[C++] PrimitiveArrays cause segmentation faults in PrettyPrinter

[sjperkins]

Describe the bug, including details regarding any error messages, version, and platform.

Arrays created with the PrimitiveArray(datatype, length, buffer) constructor aren't usable with PrettyPrinter. A workaround is to use the lower level arrow::MakeArray API.

This is likely low-priority or perhaps not an issue as the constructor isn't documented as part of the public API.

Here's a reproducer:

#include <iostream>
#include <memory>

#include <arrow/api.h>

using ::arrow::ArrayData;
using ::arrow::Buffer;
using ::arrow::PrimitiveArray;

const std::int64_t N = 1000;

int main(int argc, char * argv[]) {
  auto buffer = std::shared_ptr<arrow::Buffer>(arrow::AllocateBuffer(N * sizeof(double)).ValueOrDie());
  auto span = buffer->mutable_span_as<double>();
  for(std::size_t i=0; i<N; ++i) span[i] = double(i);

  if (true) {
    auto a = std::make_shared<PrimitiveArray>(arrow::float64(), N, buffer);
    // This is a valid array
    assert(a->ValidateFull().ok());
    // bad_cast from PrimiveArray to NumericArray<arrow::Double>
    // core dumps
    std::cout << a->ToString() << std::endl;
  }

  if (true) {
    auto buffers = std::vector<std::shared_ptr<Buffer>>{nullptr, buffer};
    auto data = std::make_shared<ArrayData>(arrow::float64(), N, std::move(buffers));
    auto a = arrow::MakeArray(data);
    // This is a valid array
    assert(a->ValidateFull().ok());
    std::cout << a->ToString() << std::endl;
  }

And an example of the backtrace it creates:

(gdb) bt
#0  __pthread_kill_implementation (no_tid=0, signo=6, threadid=<optimized out>)
    at ./nptl/pthread_kill.c:44
#1  __pthread_kill_internal (signo=6, threadid=<optimized out>)
    at ./nptl/pthread_kill.c:78
#2  __GI___pthread_kill (threadid=<optimized out>, signo=signo@entry=6)
    at ./nptl/pthread_kill.c:89
#3  0x0000790c64c4526e in __GI_raise (sig=sig@entry=6)
    at ../sysdeps/posix/raise.c:26
#4  0x0000790c64c288ff in __GI_abort () at ./stdlib/abort.c:79
#5  0x0000790c650d5fae in __gnu_cxx::__verbose_terminate_handler ()
    at ../../../../libstdc++-v3/libsupc++/vterminate.cc:95
#6  0x0000790c650d44f2 in __cxxabiv1::__terminate (handler=<optimized out>)
    at ../../../../libstdc++-v3/libsupc++/eh_terminate.cc:48
#7  0x0000790c650ce2f3 in std::terminate ()
    at ../../../../libstdc++-v3/libsupc++/eh_terminate.cc:58
#8  0x0000790c650d4712 in __cxxabiv1::__cxa_throw (obj=0x5e37424e6d60, 
    tinfo=0x790c651ef078 <typeinfo for std::bad_cast>, 
    dest=0x790c650d30fe <std::bad_cast::~bad_cast()>)
    at ../../../../libstdc++-v3/libsupc++/eh_throw.cc:98
#9  0x0000790c650ce248 in __cxxabiv1::__cxa_bad_cast ()
    at ../../../../libstdc++-v3/libsupc++/eh_aux_runtime.cc:34
#10 0x0000790c65746ef9 in arrow::internal::checked_cast<arrow::NumericArray<arrow::DoubleType> const&, arrow::Array const&> (value=...)
--Type <RET> for more, q to quit, c to continue without paging--
    at /home/simon/code/arrow/cpp/src/arrow/util/checked_cast.h:38
#11 0x0000790c65743484 in arrow::VisitArrayInline<arrow::(anonymous namespace)::ArrayPrinter> (array=..., visitor=0x7ffd5fd803d0)
    at /home/simon/code/arrow/cpp/src/arrow/visit_array_inline.h:55
#12 0x0000790c6573e6ea in arrow::(anonymous namespace)::ArrayPrinter::Print (this=0x7ffd5fd803d0, array=...) at /home/simon/code/arrow/cpp/src/arrow/pretty_print.cc:401
#13 0x0000790c6573ebe1 in arrow::PrettyPrint (arr=..., indent=0, sink=0x7ffd5fd80580) at /home/simon/code/arrow/cpp/src/arrow/pretty_print.cc:430
#14 0x0000790c65a9cf09 in arrow::Array::ToString[abi:cxx11]() const (this=0x5e37424e6a90) at /home/simon/code/arrow/cpp/src/arrow/array/array_base.cc:301
#15 0x00005e370f5cb456 in main ()

Component(s)

C++