Skip to content
This repository has been archived by the owner on Sep 18, 2023. It is now read-only.

tracking issues found from scala unit tests #207

Open
50 of 57 tasks
rui-mo opened this issue Mar 29, 2021 · 0 comments
Open
50 of 57 tasks

tracking issues found from scala unit tests #207

rui-mo opened this issue Mar 29, 2021 · 0 comments
Labels
bug Something isn't working

Comments

@rui-mo
Copy link
Collaborator

rui-mo commented Mar 29, 2021

  • case-when with multiple branches is not supported
  • null in in/inset should be specially handled
  • incorrect input cols in Expand
  • Aggregate should have a default result if all input batches are empty
  • EqualNullSafe should handle null correctly
  • Aggregate without grouping and aggregate expression should be supported
  • count literal on multiple cols should ignore null
  • NaNs or -0.0 and 0.0 should be in the same group
  • aggregate post-projection attributes mapping issue [SPARK-31620: agg with subquery (whole-stage-codegen =]
  • join: incorrect result in left/right outer join (BHJ, SHJ)
  • compilation failed in ExistenceJoin due to "not" operator in WSCG
  • NullPointerException in WSCG get metrics
  • incorrect result in semi join condition check
  • Max for NaN [NaN is greater than all other non-NaN numeric values]
  • precision loss of castVARCHAR in both gandiva and WSCG
  • some datatypes (BooleanType, DateType, StringType) may be not supported in Aggregate Actions
  • ColumnarInMemoryTableScanExec does not support row data as input
  • sum and avg should return null if all input values are null (currently zero) [SPARK-8828 sum should return null if all input values are null]
  • Aggregate (sum, avg...) for Literal should be fully supported. In this case, input batch is empty, and native calculation should be skipped. For now, only count_literal was supported. [cte-legacy.sql]
  • grouping for Literal is not supported
  • Count with condition was not supported. For now, aggregateExpression.filter was ignored. [Support filter clause for aggregate function with hash aggregate]
  • ColumnarConditionProjector: filter without project is not supported [SPARK-32788: non-partitioned table scan should not have partition filter]
  • ColumnarLike: escapeChar should be supported. [SPARK-33677: LikeSimplification should be skipped if pattern contains any escapeChar]
  • ColumnarLike: Failed to make LLVM module due to 'like' function requires a literal as the second parameter [like-all.sql]
  • SMJ: segfault [SPARK-25988: self join with aliases on partitioned tables]
  • SMJ: incorrect result in LeftAnti [scalar-subquery-select.sql]
  • In WSCG project, some expressions (not, equal, ...) needs to set "check_str_", otherwise compilation would fail. [many cases in sql test, eg, except.sql]
  • incorrect result in LeftAnti BHJ on null values in non-codegen verison caused by null_set [NOT IN predicate subquery]
  • incorrect result in LeftAnti BHJ on null values in codegen version
  • incorrect result in LeftSemi BHJ wo/ codegen caused by BooleanType is not correctly handled in HashRelationKernel [group-by.sql, ut: "groupby"]
  • Window: segfault [null inputs]
  • Window: KnownFloatingPointNormalized cannot be cast to org.apache.spark.sql.catalyst.expressions.AttributeReference [NaN and -0.0 in window partition keys]
  • Concat: different result [string concat]
  • decimalArithmeticOperations.sql: precision, handling for overflow and precision loss
  • ColumnarSorter: key not found in output attributes due to upper case and lower case [order-by-nulls-ordering.sql]
  • ConvertUtils, getAttrFromExpr: some other expressions should be handled [windowFrameCoercion.sql, postgreSQL/select_implicit.sql]
  • Aggregate group-by: a very small value (-1.2345678901234e-200) is regarded as the same as 0.0 in normalize function [union.sql]
  • exception is expected when int4 or int8 overflows [postgreSQL/int4.sql, postgreSQL/int8.sql]
  • NotImplemented: Function min_max has no kernel matching input types (array[date32[day]]) [subquery/scalar-subquery/scalar-subquery-predicate.sql]
  • max for bool has incorrect result
  • WSCG NullPointerException: s"ColumnarWSCG can't doCodeGen on ${child}". In ColumnarSortExec, doCodeGen: ColumnarCodegenContext = null [SPARK-34003: fix char/varchar fails w/ order by functions]
  • incorrect result in columnar Expand [cube]
  • nullOnDivideByZero should be supported in stddev
  • incorrect result for groupby bool
  • different result for Timestamp maybe due to timezone [in-order-by: different result for timestamp]
  • Sort doCodegn is null causing NPE
  • sort by literal
  • makeCopy, tree in ColumnarCollapseCodegenStages
  • Failed to make LLVM module due to Return type of root node int64 does not match that of expression timestamp[us, tz=UTC]
  • Not a valid date value 2014-31-12, to_date should return null instead of throwing exception [function to_date]
  • incorrect result in date and timestamp functions [DateFunctionsSuite]
  • SMJ segfault caused by case of field name
  • SMJ left/right outer result
  • runtime error in aggregate caused by the field name
  • "divide by zero" exception from gandiva in decimal divide
  • window has incorrect result ["two inner joins with condition"]
  • handling of Decimal overflow in Aggregation
@zhouyuan zhouyuan added the bug Something isn't working label Mar 30, 2021
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

2 participants