HyukjinKwon commented on a change in pull request #27366: [SPARK-30648][SQL]
Support filters pushdown in JSON datasource
URL: https://github.com/apache/spark/pull/27366#discussion_r373465776
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/json/JacksonParser.scala
##########
@@ -334,15 +336,19 @@ class JacksonParser(
private def convertObject(
parser: JsonParser,
schema: StructType,
- fieldConverters: Array[ValueConverter]): InternalRow = {
+ fieldConverters: Array[ValueConverter],
+ structFilters: StructFilters = new NoopFilters()): Option[InternalRow] =
{
val row = new GenericInternalRow(schema.length)
var badRecordException: Option[Throwable] = None
+ var skipRow = false
- while (nextUntil(parser, JsonToken.END_OBJECT)) {
+ structFilters.reset()
+ while (!skipRow && nextUntil(parser, JsonToken.END_OBJECT)) {
schema.getFieldIndex(parser.getCurrentName) match {
case Some(index) =>
try {
row.update(index, fieldConverters(index).apply(parser))
+ skipRow = structFilters.skipRow(row, index)
Review comment:
Hm, if I didn't misread, this wouldn't work properly (with
https://github.com/apache/spark/pull/27366/files#diff-635e02b2d1ce4ad1675b0350ccac0c10R266)
if we happen to push down predicates that refer nested columns.
Seems a bit odd because the previous code didn't have such assumption before
and all codes were valid for both root and nested rows. Now this method works
differently for root row and nested row. I wonder if we can decouple the
filtering logic out of this method ...
----------------------------------------------------------------
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]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
