Github user hotou commented on a diff in the pull request:
https://github.com/apache/spark/pull/4701#discussion_r25072880
--- Diff: core/src/main/scala/org/apache/spark/rdd/JdbcRDD.scala ---
@@ -64,8 +64,8 @@ class JdbcRDD[T: ClassTag](
// bounds are inclusive, hence the + 1 here and - 1 on end
val length = 1 + upperBound - lowerBound
(0 until numPartitions).map(i => {
- val start = lowerBound + ((i * length) / numPartitions).toLong
- val end = lowerBound + (((i + 1) * length) / numPartitions).toLong -
1
+ val start = lowerBound + ((BigDecimal(i) * length) /
numPartitions).toLong
--- End diff --
@rxin I actually favor the current partition algo, it's pretty neat in a
way, for example
lowerBound = 1
upperBound = 100
numPartition = 8
With fix length increase, you get
[1,13],[14,26],[27,39],[40,52],[53,65],[66,78],[79,91],[92,100]
In which you always end up with one small partition at the end
With the current algo you get
[1,12],[13,25],[26,37],[38,50],[51,62],[63,75],[76,87],[88,100]
You get more evenly distributed partitions
---
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]
