Skip to content

[SPARK-20978][SQL] Set null for malformed column when the number of tokens is less than schema in CSV read/permissive mode #18200

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
HyukjinKwon wants to merge 1 commit into from
Closed

[SPARK-20978][SQL] Set null for malformed column when the number of tokens is less than schema in CSV read/permissive mode #18200

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
3 changes: 2 additions & 1 deletion sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/csv/UnivocityParser.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,7 +53,8 @@ class UnivocityParser(

// Retrieve the raw record string.
private def getCurrentInput: UTF8String = {
UTF8String.fromString(tokenizer.getContext.currentParsedContent().stripLineEnd)
UTF8String.fromString(
Option(tokenizer.getContext.currentParsedContent()).map(_.stripLineEnd).orNull)
Copy link
Member Author

@HyukjinKwon HyukjinKwon Jun 7, 2017 •
edited
Loading

Choose a reason for hiding this comment

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

BTW, this looks only called when PERMISSIVE mode when the malformed column is defined.

Copy link
Member

@maropu maropu Jun 7, 2017

Choose a reason for hiding this comment

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

Look so, is this something wrong?

Copy link
Member Author

@HyukjinKwon HyukjinKwon Jun 7, 2017

Choose a reason for hiding this comment

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

(Sorry, I meant just a note for a review ...)

Copy link
Member

@maropu maropu Jun 7, 2017

Choose a reason for hiding this comment

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

(oh, ... sorry)

}

// This parser first picks some tokens from the input tokens, according to the required schema,
Expand Down
8 changes: 8 additions & 0 deletions sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/csv/CSVSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1174,4 +1174,12 @@ class CSVSuite extends QueryTest with SharedSQLContext with SQLTestUtils {
}
}
}

test("SPARK-20978: Set null for malformed column when the number of tokens is less than schema") {
val df = spark.read
.schema("a string, b string, unparsed string")
.option("columnNameOfCorruptRecord", "unparsed")
.csv(Seq("a").toDS())
checkAnswer(df, Row("a", null, null))
Copy link
Member

@gatorsmile gatorsmile Jun 7, 2017

Choose a reason for hiding this comment

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

I did not read the code yet, but it looks like the result is wrong? We should output the a in the corrupted column.

In addition, this scenario works if we pass the csv files that contain less column values than the schema, right?

Copy link
Member

@maropu maropu Jun 7, 2017

Choose a reason for hiding this comment

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

Sorry for cutting in, but how to handle these longer/shorter cases is some arguable as @HyukjinKwon said in #16928 (comment). IMHO currently we just regard shorter cases as not-corrupted so as to keep existing behaviour (the previous Spark releases have regarded these case as not-corrupted, so other developers said that this behaviour change was not acceptable in earlier prs).

Copy link
Member

@gatorsmile gatorsmile Jun 8, 2017

Choose a reason for hiding this comment

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

Without this PR, you can try this example.

    val corruptCVSRecords: Dataset[String] =
      spark.createDataset(spark.sparkContext.parallelize(
        """1997, "b"""" ::
          """2015""" :: Nil))(Encoders.STRING)

    val df1 = spark.read
      .schema("col1 int, col2 String, __corrupted_column_name string")
      .option("columnNameOfCorruptRecord", "__corrupted_column_name")
      .option("mode", PermissiveMode.name)
      .csv(corruptCVSRecords)
    df1.show()

The output result of Permissive mode also treats the records with less columns as corrupted.

+----+----+-----------------------+
|col1|col2|__corrupted_column_name|
+----+----+-----------------------+
|1997| "b"|                   null|
|2015|null|                   2015|
+----+----+-----------------------+

}
}