Github user twalthr commented on a diff in the pull request:
https://github.com/apache/flink/pull/6218#discussion_r199746426
--- Diff:
flink-formats/flink-avro/src/main/java/org/apache/flink/formats/avro/AvroRowSerializationSchema.java
---
@@ -37,18 +43,42 @@
import java.io.IOException;
import java.io.ObjectInputStream;
import java.io.ObjectOutputStream;
+import java.math.BigDecimal;
+import java.nio.ByteBuffer;
+import java.sql.Date;
+import java.sql.Time;
+import java.sql.Timestamp;
+import java.util.HashMap;
import java.util.List;
+import java.util.Map;
+import java.util.TimeZone;
/**
- * Serialization schema that serializes {@link Row} over {@link
SpecificRecord} into a Avro bytes.
+ * Serialization schema that serializes {@link Row} into Avro bytes.
+ *
+ * <p>Serializes objects that are represented in (nested) Flink rows. It
support types that
+ * are compatible with Flink's Table & SQL API.
+ *
+ * <p>Note: Changes in this class need to be kept in sync with the
corresponding runtime
+ * class {@link AvroRowDeserializationSchema} and schema converter {@link
AvroSchemaConverter}.
*/
public class AvroRowSerializationSchema implements
SerializationSchema<Row> {
/**
- * Avro record class.
+ * Used for time conversions into SQL types.
+ */
+ private static final TimeZone LOCAL_TZ = TimeZone.getDefault();
--- End diff --
We are using this pattern at different places. E.g.
`org.apache.flink.orc.OrcBatchReader`. The problem is that Java's SQL
time/date/timestamp are a complete design fail. They are timezone specific.
This adds/removes the local timezone from the timestamp. Such that the string
representation of the produced `Timestamp` object is always correct.
---
