Github user steveloughran commented on a diff in the pull request:
https://github.com/apache/spark/pull/18111#discussion_r133751724
--- Diff:
core/src/main/scala/org/apache/spark/internal/io/HadoopMapReduceCommitProtocol.scala
---
@@ -73,7 +73,10 @@ class HadoopMapReduceCommitProtocol(jobId: String, path:
String)
val stagingDir: String = committer match {
// For FileOutputCommitter it has its own staging path called "work
path".
- case f: FileOutputCommitter =>
Option(f.getWorkPath.toString).getOrElse(path)
+ case f: FileOutputCommitter =>
+ val workPath = f.getWorkPath
+ require(workPath != null, s"Committer has no workpath $f")
+ Option(workPath.toString).getOrElse(path)
--- End diff --
{{workPath.toString()}} triggers an NPE, which was the reason for the stack
trace.
Now, what's this code trying to do? find a directory to put stuff. If the
committer is a subclass of `FileOutputFormat`, then its getWorkPath method can
be called, which *should* return a non-null path.
If the requirement of the code is "get the workpath or return "path" if its
null", maybe the code should really be
```scala
Option(f.getWorkPath).getOrElse(path).toString
```
That'd return the workPath if non null, falling back to the `path`
variable, and then call `toString` on the returned object. I'm not sure off the
top of my head if Scala type inference likes that though. It may be less
elegant and more reliable to have
```scala
val workPath = f.getWorkPath
if (workPath != null) workPath.toString else path;
```
+maybe a bonus paranoid check for workPath being "".
I think that would actually get closer to the original goal of the code:
always return a path, even of the committer doesn't have one
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
