Zengtudor/pybind11
1
0
Fork 0
mirror of https://github.com/pybind/pybind11.git synced 2024-11-11 08:03:55 +00:00
Code Issues Packages Projects Releases Wiki Activity
46dbee7d42
pybind11/tests/test_numpy_vectorize.py

204 lines
8.7 KiB
Python
Raw Normal View History

Port tests to pytest Use simple asserts and pytest's powerful introspection to make testing simpler. This merges the old .py/.ref file pairs into simple .py files where the expected values are right next to the code being tested. This commit does not touch the C++ part of the code and replicates the Python tests exactly like the old .ref-file-based approach.
2016-08-12 11:50:00 +00:00
import pytest
Move requires_numpy, etc. decorators to globals test_eigen.py and test_numpy_*.py have the same @pytest.requires_eigen_and_numpy or @pytest.requires_numpy on every single test; this changes them to use pytest's global `pytestmark = ...` instead to disable the entire module when numpy and/or eigen aren't available.
2017-01-24 16:26:51 +00:00
pytestmark = pytest.requires_numpy
Port tests to pytest Use simple asserts and pytest's powerful introspection to make testing simpler. This merges the old .py/.ref file pairs into simple .py files where the expected values are right next to the code being tested. This commit does not touch the C++ part of the code and replicates the Python tests exactly like the old .ref-file-based approach.
2016-08-12 11:50:00 +00:00
with pytest.suppress(ImportError):
import numpy as np
def test_vectorize(capture):
from pybind11_tests import vectorized_func, vectorized_func2, vectorized_func3
assert np.isclose(vectorized_func3(np.array(3 + 7j)), [6 + 14j])
for f in [vectorized_func, vectorized_func2]:
with capture:
assert np.isclose(f(1, 2, 3), 6)
assert capture == "my_func(x:int=1, y:float=2, z:float=3)"
with capture:
assert np.isclose(f(np.array(1), np.array(2), 3), 6)
assert capture == "my_func(x:int=1, y:float=2, z:float=3)"
with capture:
assert np.allclose(f(np.array([1, 3]), np.array([2, 4]), 3), [6, 36])
assert capture == """
my_func(x:int=1, y:float=2, z:float=3)
my_func(x:int=3, y:float=4, z:float=3)
"""
vectorize: trivial handling for F-order arrays This extends the trivial handling to support trivial handling for Fortran-order arrays (i.e. column major): if inputs aren't all C-contiguous, but *are* all F-contiguous, the resulting array will be F-contiguous and we can do trivial processing. For anything else (e.g. C-contiguous, or inputs requiring non-trivial processing), the result is in (numpy-default) C-contiguous layout.
2017-03-19 00:11:59 +00:00
with capture:
a = np.array([[1, 2], [3, 4]], order='F')
b = np.array([[10, 20], [30, 40]], order='F')
c = 3
result = f(a, b, c)
assert np.allclose(result, a * b * c)
assert result.flags.f_contiguous
# All inputs are F order and full or singletons, so we the result is in col-major order:
assert capture == """
my_func(x:int=1, y:float=10, z:float=3)
my_func(x:int=3, y:float=30, z:float=3)
my_func(x:int=2, y:float=20, z:float=3)
my_func(x:int=4, y:float=40, z:float=3)
"""
Port tests to pytest Use simple asserts and pytest's powerful introspection to make testing simpler. This merges the old .py/.ref file pairs into simple .py files where the expected values are right next to the code being tested. This commit does not touch the C++ part of the code and replicates the Python tests exactly like the old .ref-file-based approach.
