test_cfunc.py

"""
Tests for @cfunc and friends.
"""


import ctypes
import os
import subprocess
import sys
from collections import namedtuple

import numpy as np

from numba import cfunc, carray, farray, njit
from numba.core import types, typing, utils
import numba.core.typing.cffi_utils as cffi_support
from numba.tests.support import (TestCase, skip_unless_cffi, tag,
                                 captured_stderr)
import unittest
from numba.np import numpy_support


def add_usecase(a, b):
    return a + b

def div_usecase(a, b):
    c = a / b
    return c

def square_usecase(a):
    return a ** 2

add_sig = "float64(float64, float64)"

div_sig = "float64(int64, int64)"

square_sig = "float64(float64)"

def objmode_usecase(a, b):
    object()
    return a + b

# Test functions for carray() and farray()

CARRAY_USECASE_OUT_LEN = 8

def make_cfarray_usecase(func):

    def cfarray_usecase(in_ptr, out_ptr, m, n):
        # Tuple shape
        in_ = func(in_ptr, (m, n))
        # Integer shape
        out = func(out_ptr, CARRAY_USECASE_OUT_LEN)
        out[0] = in_.ndim
        out[1:3] = in_.shape
        out[3:5] = in_.strides
        out[5] = in_.flags.c_contiguous
        out[6] = in_.flags.f_contiguous
        s = 0
        for i, j in np.ndindex(m, n):
            s += in_[i, j] * (i - j)
        out[7] = s

    return cfarray_usecase

carray_usecase = make_cfarray_usecase(carray)
farray_usecase = make_cfarray_usecase(farray)


def make_cfarray_dtype_usecase(func):
    # Same as make_cfarray_usecase(), but with explicit dtype.

    def cfarray_usecase(in_ptr, out_ptr, m, n):
        # Tuple shape
        in_ = func(in_ptr, (m, n), dtype=np.float32)
        # Integer shape
        out = func(out_ptr, CARRAY_USECASE_OUT_LEN, np.float32)
        out[0] = in_.ndim
        out[1:3] = in_.shape
        out[3:5] = in_.strides
        out[5] = in_.flags.c_contiguous
        out[6] = in_.flags.f_contiguous
        s = 0
        for i, j in np.ndindex(m, n):
            s += in_[i, j] * (i - j)
        out[7] = s

    return cfarray_usecase

carray_dtype_usecase = make_cfarray_dtype_usecase(carray)
farray_dtype_usecase = make_cfarray_dtype_usecase(farray)

carray_float32_usecase_sig = types.void(types.CPointer(types.float32),
                                        types.CPointer(types.float32),
                                        types.intp, types.intp)

carray_float64_usecase_sig = types.void(types.CPointer(types.float64),
                                        types.CPointer(types.float64),
                                        types.intp, types.intp)

carray_voidptr_usecase_sig = types.void(types.voidptr, types.voidptr,
                                        types.intp, types.intp)


class TestCFunc(TestCase):

    def test_basic(self):
        """
        Basic usage and properties of a cfunc.
        """
        f = cfunc(add_sig)(add_usecase)

        self.assertEqual(f.__name__, "add_usecase")
        self.assertEqual(f.__qualname__, "add_usecase")
        self.assertIs(f.__wrapped__, add_usecase)

        symbol = f.native_name
        self.assertIsInstance(symbol, str)
        self.assertIn("add_usecase", symbol)

        addr = f.address
        self.assertIsInstance(addr, int)

        ct = f.ctypes
        self.assertEqual(ctypes.cast(ct, ctypes.c_void_p).value, addr)

        self.assertPreciseEqual(ct(2.0, 3.5), 5.5)

    @skip_unless_cffi
    def test_cffi(self):
        from numba.tests import cffi_usecases
        ffi, lib = cffi_usecases.load_inline_module()

        f = cfunc(square_sig)(square_usecase)

        res = lib._numba_test_funcptr(f.cffi)
        self.assertPreciseEqual(res, 2.25)  # 1.5 ** 2

    def test_locals(self):
        # By forcing the intermediate result into an integer, we
        # truncate the ultimate function result
        f = cfunc(div_sig, locals={'c': types.int64})(div_usecase)
        self.assertPreciseEqual(f.ctypes(8, 3), 2.0)

    def test_errors(self):
        f = cfunc(div_sig)(div_usecase)

        with captured_stderr() as err:
            self.assertPreciseEqual(f.ctypes(5, 2), 2.5)
        self.assertEqual(err.getvalue(), "")

