tests/testthat/test_r_torch_share_objects.R

In rTorch: R Bindings to 'PyTorch'

library(testthat)

source("helper_utils.R")

skip_on_cran()
# skip_if_no_python()

context("print R values from within Python")

# these objects have to be global to be able to be seen from the test
A <<- 100
B <<- 250

Ap <<- r_to_py(A)
Bp <<- r_to_py(B)


test_that("R objects are multiplied within Python and printed", {
  result <- py_run_string("ab = r.A * r.B")  # multiply to R objects in Python
  expect_equal(result$ab, 25000)             # extract from the list/dictionary

  expect_equal(py_run_string("")$ab, 25000)     # return value in ab
  expect_equal(py_eval("ab"), 25000)
})

test_that("R objects are detected within Python",  {
  result <- py_run_string("ab")
  # print(names(result))
  expect_true(any(names(result) %in% c("sys", "R", "r", "ab")))
  expect_equal(py_eval("r.Ap"), 100)
  expect_equal(py_eval("r.Bp"), 250)
})


test_that("multiplication result returns via the py object", {
  expect_equal(py$ab, 25000)
  py_run_string("del ab") # remove Python object for the other tests
})




context("R and Python share variables")

test_that("Python and R share variables", {
  expr <- py_run_string("
import numpy as np
x = 10
y = 20
# we don't want to print
# print(x)
# print(y)
")
  expect_equal(py$x, 10)
  expect_equal(py$y, 20)

  # TODO: how do we get the R values in Python
  # In notebooks is easy but how do we do it in R scripts?
  X <<- 123    # have to make it global in unit tests
  Y <<- 456    # have to make it global in unit tests
  expect_equal(py_eval("r.X"), 123)
  expect_equal(py_eval("r.Y"), 456)

  A <<- 100.123
  B <<- 200.2

  # convert r to python
  # Ap <- r_to_py(A)
  # Bp <- r_to_py(B)
  # py_run_string("print(r.A)") # error:  error: object 'A' not found.
  expect_equal(py_eval("r.A"), 100.123)
  expect_equal(py_eval("r.B"), 200.2)


  # py_run_string("print(r['A'])") # untimeEvaluation error: object 'A' not found.
  # A <- r_to_py(A)
  # py_run_string("print(A)")
  # py_run_string("import numpy as np")
  # py_run_string("my_python_array = np.array([2,4,6,8])")
  # py_run_string("for item in my_python_array: print(item)")
  # py_eval("print(dir(r))")

  obj <- py_run_string("r.__dir__()")
  # print(class(obj))
  # print(obj$keys())
  # print(names(obj))
  expect_all_true(names(obj) %in% c("np", "r", "R", "sys", "x", "y"))
  # print(names(py_to_r(obj)))

  # after removing two Python objects
  py_run_string("del x")
  py_run_string("del y")
  expect_all_true(names(obj) %in% c("np", "r", "R", "sys"))
})