2016-08-12 11:50:00 +00:00
with capture:
a, b, c = np.array([[1, 3, 5], [7, 9, 11]]), np.array([[2, 4, 6], [8, 10, 12]]), 3
assert np.allclose(f(a, b, c), a * b * c)
assert capture == """
my_func(x:int=1, y:float=2, z:float=3)
my_func(x:int=3, y:float=4, z:float=3)
my_func(x:int=5, y:float=6, z:float=3)
my_func(x:int=7, y:float=8, z:float=3)
my_func(x:int=9, y:float=10, z:float=3)
my_func(x:int=11, y:float=12, z:float=3)
"""
with capture:
a, b, c = np.array([[1, 2, 3], [4, 5, 6]]), np.array([2, 3, 4]), 2
assert np.allclose(f(a, b, c), a * b * c)
assert capture == """
my_func(x:int=1, y:float=2, z:float=2)
my_func(x:int=2, y:float=3, z:float=2)
my_func(x:int=3, y:float=4, z:float=2)
my_func(x:int=4, y:float=2, z:float=2)
my_func(x:int=5, y:float=3, z:float=2)
my_func(x:int=6, y:float=4, z:float=2)
"""
with capture:
a, b, c = np.array([[1, 2, 3], [4, 5, 6]]), np.array([[2], [3]]), 2
assert np.allclose(f(a, b, c), a * b * c)
assert capture == """
my_func(x:int=1, y:float=2, z:float=2)
my_func(x:int=2, y:float=2, z:float=2)
my_func(x:int=3, y:float=2, z:float=2)
my_func(x:int=4, y:float=3, z:float=2)
my_func(x:int=5, y:float=3, z:float=2)
my_func(x:int=6, y:float=3, z:float=2)
"""
Stop forcing c-contiguous in py::vectorize The only part of the vectorize code that actually needs c-contiguous is the "trivial" broadcast; for non-trivial arguments, the code already uses strides properly (and so handles C-style, F-style, neither, slices, etc.) This commit rewrites `broadcast` to additionally check for C-contiguous storage, then takes off the `c_style` flag for the arguments, which will keep the functionality more or less the same, except for no longer requiring an array copy for non-c-contiguous input arrays. Additionally, if we're given a singleton slice (e.g. a[0::4, 0::4] for a 4x4 or smaller array), we no longer fail triviality because the trivial code path never actually uses the strides on a singleton.
2017-03-15 03:57:56 +00:00
with capture:
a, b, c = np.array([[1, 2, 3], [4, 5, 6]], order='F'), np.array([[2], [3]]), 2
assert np.allclose(f(a, b, c), a * b * c)
assert capture == """
my_func(x:int=1, y:float=2, z:float=2)
my_func(x:int=2, y:float=2, z:float=2)
my_func(x:int=3, y:float=2, z:float=2)
my_func(x:int=4, y:float=3, z:float=2)
my_func(x:int=5, y:float=3, z:float=2)
my_func(x:int=6, y:float=3, z:float=2)
"""
with capture:
a, b, c = np.array([[1, 2, 3], [4, 5, 6]])[::, ::2], np.array([[2], [3]]), 2
assert np.allclose(f(a, b, c), a * b * c)
assert capture == """
my_func(x:int=1, y:float=2, z:float=2)
my_func(x:int=3, y:float=2, z:float=2)
my_func(x:int=4, y:float=3, z:float=2)
my_func(x:int=6, y:float=3, z:float=2)
"""
with capture:
a, b, c = np.array([[1, 2, 3], [4, 5, 6]], order='F')[::, ::2], np.array([[2], [3]]), 2
assert np.allclose(f(a, b, c), a * b * c)
assert capture == """
my_func(x:int=1, y:float=2, z:float=2)
my_func(x:int=3, y:float=2, z:float=2)
my_func(x:int=4, y:float=3, z:float=2)
my_func(x:int=6, y:float=3, z:float=2)
"""
Port tests to pytest Use simple asserts and pytest's powerful introspection to make testing simpler. This merges the old .py/.ref file pairs into simple .py files where the expected values are right next to the code being tested. This commit does not touch the C++ part of the code and replicates the Python tests exactly like the old .ref-file-based approach.
2016-08-12 11:50:00 +00:00
Simplify tests by replacing output capture with asserts where possible The C++ part of the test code is modified to achieve this. As a result, this kind of test: ```python with capture: kw_func1(5, y=10) assert capture == "kw_func(x=5, y=10)" ``` can be replaced with a simple: `assert kw_func1(5, y=10) == "x=5, y=10"`
2016-08-12 20:28:31 +00:00
def test_type_selection():
Port tests to pytest Use simple asserts and pytest's powerful introspection to make testing simpler. This merges the old .py/.ref file pairs into simple .py files where the expected values are right next to the code being tested. This commit does not touch the C++ part of the code and replicates the Python tests exactly like the old .ref-file-based approach.
