I've hit a bug where splitAndTransfer on vectors throws if the vector is completely empty and the offset buffer is empty.
An easy repro is:
BufferAllocator allocator = new RootAllocator(Long.MAX_VALUE);
ListVector listVector = ListVector.empty("listVector", allocator);
listVector.getTransferPair(listVector.getAllocator()).splitAndTransfer(0, 0);
This results in the following stacktrace:
java.lang.IndexOutOfBoundsException: index: 0, length: 4 (expected: range(0, 0))
at io.netty.buffer.ArrowBuf.checkIndexD(ArrowBuf.java:335)
at io.netty.buffer.ArrowBuf.chk(ArrowBuf.java:322)
at io.netty.buffer.ArrowBuf.getInt(ArrowBuf.java:441)
at org.apache.arrow.vector.complex.ListVector$TransferImpl.splitAndTransfer(ListVector.java:484)
In production we hit this when calling VectorSchemaRoot.slice. The schema root contains a ListVector with a VarCharVector value vector. The list vector isn't empty, but all the strings in the var char vector are. splitAndTransfer on the list vector works, but then when underlying var char vector is split we get the same exception:
java.lang.IndexOutOfBoundsException: index: 0, length: 4 (expected: range(0, 0))
at io.netty.buffer.ArrowBuf.checkIndexD(ArrowBuf.java:335)
at io.netty.buffer.ArrowBuf.chk(ArrowBuf.java:322)
at io.netty.buffer.ArrowBuf.getInt(ArrowBuf.java:441)
at org.apache.arrow.vector.BaseVariableWidthVector.splitAndTransferOffsetBuffer(BaseVariableWidthVector.java:728)
at org.apache.arrow.vector.BaseVariableWidthVector.splitAndTransferTo(BaseVariableWidthVector.java:712)
at org.apache.arrow.vector.VarCharVector$TransferImpl.splitAndTransfer(VarCharVector.java:321)
at org.apache.arrow.vector.complex.ListVector$TransferImpl.splitAndTransfer(ListVector.java:496)
at org.apache.arrow.vector.VectorSchemaRoot.lambda$slice$1(VectorSchemaRoot.java:308)
at java.base/java.util.stream.ReferencePipeline$3$1.accept(ReferencePipeline.java:195)
at java.base/java.util.ArrayList$ArrayListSpliterator.forEachRemaining(ArrayList.java:1655)
at java.base/java.util.stream.AbstractPipeline.copyInto(AbstractPipeline.java:484)
at java.base/java.util.stream.AbstractPipeline.wrapAndCopyInto(AbstractPipeline.java:474)
at java.base/java.util.stream.ReduceOps$ReduceOp.evaluateSequential(ReduceOps.java:913)
at java.base/java.util.stream.AbstractPipeline.evaluate(AbstractPipeline.java:234)
at java.base/java.util.stream.ReferencePipeline.collect(ReferencePipeline.java:578)
at org.apache.arrow.vector.VectorSchemaRoot.slice(VectorSchemaRoot.java:310)
Reporter: David Vogelbacher
PRs and other links:
Note: This issue was originally created as ARROW-15382. Please see the migration documentation for further details.
I've hit a bug where
splitAndTransferon vectors throws if the vector is completely empty and the offset buffer is empty.An easy repro is:
This results in the following stacktrace:
In production we hit this when calling
VectorSchemaRoot.slice. The schema root contains aListVectorwith aVarCharVectorvalue vector. The list vector isn't empty, but all the strings in the var char vector are.splitAndTransferon the list vector works, but then when underlying var char vector is split we get the same exception:Reporter: David Vogelbacher
PRs and other links:
Note: This issue was originally created as ARROW-15382. Please see the migration documentation for further details.