/lib.rs
+++ b/crates/iceberg/src/lib.rs
@@ -104,7 +104,7 @@ pub mod encryption;
mod test_utils;
pub mod writer;
-mod delete_vector;
+pub mod delete_vector;
pub mod metadata_columns;
pub mod partitioning;
pub mod puffin;