        with captured_stderr() as err:
            res = f.ctypes(5, 0)
            # This is just a side effect of Numba zero-initializing
            # stack variables, and could change in the future.
            self.assertPreciseEqual(res, 0.0)
        err = err.getvalue()
        self.assertIn("ZeroDivisionError:", err)
        self.assertIn("Exception ignored", err)

    def test_llvm_ir(self):
        f = cfunc(add_sig)(add_usecase)
        ir = f.inspect_llvm()
        self.assertIn(f.native_name, ir)
        self.assertIn("fadd double", ir)

    def test_object_mode(self):
        """
        Object mode is currently unsupported.
        """
        with self.assertRaises(NotImplementedError):
            cfunc(add_sig, forceobj=True)(add_usecase)
        with self.assertTypingError() as raises:
            cfunc(add_sig)(objmode_usecase)
        self.assertIn("Untyped global name 'object'", str(raises.exception))


class TestCArray(TestCase):
    """
    Tests for carray() and farray().
    """

    def run_carray_usecase(self, pointer_factory, func):
        a = np.arange(10, 16).reshape((2, 3)).astype(np.float32)
        out = np.empty(CARRAY_USECASE_OUT_LEN, dtype=np.float32)
        func(pointer_factory(a), pointer_factory(out), *a.shape)
        return out

    def check_carray_usecase(self, pointer_factory, pyfunc, cfunc):
        expected = self.run_carray_usecase(pointer_factory, pyfunc)
        got = self.run_carray_usecase(pointer_factory, cfunc)
        self.assertPreciseEqual(expected, got)

    def make_voidptr(self, arr):
        return arr.ctypes.data_as(ctypes.c_void_p)

    def make_float32_pointer(self, arr):
        return arr.ctypes.data_as(ctypes.POINTER(ctypes.c_float))

    def make_float64_pointer(self, arr):
        return arr.ctypes.data_as(ctypes.POINTER(ctypes.c_double))

    def check_carray_farray(self, func, order):
        def eq(got, expected):
            # Same layout, dtype, shape, etc.
            self.assertPreciseEqual(got, expected)
            # Same underlying data
            self.assertEqual(got.ctypes.data, expected.ctypes.data)

        base = np.arange(6).reshape((2, 3)).astype(np.float32).copy(order=order)

        # With typed pointer and implied dtype
        a = func(self.make_float32_pointer(base), base.shape)
        eq(a, base)
        # Integer shape
        a = func(self.make_float32_pointer(base), base.size)
        eq(a, base.ravel('K'))

        # With typed pointer and explicit dtype
        a = func(self.make_float32_pointer(base), base.shape, base.dtype)
        eq(a, base)
        a = func(self.make_float32_pointer(base), base.shape, np.float32)
        eq(a, base)

        # With voidptr and explicit dtype
        a = func(self.make_voidptr(base), base.shape, base.dtype)
        eq(a, base)
        a = func(self.make_voidptr(base), base.shape, np.int32)
        eq(a, base.view(np.int32))

        # voidptr without dtype
        with self.assertRaises(TypeError):
            func(self.make_voidptr(base), base.shape)
        # Invalid pointer type
        with self.assertRaises(TypeError):
            func(base.ctypes.data, base.shape)
        # Mismatching dtype
        with self.assertRaises(TypeError) as raises:
            func(self.make_float32_pointer(base), base.shape, np.int32)
        self.assertIn("mismatching dtype 'int32' for pointer",
                      str(raises.exception))

    def test_carray(self):
        """
        Test pure Python carray().
        """
        self.check_carray_farray(carray, 'C')

    def test_farray(self):
        """
        Test pure Python farray().
        """
        self.check_carray_farray(farray, 'F')

    def make_carray_sigs(self, formal_sig):
        """
        Generate a bunch of concrete signatures by varying the width
        and signedness of size arguments (see issue #1923).
        """
        for actual_size in (types.intp, types.int32, types.intc,
                            types.uintp, types.uint32, types.uintc):
            args = tuple(actual_size if a == types.intp else a
                         for a in formal_sig.args)
            yield formal_sig.return_type(*args)

    def check_numba_carray_farray(self, usecase, dtype_usecase):
        # With typed pointers and implicit dtype
        pyfunc = usecase
        for sig in self.make_carray_sigs(carray_float32_usecase_sig):
            f = cfunc(sig)(pyfunc)
            self.check_carray_usecase(self.make_float32_pointer, pyfunc, f.ctypes)

