cloud-fan commented on code in PR #48908:
URL: https://github.com/apache/spark/pull/48908#discussion_r1889766930
##########
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/jsonExpressions.scala:
##########
@@ -502,111 +496,119 @@ case class JsonTuple(children: Seq[Expression])
}
}
+ @transient
+ private lazy val evaluator: JsonTupleEvaluator =
JsonTupleEvaluator(fieldExpressions.length)
+
override def eval(input: InternalRow): IterableOnce[InternalRow] = {
val json = jsonExpr.eval(input).asInstanceOf[UTF8String]
- if (json == null) {
- return nullRow
- }
- try {
- /* We know the bytes are UTF-8 encoded. Pass a Reader to avoid having
Jackson
- detect character encoding which could fail for some malformed strings */
- Utils.tryWithResource(CreateJacksonParser.utf8String(jsonFactory, json))
{ parser =>
- parseRow(parser, input)
- }
- } catch {
- case _: JsonProcessingException =>
- nullRow
- }
- }
-
- private def parseRow(parser: JsonParser, input: InternalRow):
Seq[InternalRow] = {
- // only objects are supported
- if (parser.nextToken() != JsonToken.START_OBJECT) {
- return nullRow
- }
-
- // evaluate the field names as String rather than UTF8String to
- // optimize lookups from the json token, which is also a String
+ // Evaluate the field names as String rather than UTF8String to
+ // optimize lookups from the json token, which is also a String.
val fieldNames = if (constantFields == fieldExpressions.length) {
Review Comment:
Can we simplify things if we don't have the constant field names
optimization? e.g. always pass field names to the evaluator as array of string.
--
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]
