GitHub user viirya opened a pull request:
https://github.com/apache/spark/pull/22357
[SPARK-25363][SQL] Fix schema pruning in where clause by ignoring
unnecessary root fields
## What changes were proposed in this pull request?
Schema pruning doesn't work if nested column is used in where clause.
For example,
```
sql("select name.first from contacts where name.first = 'David'")
== Physical Plan ==
*(1) Project [name#19.first AS first#40]
+- *(1) Filter (isnotnull(name#19) && (name#19.first = David))
+- *(1) FileScan parquet [name#19] Batched: false, Format: Parquet,
PartitionFilters: [],
PushedFilters: [IsNotNull(name)], ReadSchema:
struct<name:struct<first:string,middle:string,last:string>>
```
In above query plan, the scan node reads the entire schema of `name` column.
This issue is reported by:
https://github.com/apache/spark/pull/21320#issuecomment-419290197
The cause is that we infer a root field from expression `IsNotNull(name)`.
However, for such expression, we don't really use the nested fields of this
root field, so we can ignore the unnecessary nested fields.
## How was this patch tested?
Unit tests.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/viirya/spark-1 SPARK-25363
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/22357.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #22357
----
commit 3de6ee2c28e1242fead892f922d0053e31509a9f
Author: Liang-Chi Hsieh <viirya@...>
Date: 2018-09-07T05:06:07Z
Fix schema pruning when used in where clause.
----
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