2016-08-12 11:50:00 +00:00
from pybind11_tests import selective_func
Simplify tests by replacing output capture with asserts where possible The C++ part of the test code is modified to achieve this. As a result, this kind of test: ```python with capture: kw_func1(5, y=10) assert capture == "kw_func(x=5, y=10)" ``` can be replaced with a simple: `assert kw_func1(5, y=10) == "x=5, y=10"`
2016-08-12 20:28:31 +00:00
assert selective_func(np.array([1], dtype=np.int32)) == "Int branch taken."
assert selective_func(np.array([1.0], dtype=np.float32)) == "Float branch taken."
assert selective_func(np.array([1.0j], dtype=np.complex64)) == "Complex float branch taken."
Port tests to pytest Use simple asserts and pytest's powerful introspection to make testing simpler. This merges the old .py/.ref file pairs into simple .py files where the expected values are right next to the code being tested. This commit does not touch the C++ part of the code and replicates the Python tests exactly like the old .ref-file-based approach.
2016-08-12 11:50:00 +00:00
def test_docs(doc):
from pybind11_tests import vectorized_func
Set maximum line length for Python style checker (#552)
2016-12-12 23:59:28 +00:00
assert doc(vectorized_func) == """
Improve py::array_t scalar type information (#724) * Add value_type member alias to py::array_t (resolve #632) * Use numpy scalar name in py::array_t function signatures (e.g. float32/64 instead of just float)
2017-03-13 18:17:18 +00:00
vectorized_func(arg0: numpy.ndarray[int32], arg1: numpy.ndarray[float32], arg2: numpy.ndarray[float64]) -> object
Set maximum line length for Python style checker (#552)
2016-12-12 23:59:28 +00:00
""" # noqa: E501 line too long
Stop forcing c-contiguous in py::vectorize The only part of the vectorize code that actually needs c-contiguous is the "trivial" broadcast; for non-trivial arguments, the code already uses strides properly (and so handles C-style, F-style, neither, slices, etc.) This commit rewrites `broadcast` to additionally check for C-contiguous storage, then takes off the `c_style` flag for the arguments, which will keep the functionality more or less the same, except for no longer requiring an array copy for non-c-contiguous input arrays. Additionally, if we're given a singleton slice (e.g. a[0::4, 0::4] for a 4x4 or smaller array), we no longer fail triviality because the trivial code path never actually uses the strides on a singleton.
2017-03-15 03:57:56 +00:00
def test_trivial_broadcasting():
vectorize: trivial handling for F-order arrays This extends the trivial handling to support trivial handling for Fortran-order arrays (i.e. column major): if inputs aren't all C-contiguous, but *are* all F-contiguous, the resulting array will be F-contiguous and we can do trivial processing. For anything else (e.g. C-contiguous, or inputs requiring non-trivial processing), the result is in (numpy-default) C-contiguous layout.
2017-03-19 00:11:59 +00:00
from pybind11_tests import vectorized_is_trivial, trivial, vectorized_func
Stop forcing c-contiguous in py::vectorize The only part of the vectorize code that actually needs c-contiguous is the "trivial" broadcast; for non-trivial arguments, the code already uses strides properly (and so handles C-style, F-style, neither, slices, etc.) This commit rewrites `broadcast` to additionally check for C-contiguous storage, then takes off the `c_style` flag for the arguments, which will keep the functionality more or less the same, except for no longer requiring an array copy for non-c-contiguous input arrays. Additionally, if we're given a singleton slice (e.g. a[0::4, 0::4] for a 4x4 or smaller array), we no longer fail triviality because the trivial code path never actually uses the strides on a singleton.
2017-03-15 03:57:56 +00:00
vectorize: trivial handling for F-order arrays This extends the trivial handling to support trivial handling for Fortran-order arrays (i.e. column major): if inputs aren't all C-contiguous, but *are* all F-contiguous, the resulting array will be F-contiguous and we can do trivial processing. For anything else (e.g. C-contiguous, or inputs requiring non-trivial processing), the result is in (numpy-default) C-contiguous layout.
