Author: Armin Rigo <ar...@tunes.org>
Branch:
Changeset: r741:2ac4cc98111f
Date: 2012-07-30 15:05 +0200
http://bitbucket.org/cffi/cffi/changeset/2ac4cc98111f/
Log: Found out how to properly generalize the "pass a Python string as a
'char *' argument to a function call". It has the nice effect that
the documented example can be simplified.
diff --git a/c/_cffi_backend.c b/c/_cffi_backend.c
--- a/c/_cffi_backend.c
+++ b/c/_cffi_backend.c
@@ -64,7 +64,8 @@
Py_ssize_t ct_size; /* size of instances, or -1 if unknown */
Py_ssize_t ct_length; /* length of arrays, or -1 if unknown;
- or alignment of primitive and struct types */
+ or alignment of primitive and struct types;
+ always -1 for pointers */
int ct_flags; /* CT_xxx flags */
int ct_name_position; /* index in ct_name of where to put a var name */
@@ -735,78 +736,91 @@
}
static int
+convert_array_from_object(char *data, CTypeDescrObject *ct, PyObject *init)
+{
+ /* used by convert_from_object(), and also to decode lists/tuples/unicodes
+ passed as function arguments. 'ct' is an CT_ARRAY in the first case
+ and a CT_POINTER in the second case. */
+ const char *expected;
+ CTypeDescrObject *ctitem = ct->ct_itemdescr;
+
+ if (PyList_Check(init) || PyTuple_Check(init)) {
+ PyObject **items;
+ Py_ssize_t i, n;
+ n = PySequence_Fast_GET_SIZE(init);
+ if (ct->ct_length >= 0 && n > ct->ct_length) {
+ PyErr_Format(PyExc_IndexError,
+ "too many initializers for '%s' (got %zd)",
+ ct->ct_name, n);
+ return -1;
+ }
+ items = PySequence_Fast_ITEMS(init);
+ for (i=0; i<n; i++) {
+ if (convert_from_object(data, ctitem, items[i]) < 0)
+ return -1;
+ data += ctitem->ct_size;
+ }
+ return 0;
+ }
+ else if (ctitem->ct_flags & CT_PRIMITIVE_CHAR) {
+ if (ctitem->ct_size == sizeof(char)) {
+ char *srcdata;
+ Py_ssize_t n;
+ if (!PyString_Check(init)) {
+ expected = "str or list or tuple";
+ goto cannot_convert;
+ }
+ n = PyString_GET_SIZE(init);
+ if (ct->ct_length >= 0 && n > ct->ct_length) {
+ PyErr_Format(PyExc_IndexError,
+ "initializer string is too long for '%s' "
+ "(got %zd characters)", ct->ct_name, n);
+ return -1;
+ }
+ if (n != ct->ct_length)
+ n++;
+ srcdata = PyString_AS_STRING(init);
+ memcpy(data, srcdata, n);
+ return 0;
+ }
+#ifdef HAVE_WCHAR_H
+ else {
+ Py_ssize_t n;
+ if (!PyUnicode_Check(init)) {
+ expected = "unicode or list or tuple";
+ goto cannot_convert;
+ }
+ n = _my_PyUnicode_SizeAsWideChar(init);
+ if (ct->ct_length >= 0 && n > ct->ct_length) {
+ PyErr_Format(PyExc_IndexError,
+ "initializer unicode is too long for '%s' "
+ "(got %zd characters)", ct->ct_name, n);
+ return -1;
+ }
+ if (n != ct->ct_length)
+ n++;
+ _my_PyUnicode_AsWideChar(init, (wchar_t *)data, n);
+ return 0;
+ }
+#endif
+ }
+ else {
+ expected = "list or tuple";
+ goto cannot_convert;
+ }
+
+ cannot_convert:
+ return _convert_error(init, ct->ct_name, expected);
+}
+
+static int
convert_from_object(char *data, CTypeDescrObject *ct, PyObject *init)
{
const char *expected;
char buf[sizeof(PY_LONG_LONG)];
if (ct->ct_flags & CT_ARRAY) {
- CTypeDescrObject *ctitem = ct->ct_itemdescr;
-
- if (PyList_Check(init) || PyTuple_Check(init)) {
- PyObject **items;
- Py_ssize_t i, n;
- n = PySequence_Fast_GET_SIZE(init);
- if (ct->ct_length >= 0 && n > ct->ct_length) {
- PyErr_Format(PyExc_IndexError,
- "too many initializers for '%s' (got %zd)",
- ct->ct_name, n);
- return -1;
- }
- items = PySequence_Fast_ITEMS(init);
- for (i=0; i<n; i++) {
- if (convert_from_object(data, ctitem, items[i]) < 0)
- return -1;
- data += ctitem->ct_size;
- }
- return 0;
- }
- else if (ctitem->ct_flags & CT_PRIMITIVE_CHAR) {
- if (ctitem->ct_size == sizeof(char)) {
- char *srcdata;
- Py_ssize_t n;
- if (!PyString_Check(init)) {
- expected = "str or list or tuple";
