Skip to content

[SPARK-21677][SQL] json_tuple throws NullPointException when column is null as string type #18930

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
jmchung wants to merge 11 commits into from
Closed

[SPARK-21677][SQL] json_tuple throws NullPointException when column is null as string type #18930

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
596c280
fix null field name and add corresponding test
jmchung Aug 12, 2017
796041f
add description
jmchung Aug 12, 2017
f07a9f7
modify the comment about nullFieldName
jmchung Aug 13, 2017
ffa575a
make properly comment on pseudo field name
jmchung Aug 13, 2017
5d71263
discard fake field name, apply Option instead
jmchung Aug 14, 2017
0078445
added unit test that mixes constant field name and non constant one f…
jmchung Aug 15, 2017
5c69df5
use IndexedSeq instead of Array on foldableFieldNames to get better p…
jmchung Aug 15, 2017
ab16929
remove redundant comment and move toIndexedSeq after the map
jmchung Aug 15, 2017
e0e0c74
move test case from JsonSuite to JsonExpressionsSuite and add new tes…
jmchung Aug 16, 2017
5191ed4
consistent style in SQL and append the corresponding golden file of j…
jmchung Aug 17, 2017
ff3b9da
add a newline at the end and alter SQL
jmchung Aug 17, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 9 additions & 1 deletion sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/jsonExpressions.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -361,10 +361,18 @@ case class JsonTuple(children: Seq[Expression])
// the fields to query are the remaining children
@transient private lazy val fieldExpressions: Seq[Expression] = children.tail

// a field name given with constant null will be replaced with this pseudo field name
private val nullFieldName = "__NullFieldName"
Copy link
Member

@viirya viirya Aug 13, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A field name given with constant null will be replaced with this pseudo field name.

Copy link
Member

@HyukjinKwon HyukjinKwon Aug 13, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@jmchung, could we maybe compute this foldable related optimization ahead -
https://github.com/jmchung/spark/blob/ffa575a6731fef3e0731b73e0f7311cb024e831b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/jsonExpressions.scala#L425-L439 and remove this fake field name?

I think we can make a function for the above codes first and then use it for computation for each row. Did I understand correctly?

I tried a rough version I thought - https://github.com/jmchung/spark/compare/SPARK-21677...HyukjinKwon:tmp-18930?expand=1, @viirya what do you think about this?

Copy link
Member

@viirya viirya Aug 14, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, I've also considered using Option here. But don't want to come out Option version from me first, so we can experience review process. It looks good to me.

Copy link
Contributor Author

@jmchung jmchung Aug 14, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@HyukjinKwon @viirya Yep, we've discarded the fake field name and use Option here. We made a slight revision to deal with the None in foldableFieldNames instead of creating a new function.


// eagerly evaluate any foldable the field names
@transient private lazy val foldableFieldNames: IndexedSeq[String] = {
fieldExpressions.map {
case expr if expr.foldable => expr.eval().asInstanceOf[UTF8String].toString
case expr if expr.foldable =>
if (expr.eval() == null) {
nullFieldName
} else {
expr.eval().asInstanceOf[UTF8String].toString
}
case _ => null
}.toIndexedSeq
}
Expand Down
9 changes: 9 additions & 0 deletions sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/json/JsonSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2034,4 +2034,13 @@ class JsonSuite extends QueryTest with SharedSQLContext with TestJsonData {
}
}
}

test("SPARK-21677: json_tuple throws NullPointException when column is null as string type") {
Copy link
Member

@HyukjinKwon HyukjinKwon Aug 15, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could we move this to spark/sql/core/src/test/resources/sql-tests/inputs/json-functions.sql and/or spark/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/expressions/JsonExpressionsSuite.scala?

Copy link
Member

@gatorsmile gatorsmile Aug 15, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is just an end-to-end test case. We also need to add unit test cases in JsonExpressionsSuite

Copy link
Member

@viirya viirya Aug 16, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The end-to-end test at L2047 may not be able to move to JsonExpressionsSuite. We can have some unit test cases similar to L2039 in JsonExpressionsSuite as @gatorsmile suggested.

It is also good to have this end-to-end tests in json-functions.sql.

Copy link
Contributor Author

@jmchung jmchung Aug 16, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@gatorsmile has added unit test case in JsonExpressionsSuite
@viirya also add end-to-end test in json-functions.sql

checkAnswer(sql(
"""
|SELECT json_tuple('{"a" : 1, "b" : 2}'
|, cast(NULL AS STRING), 'b'
|, cast(NULL AS STRING), 'a')
""".stripMargin), Row(null, "2", null, "1"))
Copy link
Member

@viirya viirya Aug 15, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@jmchung Can we also add the test we discussed in slack which mixes constant field name and non constant one?

Copy link
Contributor Author

@jmchung jmchung Aug 15, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@viirya Done, the added test case contains column name, constant field name, and null field name.

Copy link
Member

@gatorsmile gatorsmile Aug 15, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nit: move Row(null, "2", null, "1")) to the next line.

Copy link
Contributor Author

@jmchung jmchung Aug 16, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok, thanks

}
}