def test_gentle_asarray_array_input():
"""
Test gentle_asarray with various forms of np.ndarray input,
which may need to be cast to a different data type.
"""
# Array that already bears the correct dtype:
inp = np.array([1, 2, 3, 4], dtype=np.int16)
result = util.gentle_asarray(inp, dtype=np.int16)
assert result.dtype == np.int16
assert_array_equal(result, inp)
# This will require casting, since int8 != int16:
result = util.gentle_asarray(inp, dtype=np.int8)
assert result.dtype == np.int8
assert_array_equal(result, inp)
# Test broadcasting the same 1D array to a recarray with
# two columns. The input array should be copied into
# each column and cast as necessary:
dtype = np.dtype([("col1", np.int16), ("col2", np.float32)])
result = util.gentle_asarray(inp, dtype=dtype)
assert result.dtype == dtype
assert result["col1"].dtype == np.int16
assert_array_equal(result["col1"], np.array([1, 2, 3, 4], dtype=np.int16))
assert result["col2"].dtype == np.float32
assert_array_equal(result["col1"], np.array([1, 2, 3, 4],
dtype=np.float32))
def test_gentle_asarray_invalid_conversion():
"""
Test gentle_asarray with input that cannot be cast
to an array.
"""
with pytest.raises(ValueError):
util.gentle_asarray(object(), dtype=np.float32)
def test_gentle_asarray_mismatched_field_names():
"""
Test gentle_asarray with input field names that don't
match the requested output field names.
"""
in_dtype = np.dtype([("col1", np.int16), ("col2", np.float32)])
inp = np.array([1, 2, 3, 4], dtype=in_dtype)
out_dtype = np.dtype([("col1", np.int16), ("foo", np.float32)])
with pytest.raises(ValueError):
util.gentle_asarray(inp, dtype=out_dtype)
def test_gentle_asarray_scalar_input():
"""
Test gentle_asarray with scalar input.
"""
result = util.gentle_asarray(3.14159, dtype=np.float32)
assert result.dtype == np.float32
assert_array_equal(result, np.array([3.14159], dtype=np.float32))
def test_gentle_asarray():
x = np.array([('abc', 1.0)], dtype=[('FOO', 'S3'), ('BAR', '>f8')])
new_dtype = [('foo', '|S3'), ('bar', '<f8')]
y = util.gentle_asarray(x, new_dtype)
assert y['BAR'][0] == 1.0
def test_gentle_asarray_fits_rec_input():
"""
Test gentle_asarray with FITS_rec array input.
"""
# 'e' is the FITS data type code for single-precision float:
cols = [fits.Column("col1", format="e", array=np.array([1, 2, 3, 4]))]
inp = fits.FITS_rec.from_columns(cols)
out_dtype = np.dtype([("col1", np.float32)])
result = util.gentle_asarray(inp, dtype=out_dtype)
assert result.dtype == out_dtype
assert_array_equal(result, inp)
def test_gentle_asarray_field_name_case():
"""
Test gentle_assary with input and output field names that
only differ by case.
"""
in_dtype = np.dtype([("col1", np.int16), ("col2", np.float32)])
inp = np.array([1, 2, 3, 4], dtype=in_dtype)
out_dtype = np.dtype([("COL1", np.int16), ("COL2", np.float32)])
result = util.gentle_asarray(inp, dtype=out_dtype)
assert result.dtype == out_dtype
assert_array_equal(result, inp)
def test_gentle_asarray_nested_array():
"""
Test gentle_asarray with nested arrays in one of the record
fields.
"""
# '2f' is a numpy data type code for an array of two 32-bit floats:
in_dtype = np.dtype([("col1", np.dtype("2f")), ("col2", np.int16)])
inp = np.array([1, 2, 3, 4], dtype=in_dtype)
out_dtype = np.dtype([("col1", np.dtype("2f")), ("col2", np.int8)])
result = util.gentle_asarray(inp, dtype=out_dtype)
assert result.dtype == out_dtype
assert_array_equal(result["col2"], inp["col2"])
def test_gentle_asarray_recarray_input():
"""
Test gentle_asarray input that is already an np.recarray.
"""
# This is just the simple case where the dtype already matches
# the desired output, but it follows a different code path
# from array input:
dtype = np.dtype([("col1", np.int16), ("col2", np.float32)])
inp = np.array([1, 2, 3, 4], dtype=dtype)
result = util.gentle_asarray(inp, dtype=dtype)
assert result.dtype == dtype
assert_array_equal(result, inp)
def test_gentle_asarray_fits_rec_pseudo_unsigned():
"""
Test gentle_asarray handling of a FITS_rec with a pseudo unsigned
integer column, which is a special case due to a bug in astropy.
"""
# 'j' is the FITS data type code for 32-bit integer. A column
# is only considered "pseudo unsigned" for certain values of
# bezero, one of which is 2**31.
cols = [fits.Column("col1", format="j", array=np.array([1, 2, 3, 4], np.uint32), bzero=2**31)]
inp = fits.FITS_rec.from_columns(cols)
out_dtype = np.dtype([("col1", np.uint32)])
result = util.gentle_asarray(inp, dtype=out_dtype)
# If not handled, the astropy bug would result in a signed
# int column dtype:
assert result["col1"].dtype == np.uint32