fqaiser94 commented on a change in pull request #29795:
URL: https://github.com/apache/spark/pull/29795#discussion_r494700403
##########
File path: sql/core/src/main/scala/org/apache/spark/sql/Column.scala
##########
@@ -901,39 +901,125 @@ class Column(val expr: Expression) extends Logging {
* // result: org.apache.spark.sql.AnalysisException: Ambiguous reference
to fields
* }}}
*
+ * This method supports adding/replacing nested fields directly e.g.
+ *
+ * {{{
+ * val df = sql("SELECT named_struct('a', named_struct('a', 1, 'b', 2))
struct_col")
+ * df.select($"struct_col".withField("a.c", lit(3)).withField("a.d",
lit(4)))
+ * // result: {"a":{"a":1,"b":2,"c":3,"d":4}}
+ * }}}
+ *
+ * However, if you are going to add/replace multiple nested fields, it is
more optimal to extract
+ * out the nested struct before adding/replacing multiple fields e.g.
+ *
+ * {{{
+ * val df = sql("SELECT named_struct('a', named_struct('a', 1, 'b', 2))
struct_col")
+ * df.select($"struct_col".withField("a", $"struct_col.a".withField("c",
lit(3)).withField("d", lit(4))))
+ * // result: {"a":{"a":1,"b":2,"c":3,"d":4}}
+ * }}}
+ *
* @group expr_ops
* @since 3.1.0
*/
// scalastyle:on line.size.limit
def withField(fieldName: String, col: Column): Column = withExpr {
require(fieldName != null, "fieldName cannot be null")
require(col != null, "col cannot be null")
+ updateFieldsHelper(expr, nameParts(fieldName), name => WithField(name,
col.expr))
+ }
- val nameParts = if (fieldName.isEmpty) {
+ // scalastyle:off line.size.limit
+ /**
+ * An expression that drops fields in `StructType` by name.
Review comment:
I've made this change but now that I think about it, I don't think its
actually classifies as a "noop". We still reconstruct the struct unfortunately
e.g.
```
val structType = StructType(Seq(
StructField("a", IntegerType, nullable = false),
StructField("b", IntegerType, nullable = true),
StructField("c", IntegerType, nullable = false)))
val structLevel1: DataFrame = spark.createDataFrame(
sparkContext.parallelize(Row(Row(1, null, 3)) :: Nil),
StructType(Seq(StructField("a", structType, nullable = false))))
structLevel1.withColumn("a", 'a.dropFields("d")).explain()
== Physical Plan ==
*(1) Project [named_struct(a, a#1.a, b, a#1.b, c, a#1.c) AS a#3]
+- *(1) Scan ExistingRDD[a#1]
```
Should I revert this?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