- goto cannot_convert;
- }
- n = PyString_GET_SIZE(init);
- if (ct->ct_length >= 0 && n > ct->ct_length) {
- PyErr_Format(PyExc_IndexError,
- "initializer string is too long for '%s' "
- "(got %zd characters)", ct->ct_name, n);
- return -1;
- }
- if (n != ct->ct_length)
- n++;
- srcdata = PyString_AS_STRING(init);
- memcpy(data, srcdata, n);
- return 0;
- }
-#ifdef HAVE_WCHAR_H
- else {
- Py_ssize_t n;
- if (!PyUnicode_Check(init)) {
- expected = "unicode or list or tuple";
- goto cannot_convert;
- }
- n = _my_PyUnicode_SizeAsWideChar(init);
- if (ct->ct_length >= 0 && n > ct->ct_length) {
- PyErr_Format(PyExc_IndexError,
- "initializer unicode is too long for '%s' "
- "(got %zd characters)", ct->ct_name, n);
- return -1;
- }
- if (n != ct->ct_length)
- n++;
- _my_PyUnicode_AsWideChar(init, (wchar_t *)data, n);
- return 0;
- }
-#endif
- }
- else {
- expected = "list or tuple";
- goto cannot_convert;
- }
+ return convert_array_from_object(data, ct, init);
}
if (ct->ct_flags & (CT_POINTER|CT_FUNCTIONPTR)) {
char *ptrdata;
@@ -1599,14 +1613,72 @@
return ct_int;
}
+static PyObject *
+_prepare_pointer_call_argument(CTypeDescrObject *ctptr, PyObject *init)
+{
+ /* 'ctptr' is here a pointer type 'ITEM *'. Accept as argument an
+ initializer for an array 'ITEM[]'. This includes the case of
+ passing a Python string to a 'char *' argument. */
+ Py_ssize_t length, datasize;
+ CTypeDescrObject *ctitem = ctptr->ct_itemdescr;
+ PyObject *result;
+ char *data;
+
+ /* XXX some code duplication, how to avoid it? */
+ if (PyString_Check(init)) {
+ /* from a string: just returning the string here is fine.
+ We assume that the C code won't modify the 'char *' data. */
+ if ((ctitem->ct_flags & CT_PRIMITIVE_CHAR) &&
+ (ctitem->ct_size == sizeof(char))) {
+ Py_INCREF(init);
+ return init;
+ }
+ else
+ return Py_None;
+ }
+ else if (PyList_Check(init) || PyTuple_Check(init)) {
+ length = PySequence_Fast_GET_SIZE(init);
+ }
+ else if (PyUnicode_Check(init)) {
+ /* from a unicode, we add the null terminator */
+ length = _my_PyUnicode_SizeAsWideChar(init) + 1;
+ }
+ else {
+ /* refuse to receive just an integer (and interpret it
+ as the array size) */
+ return Py_None;
+ }
+
+ if (ctitem->ct_size <= 0)
+ return Py_None;
+ datasize = length * ctitem->ct_size;
+ if ((datasize / ctitem->ct_size) != length) {
+ PyErr_SetString(PyExc_OverflowError,
+ "array size would overflow a Py_ssize_t");
+ return NULL;
+ }
+
+ result = PyString_FromStringAndSize(NULL, datasize);
+ if (result == NULL)
+ return NULL;
+
+ data = PyString_AS_STRING(result);
+ memset(data, 0, datasize);
+ if (convert_array_from_object(data, ctptr, init) < 0) {
+ Py_DECREF(result);
+ return NULL;
+ }
+ return result;
+}
+
static PyObject*
cdata_call(CDataObject *cd, PyObject *args, PyObject *kwds)
{
char *buffer;
void** buffer_array;
cif_description_t *cif_descr;
- Py_ssize_t i, nargs, nargs_declared;
- PyObject *signature, *res, *fvarargs;
+ Py_ssize_t i, nargs, nargs_declared, free_me_until = 0;
+ PyObject *signature, *res = NULL, *fvarargs;
CTypeDescrObject *fresult;
char *resultdata;
char *errormsg;
@@ -1636,7 +1708,10 @@
/* regular case: this function does not take '...' arguments */
if (nargs != nargs_declared) {
errormsg = "'%s' expects %zd arguments, got %zd";
- goto bad_number_of_arguments;
+ bad_number_of_arguments:
+ PyErr_Format(PyExc_TypeError, errormsg,
+ cd->c_type->ct_name, nargs_declared, nargs);
+ goto error;
}
}
else {
@@ -1708,26 +1783,21 @@
else
argtype = (CTypeDescrObject *)PyTuple_GET_ITEM(fvarargs, i);
- if ((argtype->ct_flags & CT_POINTER) &&
- (argtype->ct_itemdescr->ct_flags & CT_PRIMITIVE_CHAR)) {
- if (argtype->ct_itemdescr->ct_size == sizeof(char)) {
