Github user cloud-fan commented on a diff in the pull request:
https://github.com/apache/spark/pull/12640#discussion_r61539393
--- Diff:
sql/catalyst/src/main/java/org/apache/spark/sql/catalyst/expressions/UnsafeArrayData.java
---
@@ -336,4 +336,62 @@ public UnsafeArrayData copy() {
arrayCopy.pointTo(arrayDataCopy, Platform.BYTE_ARRAY_OFFSET,
sizeInBytes);
return arrayCopy;
}
+
+ public int[] toPrimitiveIntArray() {
+ int[] result = new int[numElements];
+ Platform.copyMemory(baseObject, baseOffset + 4 + 4 * numElements,
+ result, Platform.INT_ARRAY_OFFSET, 4 * numElements);
+ return result;
+ }
+
+ public double[] toPrimitiveDoubleArray() {
+ double[] result = new double[numElements];
+ Platform.copyMemory(baseObject, baseOffset + 4 + 4 * numElements,
+ result, Platform.DOUBLE_ARRAY_OFFSET, 8 * numElements);
+ return result;
+ }
+
+ public static UnsafeArrayData fromPrimitiveArray(int[] arr) {
+ int offsetRegionSize = 4 * arr.length;
+ int valueRegionSize = 4 * arr.length;
+ int totalSize = 4 + offsetRegionSize + valueRegionSize;
+ byte[] data = new byte[totalSize];
--- End diff --
After think about it, we can't support very long array in
`UnsafeArrayData`, as the offset region only use 4 bytes to encode the value
offset. I'm going to add a validation so that we won't overflow if given a big
array.
Actually we can improve the unsafe format for array data, which is similar
to unsafe row:
```
[null bits] [values] [variable length portion]
```
One difference is that, the `values` region in unsafe row is 8-byte per
field, but for array, it should depend on the data size, e.g. 1 byte for
boolean. Then we can just memory copy primitive array into `values` region and
no need to care about the offset region.
---
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]
