This is an automated email from the ASF dual-hosted git repository.
JingsongLi pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/paimon.git
The following commit(s) were added to refs/heads/master by this push:
new a9aaba86f1 [common] Always accept the first entry in bitmap index
block packing (#9614)
a9aaba86f1 is described below
commit a9aaba86f1c16ee478da8be2506089e86c4a4075
Author: YangJie <[email protected]>
AuthorDate: Thu Sep 10 02:33:59 2026 -0400
[common] Always accept the first entry in bitmap index block packing (#9614)
---
.../fileindex/bitmap/BitmapFileIndexMetaV2.java | 5 ++-
.../fileindex/bitmapindex/BitmapFileIndexTest.java | 36 ++++++++++++++++++++++
2 files changed, 40 insertions(+), 1 deletion(-)
diff --git
a/paimon-common/src/main/java/org/apache/paimon/fileindex/bitmap/BitmapFileIndexMetaV2.java
b/paimon-common/src/main/java/org/apache/paimon/fileindex/bitmap/BitmapFileIndexMetaV2.java
index f92e008162..8e6d79a0a1 100644
---
a/paimon-common/src/main/java/org/apache/paimon/fileindex/bitmap/BitmapFileIndexMetaV2.java
+++
b/paimon-common/src/main/java/org/apache/paimon/fileindex/bitmap/BitmapFileIndexMetaV2.java
@@ -377,7 +377,10 @@ public class BitmapFileIndexMetaV2 extends
BitmapFileIndexMeta {
key = entry.key;
}
int entryBytes = 2 * Integer.BYTES +
keyBytesMapper.apply(entry.key);
- if (serializedBytes + entryBytes > blockSizeLimit) {
+ // an empty block always accepts its first entry: the block size
limit is a
+ // packing target, and rejecting a single oversized key would fail
the whole
+ // index serialization
+ if (!entryList.isEmpty() && serializedBytes + entryBytes >
blockSizeLimit) {
return false;
}
serializedBytes += entryBytes;
diff --git
a/paimon-common/src/test/java/org/apache/paimon/fileindex/bitmapindex/BitmapFileIndexTest.java
b/paimon-common/src/test/java/org/apache/paimon/fileindex/bitmapindex/BitmapFileIndexTest.java
index cf67e9c2fe..1c56662ab1 100644
---
a/paimon-common/src/test/java/org/apache/paimon/fileindex/bitmapindex/BitmapFileIndexTest.java
+++
b/paimon-common/src/test/java/org/apache/paimon/fileindex/bitmapindex/BitmapFileIndexTest.java
@@ -34,6 +34,7 @@ import org.apache.paimon.types.DataTypes;
import org.apache.paimon.types.IntType;
import org.apache.paimon.types.VarCharType;
import org.apache.paimon.utils.RoaringBitmap32;
+import org.apache.paimon.utils.StringUtils;
import org.apache.commons.io.FileUtils;
import org.junit.Rule;
@@ -148,6 +149,41 @@ public class BitmapFileIndexTest {
return bitmapFileIndex.createReader(localSeekableInputStream, 0, 0);
}
+ @Test
+ public void testV2EntryLargerThanBlockSize() throws Exception {
+ FieldRef fieldRef = new FieldRef(0, "", DataTypes.STRING());
+ // "big" serializes to more than the default 16kb index-block-size, so
it cannot
+ // share a block with any other key; the writer must still produce a
readable
+ // index instead of failing with "index fail". "a" and "b" pack into
the first
+ // block, so the ordinary size check is exercised too.
+ BinaryString big = BinaryString.fromString(StringUtils.repeat("x",
20_000));
+ BinaryString a = BinaryString.fromString("a");
+ BinaryString b = BinaryString.fromString("b");
+ Object[] dataColumn = {big, null, a, big, b};
+ FileIndexReader reader =
+ createTestReaderOnWriter(
+ BitmapFileIndex.VERSION_2,
+ 16 * 1024,
+ DataTypes.STRING(),
+ writer -> {
+ for (Object o : dataColumn) {
+ writer.write(o);
+ }
+ });
+ assert ((BitmapIndexResult) reader.visitEqual(fieldRef, big))
+ .get()
+ .equals(RoaringBitmap32.bitmapOf(0, 3));
+ assert ((BitmapIndexResult) reader.visitEqual(fieldRef, a))
+ .get()
+ .equals(RoaringBitmap32.bitmapOf(2));
+ assert ((BitmapIndexResult) reader.visitEqual(fieldRef, b))
+ .get()
+ .equals(RoaringBitmap32.bitmapOf(4));
+ assert ((BitmapIndexResult) reader.visitIsNull(fieldRef))
+ .get()
+ .equals(RoaringBitmap32.bitmapOf(1));
+ }
+
private void testStringType(int version) throws Exception {
FieldRef fieldRef = new FieldRef(0, "", DataTypes.STRING());
BinaryString a = BinaryString.fromString("a");
