Github user nongli commented on a diff in the pull request:
https://github.com/apache/spark/pull/11055#discussion_r51922204
--- Diff:
sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/UnsafeRowParquetRecordReader.java
---
@@ -645,7 +676,15 @@ private void decodeDictionaryIds(int rowId, int num,
ColumnVector column) {
}
} else if (column.dataType() == DataTypes.ByteType) {
for (int i = rowId; i < rowId + num; ++i) {
- column.putByte(i,
(byte)dictionary.decodeToInt(dictionaryIds.getInt(i)));
+ column.putByte(i, (byte)
dictionary.decodeToInt(dictionaryIds.getInt(i)));
+ }
+ } else if (column.dataType() == DataTypes.ShortType) {
+ for (int i = rowId; i < rowId + num; ++i) {
+ column.putShort(i, (short)
dictionary.decodeToInt(dictionaryIds.getInt(i)));
+ }
+ } else if (DecimalType.is64BitDecimalType(column.dataType())) {
--- End diff --
This code is like a nested loop. On the outer, we loop over the parquet
type and on the inner, the column vector type.
In the case you mentioned, the parquet type is int64 and the column vector
type is long so I think this just works. Let me make that more explicit by
adding a check in the INT64 branch.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
