lazy vector: An unloaded lazy vector cannot be wrapped by two different top level vectors in hash join

[zml1206]

Bug description

    withTable("t1", "t2") {
      Seq((1, "a"), (2, "b")).toDF("c1", "c2").write.saveAsTable("t1")
      Seq(1, 2, 3).toDF("c1").write.saveAsTable("t2")
      val query =
        """
          |select t3.* from
          |(select c1, c2 as a,c2 as b from t1) t3
          |join t2
          |on t3.c1 = t2.c1
          |""".stripMargin
      sql(query).show()
    }

ERROR:
An unloaded lazy vector cannot be wrapped by two different top level vectors.

The vector of c2 is rewrapped here, and there seems to be HashProbe do not load LazyVector before wrapping.

System information

Relevant logs

Reason: An unloaded lazy vector cannot be wrapped by two different top level vectors.
Retriable: False
Expression: dictionaryValues_->markAsContainingLazyAndWrapped()
Context: Operator: HashProbe[3] 2
Function: setInternalState
File: /opt/gluten/ep/build-velox/build/velox_ep/./velox/vector/DictionaryVector-inl.h
Line: 35
Stack trace:
# 0  _ZN8facebook5velox7process10StackTraceC1Ei
# 1  _ZN8facebook5velox14VeloxExceptionC1EPKcmS3_St17basic_string_viewIcSt11char_traitsIcEES7_S7_S7_bNS1_4TypeES7_
# 2  _ZN8facebook5velox6detail14veloxCheckFailINS0_17VeloxRuntimeErrorEPKcEEvRKNS1_18VeloxCheckFailArgsET0_
# 3  _ZN8facebook5velox16DictionaryVectorINS0_10StringViewEE16setInternalStateEv
# 4  _ZN8facebook5velox16DictionaryVectorINS0_10StringViewEEC1EPNS0_6memory10MemoryPoolEN5boost13intrusive_ptrINS0_6BufferEEEmSt10shared_ptrINS0_10BaseVectorEESA_RKNS0_17SimpleVectorStatsIS2_EESt8optionalIiESJ_SI_IbESJ_SJ_
# 5  _ZN8facebook5veloxL13addDictionaryILNS0_8TypeKindE7EEESt10shared_ptrINS0_10BaseVectorEEN5boost13intrusive_ptrINS0_6BufferEEES9_mS5_
# 6  _ZN8facebook5velox10BaseVector16wrapInDictionaryEN5boost13intrusive_ptrINS0_6BufferEEES5_iSt10shared_ptrIS1_Eb
# 7  _ZN8facebook5velox4exec9wrapChildEiN5boost13intrusive_ptrINS0_6BufferEEERKSt10shared_ptrINS0_10BaseVectorEES5_
# 8  _ZN8facebook5velox4exec9HashProbe10fillOutputEi
# 9  _ZN8facebook5velox4exec9HashProbe17getOutputInternalEb
# 10 _ZN8facebook5velox4exec9HashProbe9getOutputEv
# 11 _ZZN8facebook5velox4exec6Driver11runInternalERSt10shared_ptrIS2_ERS3_INS1_13BlockingStateEERS3_INS0_9RowVectorEEENKUlvE3_clEv
# 12 _ZN8facebook5velox4exec6Driver11runInternalERSt10shared_ptrIS2_ERS3_INS1_13BlockingStateEERS3_INS0_9RowVectorEE
# 13 _ZN8facebook5velox4exec6Driver4nextEPN5folly10SemiFutureINS3_4UnitEEE
# 14 _ZN8facebook5velox4exec4Task4nextEPN5folly10SemiFutureINS3_4UnitEEE
# 15 _ZN6gluten24WholeStageResultIterator4nextEv
# 16 Java_org_apache_gluten_vectorized_ColumnarBatchOutIterator_nativeHasNext
# 17 0x00007f97526413b0

[pedroerp]

Cc: @xiaoxmeng @tanjialiang

[leoluan2009]

We ran into the same issue, Any updates or progress so far? @pedroerp

[zml1206]

The SQL in the description previously added a filter by mistake, which made it impossible to reproduce. It has been modified.

[Yuhta]

This is a bug in optimizer, Velox does not allow duplicated projections in the physical plan.