GideonPotok commented on code in PR #47154:
URL: https://github.com/apache/spark/pull/47154#discussion_r1689766242
##########
sql/core/src/test/scala/org/apache/spark/sql/CollationSQLExpressionsSuite.scala:
##########
@@ -1789,44 +1798,90 @@ class CollationSQLExpressionsSuite
s"named_struct('f2', collate('$elt', '${t.collationId}')), 'f3',
1)").mkString(",")
}.mkString(",")
- val tableName = s"t_${t.collationId}_mode_nested_struct"
+ val tableName = s"t_${t.collationId}_mode_nested_struct1"
withTable(tableName) {
sql(s"CREATE TABLE ${tableName}(i STRUCT<f1: STRUCT<f2: STRING COLLATE
" +
t.collationId + ">, f3: INT>) USING parquet")
sql(s"INSERT INTO ${tableName} VALUES " + valuesToAdd)
val query = s"SELECT lower(mode(i).f1.f2) FROM ${tableName}"
- if(t.collationId == "UTF8_LCASE" ||
- t.collationId == "unicode_ci" ||
- t.collationId == "unicode") {
- // Cannot resolve "mode(i)" due to data type mismatch:
- // Input to function mode was a complex type with strings collated
on non-binary
- // collations, which is not yet supported.. SQLSTATE: 42K09; line 1
pos 13;
- val params = Seq(("sqlExpr", "\"mode(i)\""),
- ("msg", "The input to the function 'mode' " +
- "was a type of binary-unstable type that is not currently
supported by mode."),
- ("hint", "")).toMap
- checkError(
- exception = intercept[AnalysisException] {
- sql(query)
- },
- errorClass = "DATATYPE_MISMATCH.TYPE_CHECK_FAILURE_WITH_HINT",
- parameters = params,
- queryContext = Array(
- ExpectedContext(objectType = "",
- objectName = "",
- startIndex = 13,
- stopIndex = 19,
- fragment = "mode(i)")
- )
- )
- } else {
- checkAnswer(sql(query), Row(t.result))
- }
+ checkAnswer(sql(query), Row(t.result))
}
})
}
test("Support mode for string expression with collated strings in array
complex type") {
+ case class ModeTestCase[R](collationId: String, bufferValues: Map[String,
Long], result: R)
+ val testCases = Seq(
+ ModeTestCase("utf8_binary", Map("a" -> 3L, "b" -> 2L, "B" -> 2L), "a"),
+ ModeTestCase("UTF8_LCASE", Map("a" -> 3L, "b" -> 2L, "B" -> 2L), "b"),
+ ModeTestCase("unicode", Map("a" -> 3L, "b" -> 2L, "B" -> 2L), "a"),
+ ModeTestCase("unicode_ci", Map("a" -> 3L, "b" -> 2L, "B" -> 2L), "b")
+ )
+ testCases.foreach(t => {
+ val valuesToAdd = t.bufferValues.map { case (elt, numRepeats) =>
+ (0L to numRepeats).map(_ => s"array(named_struct('f2', " +
+ s"collate('$elt', '${t.collationId}'), 'f3', 1))").mkString(",")
+ }.mkString(",")
+
+ val tableName = s"t_${t.collationId}_mode_nested_struct2"
+ withTable(tableName) {
+ sql(s"CREATE TABLE ${tableName}(" +
+ s"i ARRAY< STRUCT<f2: STRING COLLATE ${t.collationId}, f3: INT>>)" +
+ s" USING parquet")
+ sql(s"INSERT INTO ${tableName} VALUES " + valuesToAdd)
+ val query = s"SELECT lower(element_at(mode(i).f2, 1)) FROM
${tableName}"
+ checkAnswer(sql(query), Row(t.result))
+ }
+ })
+ }
+
+ test("Support mode for string expression with collated strings in 3D array
type") {
Review Comment:
@uros-db
How about I replace it with this, calling recursive functions within
getCollationAwareBuffer, like so:
```
private def getCollationAwareBuffer(
childDataType: DataType,
buffer: OpenHashMap[AnyRef, Long]): Iterable[(AnyRef, Long)] = {
def getBuffer(groupingFunction: AnyRef => _): Iterable[(AnyRef, Long)] =
{
buffer.groupMapReduce(t =>
groupingFunction(t._1))(x => x)((x, y) => (x._1, x._2 + y._2)).values
}
childDataType match {
// Short-circuit if there is no collation.
case _ if UnsafeRowUtils.isBinaryStable(child.dataType) => buffer
case c: StringType => getBuffer(k =>
CollationFactory.getCollationKey(k.asInstanceOf[UTF8String],
c.collationId))
case at: ArrayType => getBuffer(k =>
recursivelyGetBufferForArrayType(at, k.asInstanceOf[ArrayData]))
case st: StructType =>
getBuffer(k => recursivelyGetBufferForStructType(
k.asInstanceOf[InternalRow].toSeq(st).zip(st.fields)))
// Not supported: MapType
case _ => buffer
}
}
```
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