2017-03-19 00:11:59 +00:00
assert vectorized_is_trivial(1, 2, 3) == trivial.c_trivial
assert vectorized_is_trivial(np.array(1), np.array(2), 3) == trivial.c_trivial
assert vectorized_is_trivial(np.array([1, 3]), np.array([2, 4]), 3) == trivial.c_trivial
assert trivial.c_trivial == vectorized_is_trivial(
Stop forcing c-contiguous in py::vectorize The only part of the vectorize code that actually needs c-contiguous is the "trivial" broadcast; for non-trivial arguments, the code already uses strides properly (and so handles C-style, F-style, neither, slices, etc.) This commit rewrites `broadcast` to additionally check for C-contiguous storage, then takes off the `c_style` flag for the arguments, which will keep the functionality more or less the same, except for no longer requiring an array copy for non-c-contiguous input arrays. Additionally, if we're given a singleton slice (e.g. a[0::4, 0::4] for a 4x4 or smaller array), we no longer fail triviality because the trivial code path never actually uses the strides on a singleton.
2017-03-15 03:57:56 +00:00
np.array([[1, 3, 5], [7, 9, 11]]), np.array([[2, 4, 6], [8, 10, 12]]), 3)
vectorize: trivial handling for F-order arrays This extends the trivial handling to support trivial handling for Fortran-order arrays (i.e. column major): if inputs aren't all C-contiguous, but *are* all F-contiguous, the resulting array will be F-contiguous and we can do trivial processing. For anything else (e.g. C-contiguous, or inputs requiring non-trivial processing), the result is in (numpy-default) C-contiguous layout.
2017-03-19 00:11:59 +00:00
assert vectorized_is_trivial(
np.array([[1, 2, 3], [4, 5, 6]]), np.array([2, 3, 4]), 2) == trivial.non_trivial
assert vectorized_is_trivial(
np.array([[1, 2, 3], [4, 5, 6]]), np.array([[2], [3]]), 2) == trivial.non_trivial
Stop forcing c-contiguous in py::vectorize The only part of the vectorize code that actually needs c-contiguous is the "trivial" broadcast; for non-trivial arguments, the code already uses strides properly (and so handles C-style, F-style, neither, slices, etc.) This commit rewrites `broadcast` to additionally check for C-contiguous storage, then takes off the `c_style` flag for the arguments, which will keep the functionality more or less the same, except for no longer requiring an array copy for non-c-contiguous input arrays. Additionally, if we're given a singleton slice (e.g. a[0::4, 0::4] for a 4x4 or smaller array), we no longer fail triviality because the trivial code path never actually uses the strides on a singleton.
2017-03-15 03:57:56 +00:00
z1 = np.array([[1, 2, 3, 4], [5, 6, 7, 8]], dtype='int32')
z2 = np.array(z1, dtype='float32')
z3 = np.array(z1, dtype='float64')
vectorize: trivial handling for F-order arrays This extends the trivial handling to support trivial handling for Fortran-order arrays (i.e. column major): if inputs aren't all C-contiguous, but *are* all F-contiguous, the resulting array will be F-contiguous and we can do trivial processing. For anything else (e.g. C-contiguous, or inputs requiring non-trivial processing), the result is in (numpy-default) C-contiguous layout.
2017-03-19 00:11:59 +00:00
assert vectorized_is_trivial(z1, z2, z3) == trivial.c_trivial
assert vectorized_is_trivial(1, z2, z3) == trivial.c_trivial
assert vectorized_is_trivial(z1, 1, z3) == trivial.c_trivial
assert vectorized_is_trivial(z1, z2, 1) == trivial.c_trivial
assert vectorized_is_trivial(z1[::2, ::2], 1, 1) == trivial.non_trivial
assert vectorized_is_trivial(1, 1, z1[::2, ::2]) == trivial.c_trivial
assert vectorized_is_trivial(1, 1, z3[::2, ::2]) == trivial.non_trivial
assert vectorized_is_trivial(z1, 1, z3[1::4, 1::4]) == trivial.c_trivial
Stop forcing c-contiguous in py::vectorize The only part of the vectorize code that actually needs c-contiguous is the "trivial" broadcast; for non-trivial arguments, the code already uses strides properly (and so handles C-style, F-style, neither, slices, etc.) This commit rewrites `broadcast` to additionally check for C-contiguous storage, then takes off the `c_style` flag for the arguments, which will keep the functionality more or less the same, except for no longer requiring an array copy for non-c-contiguous input arrays. Additionally, if we're given a singleton slice (e.g. a[0::4, 0::4] for a 4x4 or smaller array), we no longer fail triviality because the trivial code path never actually uses the strides on a singleton.
