Python integration via reticulate

These functions enable Python wrapper objects created via reticulate to be used with any function that uses as_nanoarrow_array() or as_nanoarrow_array_stream() to accept generic "arrowable" input. Implementations for reticulate::py_to_r() and reticulate::r_to_py() are also included such that nanoarrow's array/schema/array stream objects can be passed as arguments to Python functions that would otherwise accept an object implementing the Arrow PyCapsule protocol.

Usage

# S3 method for class 'python.builtin.object'
as_nanoarrow_schema(x, ...)

# S3 method for class 'python.builtin.object'
as_nanoarrow_array(x, ..., schema = NULL)

# S3 method for class 'python.builtin.object'
as_nanoarrow_array_stream(x, ..., schema = NULL)

test_reticulate_with_nanoarrow()

Arguments

x

An Python object to convert

...

Unused

schema

A requested schema, which may or may not be honoured depending on the capabilities of the producer

Value

• as_nanoarrow_schema() returns an object of class nanoarrow_schema

• as_nanoarrow_array() returns an object of class nanoarrow_array

• as_nanoarrow_array_stream() returns an object of class nanoarrow_array_stream.

Details

This implementation uses the Arrow PyCapsule protocol to interpret an arbitrary Python object as an Arrow array/schema/array stream and produces Python objects that implement this protocol. This is currently implemented using the nanoarrow Python package which provides similar primitives for facilitating interchange in Python.

Examples

if (FALSE) { # test_reticulate_with_nanoarrow()
library(reticulate)

py_require("nanoarrow")

na <- import("nanoarrow", convert = FALSE)
python_arrayish_thing <- na$Array(1:3, na_int32())
as_nanoarrow_array(python_arrayish_thing)

r_to_py(as_nanoarrow_array(1:3))
}