zhengruifeng commented on code in PR #47742:
URL: https://github.com/apache/spark/pull/47742#discussion_r1716150562
##########
mllib/src/main/scala/org/apache/spark/ml/feature/StringIndexer.scala:
##########
@@ -195,31 +193,27 @@ class StringIndexer @Since("1.4.0") (
} else {
// We don't count for NaN values. Because `StringIndexerAggregator`
only processes strings,
// we replace NaNs with null in advance.
- new Column(If(col.isNaN.expr, Literal(null),
col.expr)).cast(StringType)
+ when(!col.isNaN, col.cast(StringType)).otherwise(lit(null))
}
}
}
private def countByValue(
dataset: Dataset[_],
- inputCols: Array[String]): Array[OpenHashMap[String, Long]] = {
+ inputCols: Array[String]): Seq[Map[String, Long]] = {
+ val selectedCols = getSelectedCols(dataset,
inputCols.toImmutableArraySeq).map(grouped_count)
- val aggregator = new StringIndexerAggregator(inputCols.length)
- implicit val encoder = Encoders.kryo[Array[OpenHashMap[String, Long]]]
-
- val selectedCols = getSelectedCols(dataset, inputCols.toImmutableArraySeq)
- dataset.select(selectedCols: _*)
- .toDF()
- .agg(aggregator.toColumn)
- .as[Array[OpenHashMap[String, Long]]]
- .collect()(0)
+ dataset
Review Comment:
Great! It require the datatypes of all columns to be the same, but in
existing two places (ml and pyspark) the datatyes can be cast to the same one,
so in this way we don't need a new expression then.
Thanks!
--
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]
