This is an automated email from the ASF dual-hosted git repository.
zeroshade pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/arrow-go.git
The following commit(s) were added to refs/heads/main by this push:
new ecf655dd fix(parquet/schema): reject invalid fixed lengths (#1035)
ecf655dd is described below
commit ecf655dda477573d000a6329c731c6b4f66c61f3
Author: Minh Vu <[email protected]>
AuthorDate: Mon Jul 27 23:14:01 2026 +0200
fix(parquet/schema): reject invalid fixed lengths (#1035)
### Rationale for this change
Fixed-length byte-array constructors reject non-positive widths, but
`PrimitiveNode.SetTypeLength` allowed an existing valid node to be
mutated to zero or a negative width. The invalid node could then be
serialized or passed to encoders.
### What changes are included in this PR?
* Reject non-positive widths when mutating fixed-length byte-array
nodes.
* Leave non-fixed-width primitive nodes unchanged, preserving the
existing no-op behavior.
* Document the panic contract on the setter.
### Are these changes tested?
Yes. The regression test covers zero and negative widths, verifies that
failed mutations preserve the previous width, and checks the existing
no-op behavior for other physical types. The full `parquet/schema`
package passes.
---
parquet/schema/node.go | 7 +++++--
parquet/schema/schema_test.go | 14 ++++++++++++++
2 files changed, 19 insertions(+), 2 deletions(-)
diff --git a/parquet/schema/node.go b/parquet/schema/node.go
index f95c6164..75ae1c5e 100644
--- a/parquet/schema/node.go
+++ b/parquet/schema/node.go
@@ -319,10 +319,13 @@ func (p *PrimitiveNode) Equals(rhs Node) bool {
// to store the values in this column.
func (p *PrimitiveNode) PhysicalType() parquet.Type { return p.physicalType }
-// SetTypeLength will change the type length of the node, has no effect if the
-// physical type is not FixedLength Byte Array
+// SetTypeLength will change the type length of the node, and has no effect if
the
+// physical type is not FixedLength Byte Array. It panics if length is not
positive.
func (p *PrimitiveNode) SetTypeLength(length int) {
if p.PhysicalType() == parquet.Types.FixedLenByteArray {
+ if length <= 0 {
+ panic("parquet: fixed length byte array length must be
positive")
+ }
p.typeLen = length
}
}
diff --git a/parquet/schema/schema_test.go b/parquet/schema/schema_test.go
index f7d2818b..fb3dafe2 100644
--- a/parquet/schema/schema_test.go
+++ b/parquet/schema/schema_test.go
@@ -191,6 +191,20 @@ func (p *PrimitiveNodeTestSuite) TestEquals() {
p.False(flba1.Equals(flba5))
}
+func (p *PrimitiveNodeTestSuite) TestSetTypeLengthRejectsInvalidLength() {
+ node := schema.NewFixedLenByteArrayNode("value",
parquet.Repetitions.Required, 4, -1)
+
+ for _, length := range []int{0, -1} {
+ p.PanicsWithValue("parquet: fixed length byte array length must
be positive", func() {
+ node.SetTypeLength(length)
+ })
+ p.Equal(4, node.TypeLength())
+ }
+
+ other := schema.NewInt32Node("value", parquet.Repetitions.Required, -1)
+ p.NotPanics(func() { other.SetTypeLength(0) })
+}
+
func (p *PrimitiveNodeTestSuite) TestPhysicalLogicalMapping() {
tests := []struct {
typ parquet.Type
