[SPARK-23588][SQL] CatalystToExternalMap should support interpreted execution #20979
Conversation
| private lazy val toScalaValue: Any => Any = { | ||
| assert(inputData.dataType.isInstanceOf[MapType]) | ||
| val mapType = inputData.dataType.asInstanceOf[MapType] | ||
| CatalystTypeConverters.createToScalaConverter(mapType) |
There was a problem hiding this comment.
hmm, don't we need to support different types of resulting collection specified by collClass?
There was a problem hiding this comment.
yea, you're right. I'll update soon.
I checked the related code; it seems the input type is always the same with the output type specified by collClass? ex.)
scala> Seq(Map(1 -> 1, 2 -> 2)).toDS.as[Map[Long, String]].map(y => y).explain(true)
== Physical Plan ==
*(1) SerializeFromObject [externalmaptocatalyst(ExternalMapToCatalyst_key34, false, LongType, lambdavariable(ExternalMapToCatalyst_key34, false, LongType, false), ExternalMapToCatalyst_value34, ExternalMapToCatalyst_value_isNull34, ObjectType(class java.lang.String), staticinvoke(class org.apache.spark.unsafe.types.UTF8String, StringType, fromString, lambdavariable(ExternalMapToCatalyst_value34, ExternalMapToCatalyst_value_isNull34, ObjectType(class java.lang.String), true), true, false), input[0, scala.collection.immutable.Map, true]) AS value#137]
+- *(1) MapElements <function1>, obj#136: scala.collection.immutable.Map
+- *(1) DeserializeToObject catalysttoexternalmap(CatalystToExternalMap_keyLoopValue33, lambdavariable(CatalystToExternalMap_keyLoopValue33, , LongType, false), CatalystToExternalMap_valueLoopValue33, CatalystToExternalMap_valueLoopIsNull33, lambdavariable(CatalystToExternalMap_valueLoopValue33, CatalystToExternalMap_valueLoopIsNull33, StringType, true).toString, cast(value#130 as map<bigint,string>), interface scala.collection.immutable.Map, MapType(LongType,StringType,true)), obj#135: scala.collection.immutable.Map
+- LocalTableScan [value#130]
deserializerFor adds casts implicitly?
There was a problem hiding this comment.
I am not sure if I buy that argument. This expression can produce a map type other than scala.collection.immutable.Map, so we should either implement this, or we should remove the ability to produce different map types.
|
Test build #88925 has finished for PR 20979 at commit
|
|
retest this please |
|
Test build #88935 has finished for PR 20979 at commit
|
|
Test build #89201 has finished for PR 20979 at commit
|
|
Test build #89548 has finished for PR 20979 at commit
|
| private lazy val valueConverter = | ||
| CatalystTypeConverters.createToScalaConverter(inputMapType.valueType) | ||
|
|
||
| private def newMapBuilder(): Builder[AnyRef, AnyRef] = { |
There was a problem hiding this comment.
Can you make sure the builder only get's resolved once per expression instead of once per row. Can you address this in a follow-up?
|
Disregard my earlier comment. Merging to master. Thanks! Can you address my last comment in a follow-up? |
|
Test build #89567 has finished for PR 20979 at commit
|
|
ok, thanks! |
…ion in CatalystToExternalMap ## What changes were proposed in this pull request? This pr is a follow-up pr of apache#20979 and fixes code to resolve a map builder method per execution instead of per row in `CatalystToExternalMap`. ## How was this patch tested? Existing tests. Author: Takeshi Yamamuro <[email protected]> Closes apache#21112 from maropu/SPARK-23588-FOLLOWUP.
What changes were proposed in this pull request?
This pr supported interpreted mode for
CatalystToExternalMap.How was this patch tested?
Added tests in
ObjectExpressionsSuite.