        # With typed pointers and explicit (matching) dtype
        pyfunc = dtype_usecase
        for sig in self.make_carray_sigs(carray_float32_usecase_sig):
            f = cfunc(sig)(pyfunc)
            self.check_carray_usecase(self.make_float32_pointer, pyfunc, f.ctypes)
        # With typed pointers and mismatching dtype
        with self.assertTypingError() as raises:
            f = cfunc(carray_float64_usecase_sig)(pyfunc)
        self.assertIn("mismatching dtype 'float32' for pointer type 'float64*'",
                      str(raises.exception))

        # With voidptr
        pyfunc = dtype_usecase
        for sig in self.make_carray_sigs(carray_voidptr_usecase_sig):
            f = cfunc(sig)(pyfunc)
            self.check_carray_usecase(self.make_float32_pointer, pyfunc, f.ctypes)

    def test_numba_carray(self):
        """
        Test Numba-compiled carray() against pure Python carray()
        """
        self.check_numba_carray_farray(carray_usecase, carray_dtype_usecase)

    def test_numba_farray(self):
        """
        Test Numba-compiled farray() against pure Python farray()
        """
        self.check_numba_carray_farray(farray_usecase, farray_dtype_usecase)


@skip_unless_cffi
class TestCffiStruct(TestCase):
    c_source = """
typedef struct _big_struct {
    int    i1;
    float  f2;
    double d3;
    float  af4[9];
} big_struct;

typedef struct _error {
    int bits:4;
} error;

typedef double (*myfunc)(big_struct*, size_t);
"""

    def get_ffi(self, src=c_source):
        from cffi import FFI

        ffi = FFI()
        ffi.cdef(src)
        return ffi

    def test_type_parsing(self):
        ffi = self.get_ffi()
        # Check struct typedef
        big_struct = ffi.typeof('big_struct')
        nbtype = cffi_support.map_type(big_struct, use_record_dtype=True)
        self.assertIsInstance(nbtype, types.Record)
        self.assertEqual(len(nbtype), 4)
        self.assertEqual(nbtype.typeof('i1'), types.int32)
        self.assertEqual(nbtype.typeof('f2'), types.float32)
        self.assertEqual(nbtype.typeof('d3'), types.float64)
        self.assertEqual(
            nbtype.typeof('af4'),
            types.NestedArray(dtype=types.float32, shape=(9,)),
        )

        # Check function typedef
        myfunc = ffi.typeof('myfunc')
        sig = cffi_support.map_type(myfunc, use_record_dtype=True)
        self.assertIsInstance(sig, typing.Signature)
        self.assertEqual(sig.args[0], types.CPointer(nbtype))
        self.assertEqual(sig.args[1], types.uintp)
        self.assertEqual(sig.return_type, types.float64)

    def test_cfunc_callback(self):
        ffi = self.get_ffi()
        big_struct = ffi.typeof('big_struct')
        nb_big_struct = cffi_support.map_type(big_struct, use_record_dtype=True)
        sig = cffi_support.map_type(ffi.typeof('myfunc'), use_record_dtype=True)

        @njit
        def calc(base):
            tmp = 0
            for i in range(base.size):
                elem = base[i]
                tmp += elem.i1 * elem.f2 / elem.d3
                tmp += base[i].af4.sum()
            return tmp

        @cfunc(sig)
        def foo(ptr, n):
            base = carray(ptr, n)
            return calc(base)

        # Make data
        mydata = ffi.new('big_struct[3]')
        ptr = ffi.cast('big_struct*', mydata)
        for i in range(3):
            ptr[i].i1 = i * 123
            ptr[i].f2 = i * 213
            ptr[i].d3 = (1 + i) * 213
            for j in range(9):
                ptr[i].af4[j] = i * 10 + j

        # Address of my data
        addr = int(ffi.cast('size_t', ptr))
        got = foo.ctypes(addr, 3)

        # Make numpy array from the cffi buffer
        array = np.ndarray(
            buffer=ffi.buffer(mydata),
            dtype=numpy_support.as_dtype(nb_big_struct),
            shape=3,
            )
        expect = calc(array)
        self.assertEqual(got, expect)

    def test_unsupport_bitsize(self):
        ffi = self.get_ffi()
        with self.assertRaises(ValueError) as raises:
            cffi_support.map_type(
                ffi.typeof('error'),
                use_record_dtype=True,
            )
        # When bitsize is provided, bitshift defaults to 0.
        self.assertEqual(
            "field 'bits' has bitshift, this is not supported",
            str(raises.exception)
        )


if __name__ == "__main__":
    unittest.main()