[
https://issues.apache.org/jira/browse/SPARK-32796?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
L. C. Hsieh updated SPARK-32796:
--------------------------------
Description:
Currently {{Column.withField}} only supports {{StructType}}. For nested struct
in {{ArrayType}}, it doesn't support. We can support {{ArrayType}} to make the
API more general and useful.
It could significantly simplify how we add/replace deeply nested fields.
For example,
{code}
private lazy val arrayType = ArrayType(
StructType(Seq(
StructField("a", IntegerType, nullable = false),
StructField("b", IntegerType, nullable = true),
StructField("c", IntegerType, nullable = false))),
containsNull = true)
private lazy val arrayStructArrayLevel1: DataFrame = spark.createDataFrame(
sparkContext.parallelize(Row(Array(Row(Array(Row(1, null, 3)), null, 3)))
:: Nil),
StructType(
Seq(StructField("a", ArrayType(
StructType(Seq(
StructField("a", arrayType, nullable = false),
StructField("b", IntegerType, nullable = true),
StructField("c", IntegerType, nullable = false))),
containsNull = false)))))
{code}
The data looks like:
{code}
+---------------------------+
|a |
+---------------------------+
|[{[{1, null, 3}], null, 3}]|
+---------------------------+
{code}
In order to replace deeply nested b column, like:
{code}
+------------------------+
|a |
+------------------------+
|[{[{1, 2, 3}], null, 3}]|
+------------------------+
{code}
Currently by using transform + withField, we probably need:
{code}
arrayStructArrayLevel1.withColumn("a",
transform($"a", _.withField("a",
flatten(transform($"a.a", transform(_, _.withField("b", lit(2))))))))
{code}
Using modified withField, we can do it like:
{code}
arrayStructArrayLevel1.withColumn("a", $"a".withField("a.b", lit(2)))
{code}
was:
Currently {{Column.withField}} only supports {{StructType}}. For nested struct
in {{ArrayType}}, it doesn't support. We can support {{ArrayType}} to make the
API more general and useful.
> Make withField API support nested struct in array
> -------------------------------------------------
>
> Key: SPARK-32796
> URL: https://issues.apache.org/jira/browse/SPARK-32796
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Affects Versions: 3.1.0
> Reporter: L. C. Hsieh
> Assignee: L. C. Hsieh
> Priority: Major
>
> Currently {{Column.withField}} only supports {{StructType}}. For nested
> struct in {{ArrayType}}, it doesn't support. We can support {{ArrayType}} to
> make the API more general and useful.
> It could significantly simplify how we add/replace deeply nested fields.
> For example,
> {code}
> private lazy val arrayType = ArrayType(
> StructType(Seq(
> StructField("a", IntegerType, nullable = false),
> StructField("b", IntegerType, nullable = true),
> StructField("c", IntegerType, nullable = false))),
> containsNull = true)
> private lazy val arrayStructArrayLevel1: DataFrame = spark.createDataFrame(
> sparkContext.parallelize(Row(Array(Row(Array(Row(1, null, 3)), null, 3)))
> :: Nil),
> StructType(
> Seq(StructField("a", ArrayType(
> StructType(Seq(
> StructField("a", arrayType, nullable = false),
> StructField("b", IntegerType, nullable = true),
> StructField("c", IntegerType, nullable = false))),
> containsNull = false)))))
> {code}
> The data looks like:
> {code}
> +---------------------------+
> |a |
> +---------------------------+
> |[{[{1, null, 3}], null, 3}]|
> +---------------------------+
> {code}
> In order to replace deeply nested b column, like:
> {code}
> +------------------------+
> |a |
> +------------------------+
> |[{[{1, 2, 3}], null, 3}]|
> +------------------------+
> {code}
> Currently by using transform + withField, we probably need:
> {code}
> arrayStructArrayLevel1.withColumn("a",
> transform($"a", _.withField("a",
> flatten(transform($"a.a", transform(_, _.withField("b", lit(2))))))))
> {code}
> Using modified withField, we can do it like:
> {code}
> arrayStructArrayLevel1.withColumn("a", $"a".withField("a.b", lit(2)))
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
