Creating Arrow Objects

Recipes related to the creation of Arrays, Tables, Tensors and all other Arrow entities.

Create Arrays from Standard C++

Typed subclasses of arrow::ArrayBuilder make it easy to efficiently create Arrow arrays from existing C++ data:

Creating an array from C++ primitives
arrow::Int32Builder builder;
ASSERT_OK(builder.Append(1));
ASSERT_OK(builder.Append(2));
ASSERT_OK(builder.Append(3));
ASSERT_OK_AND_ASSIGN(std::shared_ptr<arrow::Array> arr, builder.Finish())
rout << arr->ToString() << std::endl;
Code Output
[
  1,
  2,
  3
]

Note

Builders will allocate data as needed and insertion should have constant amortized time.

Builders can also consume standard C++ containers:

// Raw pointers
arrow::Int64Builder long_builder = arrow::Int64Builder();
std::array<int64_t, 4> values = {1, 2, 3, 4};
ASSERT_OK(long_builder.AppendValues(values.data(), values.size()));
ASSERT_OK_AND_ASSIGN(arr, long_builder.Finish());
rout << arr->ToString() << std::endl;

// Vectors
arrow::StringBuilder str_builder = arrow::StringBuilder();
std::vector<std::string> strvals = {"x", "y", "z"};
ASSERT_OK(str_builder.AppendValues(strvals));
ASSERT_OK_AND_ASSIGN(arr, str_builder.Finish());
rout << arr->ToString() << std::endl;

// Iterators
arrow::DoubleBuilder dbl_builder = arrow::DoubleBuilder();
std::set<double> dblvals = {1.1, 1.1, 2.3};
ASSERT_OK(dbl_builder.AppendValues(dblvals.begin(), dblvals.end()));
ASSERT_OK_AND_ASSIGN(arr, dbl_builder.Finish());
rout << arr->ToString() << std::endl;
Code Output
[
  1,
  2,
  3,
  4
]
[
  "x",
  "y",
  "z"
]
[
  1.1,
  2.3
]

Note

Builders will not take ownership of data in containers and will make a copy of the underlying data.