Github user rxin commented on a diff in the pull request:
https://github.com/apache/spark/pull/13990#discussion_r69676002
--- Diff:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/complexTypeCreator.scala
---
@@ -393,3 +393,71 @@ case class CreateNamedStructUnsafe(children:
Seq[Expression]) extends Expression
override def prettyName: String = "named_struct_unsafe"
}
+
+/**
+ * Creates a map after splitting the input text into key/value pairs using
delimeters
+ */
+@ExpressionDescription(
+ usage = """_FUNC_(text[, pairDelim, keyValueDelim]) - Creates a map
after splitting the text into
+ key/value pairs using delimiters.
+ Default delimiters are ',' for pairDelim and '=' for keyValueDelim.""")
+case class StringToMap(text: Expression, pairDelim: Expression,
keyValueDelim: Expression)
+ extends TernaryExpression with ExpectsInputTypes {
+
+ def this(child: Expression) = {
+ this(child, Literal(","), Literal("="))
+ }
+
+ override def children: Seq[Expression] = Seq(text, pairDelim,
keyValueDelim)
+
+ override def inputTypes: Seq[AbstractDataType] = Seq(StringType,
StringType, StringType)
+
+ override def dataType: DataType = MapType(StringType, StringType,
valueContainsNull = false)
+
+ override def nullSafeEval(str: Any, delim1: Any, delim2: Any): Any = {
+ val array = str.asInstanceOf[UTF8String]
+ .split(delim1.asInstanceOf[UTF8String], -1)
+ .map{_.split(delim2.asInstanceOf[UTF8String], 2)}
+
+ ArrayBasedMapData(array.map(_(0)),
array.map(_(1))).asInstanceOf[MapData]
+ }
+
+ override def doGenCode(ctx: CodegenContext, ev: ExprCode): ExprCode = {
+
+ nullSafeCodeGen(ctx, ev, (text, delim1, delim2) => {
+ val arrayClass = classOf[GenericArrayData].getName
+ val mapClass = classOf[ArrayBasedMapData].getName
+ val keyArray = ctx.freshName("keyArray")
+ val valueArray = ctx.freshName("valueArray")
+ ctx.addMutableState("UTF8String[]", keyArray, s"this.$keyArray =
null;")
+ ctx.addMutableState("UTF8String[]", valueArray, s"this.$valueArray =
null;")
--- End diff --
Why don't we remove code gen from this. It is making the pull request more
complicated, and in general I'm not sure how often these functions will be
called, so the value of codegen isn't great. Plus, str_to_map is a pretty
expensive function anyway.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
