[
https://issues.apache.org/jira/browse/SPARK-37575?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17455043#comment-17455043
]
Guo Wei edited comment on SPARK-37575 at 12/8/21, 8:40 AM:
-----------------------------------------------------------
I think I maybe have found the root cause through debug Spark source code.
In {color:#0747a6}UnivocityGenerator{color}, when the value of column is null
values, column's value has been changed to
{color:#00875a}options.nullValue{color}, default value is ""
{code:java}
private def convertRow(row: InternalRow): Seq[String] = {
var i = 0
val values = new Array[String](row.numFields)
while (i < row.numFields) {
if (!row.isNullAt(i)) {
values(i) = valueConverters(i).apply(row, i)
} else {
values(i) = options.nullValue
}
i += 1
}
values
} {code}
So，in {color:#0747a6}univocity-parsers lib{color}(depended by Spark)
{color:#0747a6}AbstractWriter{color} class, element( is original null values)
has been changed to '' in UnivocityGenerator，not satisfied condition(element
== null)，finally equals emptyValue, default value is "\"\""
{code:java}
protected String getStringValue(Object element) {
usingNullOrEmptyValue = false;
if (element == null) {
usingNullOrEmptyValue = true;
return nullValue;
}
String string = String.valueOf(element);
if (string.isEmpty()) {
usingNullOrEmptyValue = true;
return emptyValue;
}
return string;
} {code}
[~hyukjin.kwon] Should we fix the change(isNullAt) in
{color:#0747a6}UnivocityGenerator?{color}
was (Author: wayne guo):
I think I maybe have found the root cause through debug Spark source code.
In {color:#0747a6}UnivocityGenerator{color}, when the value of column is null
values, column's value has been changed to
{color:#00875a}options.nullValue{color}, default value is ""
{code:java}
private def convertRow(row: InternalRow): Seq[String] = {
var i = 0
val values = new Array[String](row.numFields)
while (i < row.numFields) {
if (!row.isNullAt(i)) {
values(i) = valueConverters(i).apply(row, i)
} else {
values(i) = options.nullValue
}
i += 1
}
values
} {code}
So，in {color:#0747a6}univocity-parsers lib{color}(depended by Spark)
{color:#0747a6}AbstractWriter{color} class, element( is original null values)
has been changed to '' in UnivocityGenerator，not satisfied condition(element
== null)，finally equals emptyValue, default value is "\"\""
{code:java}
protected String getStringValue(Object element) {
usingNullOrEmptyValue = false;
if (element == null) {
usingNullOrEmptyValue = true;
return nullValue;
}
String string = String.valueOf(element);
if (string.isEmpty()) {
usingNullOrEmptyValue = true;
return emptyValue;
}
return string;
} {code}
[~hyukjin.kwon] Should we fix the change(isNullAt) in
{color:#0747a6}UnivocityGenerator?{color}
> Empty strings and null values are both saved as quoted empty Strings ""
> rather than "" (for empty strings) and nothing(for null values)
> ---------------------------------------------------------------------------------------------------------------------------------------
>
> Key: SPARK-37575
> URL: https://issues.apache.org/jira/browse/SPARK-37575
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 2.4.0, 3.2.0
> Reporter: Guo Wei
> Priority: Major
>
> As mentioned in sql migration
> guide([https://spark.apache.org/docs/latest/sql-migration-guide.html#upgrading-from-spark-sql-23-to-24]),
> {noformat}
> Since Spark 2.4, empty strings are saved as quoted empty strings "". In
> version 2.3 and earlier, empty strings are equal to null values and do not
> reflect to any characters in saved CSV files. For example, the row of "a",
> null, "", 1 was written as a,,,1. Since Spark 2.4, the same row is saved as
> a,,"",1. To restore the previous behavior, set the CSV option emptyValue to
> empty (not quoted) string.{noformat}
>
> But actually, both empty strings and null values are saved as quoted empty
> Strings "" rather than "" (for empty strings) and nothing(for null values)。
> code:
> {code:java}
> val data = List("spark", null, "").toDF("name")
> data.coalesce(1).write.csv("spark_csv_test")
> {code}
> actual result:
> {noformat}
> line1: spark
> line2: ""
> line3: ""{noformat}
> expected result:
> {noformat}
> line1: spark
> line2:
> line3: ""
> {noformat}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
