Github user jinxing64 commented on a diff in the pull request:
https://github.com/apache/spark/pull/19868#discussion_r181014951
--- Diff:
sql/hive/src/main/scala/org/apache/spark/sql/hive/TableReader.scala ---
@@ -176,12 +176,13 @@ class HadoopTableReader(
val matches = fs.globStatus(pathPattern)
matches.foreach(fileStatus => existPathSet +=
fileStatus.getPath.toString)
}
- // convert /demo/data/year/month/day to /demo/data/*/*/*/
+ // convert /demo/data/year/month/day to
/demo/data/year/month/*/
--- End diff --
@cloud-fan @jiangxb1987
Thanks a lot for review.
> Em... It seems we have to check all the levels unless we have specified a
value for each partition column. We can make some improvement here but seems
that require more complicated approach.
Yes, true. In this change, I only optimize when user specify for each
partition column, which is very common in the production -- as our user always
did: `select xxx from yyy where year=yy and month=mm and day=dd`
I'm not sure about you guys idea: leave the current logic as it is(at
least the code logic now is very simple)? or implement a more complicated
approach and defend as many cases as possible? or do some improvement based on
this pr and cover some very common cases?
Thanks again for review :)
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
