You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Gurpal SINGH (Jira)" <ji...@apache.org> on 2025/08/21 12:43:00 UTC
[jira] [Created] (SPARK-53347) Spark from_protobuf() incorrectly deserializes "false" boolean values as null
Gurpal SINGH created SPARK-53347:
------------------------------------
Summary: Spark from_protobuf() incorrectly deserializes "false" boolean values as null
Key: SPARK-53347
URL: https://issues.apache.org/jira/browse/SPARK-53347
Project: Spark
Issue Type: Bug
Components: Protobuf
Affects Versions: 4.0.0, 3.5.2, 3.5.1, 3.5.0
Environment: Scala 2.13,
Spark 3.5.2
JDK 17
Maven 3.9.11
Reporter: Gurpal SINGH
*Problem*
When deserializing a Protobuf message using `{_}from_protobuf(){_}` in Spark, boolean fields with the value `false` are incorrectly deserialized as `null`. This leads to incorrect data in the resulting DataFrame and breaks semantic expectations.
*Reproduction*
Given a Protobuf message like the following (using `{_}google.protobuf.BoolValue{_}`):
{code:java}
syntax = "proto3";
message Example {
google.protobuf.BoolValue is_active = 1;
}{code}
And a message where is_active is explicitly set to false, the result of _from_protobuf()_ will show null instead of false.
*Root Cause*
In {_}ProtobufDeserializer.scala{_}, the logic for deserializing google.protobuf.BoolValue relies on the getFieldValue() method, which uses the following condition:
{code:java}
if (field.isRepeated || record.hasField(field) || field.hasDefaultValue || (!field.hasPresence && this.emitDefaultValues)) {
record.getField(field)
} else {
null
} {code}
However, for BoolValue, even when the inner value is false, record.hasField(field) returns false — as the field is present but set to false (not "unset"). As a result, _getFieldValue()_ returns null instead of false.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org