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.
---