Github user liancheng commented on a diff in the pull request:
https://github.com/apache/spark/pull/1919#discussion_r16346630
--- Diff:
sql/hive/src/main/scala/org/apache/spark/sql/hive/execution/InsertIntoHiveTable.scala
---
@@ -191,7 +257,10 @@ case class InsertIntoHiveTable(
val outputData = new Array[Any](fieldOIs.length)
iter.map { row =>
var i = 0
- while (i < row.length) {
+ while (i < fieldOIs.length) {
+ if (fieldOIs.length < row.length && row.length - fieldOIs.length
== dynamicPartNum) {
+ dynamicPartPath =
getDynamicPartDir(fileSinkConf.getTableInfo, row, dynamicPartNum,
jobConfSer.value)
--- End diff --
OK, I finally understand the trick here... Although `dynamicPartPath` is
defined out of the closure, the `dynamicPartPath` in this line and the the one
used in `writeToFile2` are the same instance for a single row as the two
iterators are pipelined.
I admit I had never thought that we can use Spark in this way :) But this
is too hacky to follow. I'd suggest to define `dynamicPartPath` within this
closure and pass it as part of the output this RDD. Namely, change [this
line](https://github.com/apache/spark/pull/1919/files#diff-d579db9a8f27e0bbef37720ab14ec3f6L200)
to:
```scala
serializer.serialize(outputData, standardOI) -> dynamicPartPath
```
Then make `writeToFile2` receive an `Iterator[(Writable, String)]` instead
of an `Iterator[Writable]`.
---
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]