2017-03-15 03:57:56 +00:00
y1 = np.array(z1, order='F')
y2 = np.array(y1)
y3 = np.array(y1)
vectorize: trivial handling for F-order arrays This extends the trivial handling to support trivial handling for Fortran-order arrays (i.e. column major): if inputs aren't all C-contiguous, but *are* all F-contiguous, the resulting array will be F-contiguous and we can do trivial processing. For anything else (e.g. C-contiguous, or inputs requiring non-trivial processing), the result is in (numpy-default) C-contiguous layout.
2017-03-19 00:11:59 +00:00
assert vectorized_is_trivial(y1, y2, y3) == trivial.f_trivial
assert vectorized_is_trivial(y1, 1, 1) == trivial.f_trivial
assert vectorized_is_trivial(1, y2, 1) == trivial.f_trivial
assert vectorized_is_trivial(1, 1, y3) == trivial.f_trivial
assert vectorized_is_trivial(y1, z2, 1) == trivial.non_trivial
assert vectorized_is_trivial(z1[1::4, 1::4], y2, 1) == trivial.f_trivial
assert vectorized_is_trivial(y1[1::4, 1::4], z2, 1) == trivial.c_trivial
assert vectorized_func(z1, z2, z3).flags.c_contiguous
assert vectorized_func(y1, y2, y3).flags.f_contiguous
assert vectorized_func(z1, 1, 1).flags.c_contiguous
assert vectorized_func(1, y2, 1).flags.f_contiguous
assert vectorized_func(z1[1::4, 1::4], y2, 1).flags.f_contiguous
assert vectorized_func(y1[1::4, 1::4], z2, 1).flags.c_contiguous
vectorize: pass-through of non-vectorizable args This extends py::vectorize to automatically pass through non-vectorizable arguments. This removes the need for the documented "explicitly exclude an argument" workaround. Vectorization now applies to arithmetic, std::complex, and POD types, passed as plain value or by const lvalue reference (previously only pass-by-value types were supported). Non-const lvalue references and any other types are passed through as-is. Functions with rvalue reference arguments (whether vectorizable or not) are explicitly prohibited: an rvalue reference is inherently not something that can be passed multiple times and is thus unsuitable to being in a vectorized function. The vectorize returned value is also now more sensitive to inputs: previously it would return by value when all inputs are of size 1; this is now amended to having all inputs of size 1 *and* 0 dimensions. Thus if you pass in, for example, [[1]], you get back a 1x1, 2D array, while previously you got back just the resulting single value. Vectorization of member function specializations is now also supported via `py::vectorize(&Class::method)`; this required passthrough support for the initial object pointer on the wrapping function pointer.
2017-03-26 03:51:40 +00:00
def test_passthrough_arguments(doc):
from pybind11_tests import vec_passthrough, NonPODClass
assert doc(vec_passthrough) == (
"vec_passthrough("
"arg0: float, arg1: numpy.ndarray[float64], arg2: numpy.ndarray[float64], "
"arg3: numpy.ndarray[int32], arg4: int, arg5: m.NonPODClass, arg6: numpy.ndarray[float64]"
") -> object")
b = np.array([[10, 20, 30]], dtype='float64')
c = np.array([100, 200]) # NOT a vectorized argument
d = np.array([[1000], [2000], [3000]], dtype='int')
g = np.array([[1000000, 2000000, 3000000]], dtype='int') # requires casting
assert np.all(
vec_passthrough(1, b, c, d, 10000, NonPODClass(100000), g) ==
np.array([[1111111, 2111121, 3111131],
[1112111, 2112121, 3112131],
[1113111, 2113121, 3113131]]))
def test_method_vectorization():
from pybind11_tests import VectorizeTestClass
o = VectorizeTestClass(3)
x = np.array([1, 2], dtype='int')
y = np.array([[10], [20]], dtype='float32')
assert np.all(o.method(x, y) == [[14, 15], [24, 25]])
def test_array_collapse():
from pybind11_tests import vectorized_func
assert not isinstance(vectorized_func(1, 2, 3), np.ndarray)
assert not isinstance(vectorized_func(np.array(1), 2, 3), np.ndarray)
z = vectorized_func([1], 2, 3)
assert isinstance(z, np.ndarray)
assert z.shape == (1, )
z = vectorized_func(1, [[[2]]], 3)
assert isinstance(z, np.ndarray)
assert z.shape == (1, 1, 1)
Reference in New Issue Copy Permalink