You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@iceberg.apache.org by at...@apache.org on 2026/07/24 20:06:31 UTC
(iceberg-go) branch main updated: fix(table): reject null position deletes (#1518)
This is an automated email from the ASF dual-hosted git repository.
laskoviymishka pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/iceberg-go.git
The following commit(s) were added to refs/heads/main by this push:
new 9097100f4 fix(table): reject null position deletes (#1518)
9097100f4 is described below
commit 9097100f4aa3bae419644e3b1f25d3fe29ce8fb7
Author: Minh Vu <vu...@gmail.com>
AuthorDate: Fri Jul 24 22:06:26 2026 +0200
fix(table): reject null position deletes (#1518)
## Motivation
The Iceberg position delete schema requires `pos` values to be non-null.
The Arrow scanner currently reads the raw `Int64Values` backing slice
without checking the validity bitmap, so a malformed delete file
containing a null position may be interpreted as a real row position
instead of returning an invalid-schema error.
## Changes
- Reject null values in position delete `pos` chunks with
`iceberg.ErrInvalidSchema`.
- Add a regression test covering a nullable position array and verifying
the error contract.
---
table/arrow_scanner.go | 3 ++
table/arrow_scanner_posdelete_regression_test.go | 35 ++++++++++++++++++++++++
2 files changed, 38 insertions(+)
diff --git a/table/arrow_scanner.go b/table/arrow_scanner.go
index 02da7fa1d..807629606 100644
--- a/table/arrow_scanner.go
+++ b/table/arrow_scanner.go
@@ -451,6 +451,9 @@ func readDeletes(ctx context.Context, fs iceio.IO, dataFile iceberg.DataFile) (_
filePathCol := tbl.Column(tbl.Schema().FieldIndices("file_path")[0]).Data()
posCol := tbl.Column(tbl.Schema().FieldIndices("pos")[0]).Data()
+ if posCol.NullN() > 0 {
+ return nil, fmt.Errorf("%w: null pos in position delete file", iceberg.ErrInvalidSchema)
+ }
return groupPosDeletesByFilePath(ctx, filePathCol, posCol)
}
diff --git a/table/arrow_scanner_posdelete_regression_test.go b/table/arrow_scanner_posdelete_regression_test.go
index 955809bca..52d6634eb 100644
--- a/table/arrow_scanner_posdelete_regression_test.go
+++ b/table/arrow_scanner_posdelete_regression_test.go
@@ -24,7 +24,10 @@ import (
"github.com/apache/arrow-go/v18/arrow/array"
"github.com/apache/arrow-go/v18/arrow/compute"
"github.com/apache/arrow-go/v18/arrow/memory"
+ "github.com/apache/arrow-go/v18/parquet"
+ "github.com/apache/arrow-go/v18/parquet/pqarrow"
"github.com/apache/iceberg-go"
+ iceio "github.com/apache/iceberg-go/io"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require"
)
@@ -265,6 +268,38 @@ func TestCollectPosDeletePositionsRejectsUnsupportedPosType(t *testing.T) {
assert.Contains(t, err.Error(), "unsupported pos column type")
}
+func TestReadDeletesRejectsNullPos(t *testing.T) {
+ mem := memory.NewCheckedAllocator(memory.DefaultAllocator)
+ ctx := compute.WithAllocator(t.Context(), mem)
+ defer mem.AssertSize(t, 0)
+
+ fields := PositionalDeleteArrowSchema.Fields()
+ fields[1].Nullable = true
+ deleteSchema := arrow.NewSchema(fields, nil)
+ deletePath := "mem://bucket/deletes/null-pos.parquet"
+ dataPath := "mem://bucket/data/data.parquet"
+
+ rec := mustLoadRecordBatchFromJSON(deleteSchema, `[
+ {"file_path": "`+dataPath+`", "pos": null}
+ ]`)
+ defer rec.Release()
+ tbl := array.NewTableFromRecords(deleteSchema, []arrow.RecordBatch{rec})
+ defer tbl.Release()
+
+ memFS := iceio.NewMemFS()
+ fw, err := memFS.Create(deletePath)
+ require.NoError(t, err)
+ require.NoError(t, pqarrow.WriteTable(tbl, fw, rec.NumRows(),
+ parquet.NewWriterProperties(parquet.WithStats(true)),
+ pqarrow.DefaultWriterProps()))
+ require.NoError(t, fw.Close())
+
+ deletes, err := readDeletes(ctx, memFS, newPosDeleteFile(t, deletePath, 1, 128))
+ require.ErrorIs(t, err, iceberg.ErrInvalidSchema)
+ assert.Nil(t, deletes)
+ assert.Contains(t, err.Error(), "null pos in position delete file")
+}
+
// TestProcessPositionalDeletesAcrossBatches is the regression net for the
// positional-delete index bug: processPositionalDeletes applies deletes one Arrow
// batch at a time, but the surviving-row indices must index into the *current