- if (PyString_Check(obj)) {
- /* special case: Python string -> cdata 'char *' */
- *(char **)data = PyString_AS_STRING(obj);
+ if (argtype->ct_flags & CT_POINTER) {
+ PyObject *string;
+ if (!CData_Check(obj)) {
+ string = _prepare_pointer_call_argument(argtype, obj);
+ if (string != Py_None) {
+ if (string == NULL)
+ goto error;
+ ((char **)data)[0] = PyString_AS_STRING(string);
+ ((char **)data)[1] = (char *)string;
+ assert(i < nargs_declared); /* otherwise, obj is a CData */
+ free_me_until = i + 1;
continue;
}
}
-#ifdef HAVE_WCHAR_H
- else {
- if (PyUnicode_Check(obj)) {
- /* Python Unicode string -> cdata 'wchar_t *':
- not supported yet */
- PyErr_SetString(PyExc_NotImplementedError,
- "automatic unicode-to-'wchar_t *' conversion");
- goto error;
- }
- }
-#endif
+ ((char **)data)[1] = NULL;
}
if (convert_from_object(data, argtype, obj) < 0)
goto error;
@@ -1761,23 +1831,26 @@
else {
res = convert_to_object(resultdata, fresult);
}
- PyObject_Free(buffer);
- done:
+ /* fall-through */
+
+ error:
+ for (i=0; i<free_me_until; i++) {
+ CTypeDescrObject *argtype;
+ argtype = (CTypeDescrObject *)PyTuple_GET_ITEM(signature, 2 + i);
+ if (argtype->ct_flags & CT_POINTER) {
+ char *data = buffer + cif_descr->exchange_offset_arg[1 + i];
+ PyObject *string_or_null = (PyObject *)(((char **)data)[1]);
+ Py_XDECREF(string_or_null);
+ }
+ }
+ if (buffer)
+ PyObject_Free(buffer);
if (fvarargs != NULL) {
Py_DECREF(fvarargs);
if (cif_descr != NULL) /* but only if fvarargs != NULL, if variadic */
PyObject_Free(cif_descr);
}
return res;
-
- bad_number_of_arguments:
- PyErr_Format(PyExc_TypeError, errormsg,
- cd->c_type->ct_name, nargs_declared, nargs);
- error:
- if (buffer)
- PyObject_Free(buffer);
- res = NULL;
- goto done;
}
static PyObject *cdata_iter(CDataObject *);
@@ -2619,6 +2692,7 @@
return NULL;
td->ct_size = sizeof(void *);
+ td->ct_length = -1;
td->ct_flags = CT_POINTER;
if (ctitem->ct_flags & (CT_STRUCT|CT_UNION))
td->ct_flags |= CT_IS_PTR_TO_OWNED;
@@ -3131,6 +3205,15 @@
exchange_offset = ALIGN_ARG(exchange_offset);
cif_descr->exchange_offset_arg[1 + i] = exchange_offset;
exchange_offset += atype->size;
+ /* if 'farg' is a pointer type 'ITEM *', then we might receive
+ as argument to the function call what is an initializer
+ for an array 'ITEM[]'. This includes the case of passing a
+ Python string to a 'char *' argument. In this case, we
+ convert the initializer to a cdata 'ITEM[]' that gets
+ temporarily stored here: */
+ if (farg->ct_flags & CT_POINTER) {
+ exchange_offset += sizeof(PyObject *);
+ }
}
}
@@ -3898,6 +3981,11 @@
return result;
}
+static int _testfunc18(struct _testfunc17_s *ptr)
+{
+ return ptr->a1 + (int)ptr->a2;
+}
+
static PyObject *b__testfunc(PyObject *self, PyObject *args)
{
/* for testing only */
@@ -3924,6 +4012,7 @@
case 15: f = &_testfunc15; break;
case 16: f = &_testfunc16; break;
case 17: f = &_testfunc17; break;
+ case 18: f = &_testfunc18; break;
default:
PyErr_SetNone(PyExc_ValueError);
return NULL;
diff --git a/c/test_c.py b/c/test_c.py
--- a/c/test_c.py
+++ b/c/test_c.py
@@ -763,12 +763,22 @@
BFunc6bis = new_function_type((BIntArray,), BIntPtr, False)
f = cast(BFunc6bis, _testfunc(6))
#
- py.test.raises(TypeError, f, [142])
+ res = f([142])
+ assert typeof(res) is BIntPtr
+ assert res[0] == 142 - 1000
+ #
+ res = f((143,))
+ assert typeof(res) is BIntPtr
+ assert res[0] == 143 - 1000
#
x = newp(BIntArray, [242])
res = f(x)
assert typeof(res) is BIntPtr
assert res[0] == 242 - 1000
+ #
+ py.test.raises(TypeError, f, 123456)
+ py.test.raises(TypeError, f, "foo")
+ py.test.raises(TypeError, f, u"bar")
def test_call_function_7():
BChar = new_primitive_type("char")
@@ -1332,6 +1342,14 @@
assert repr(s) == "<cdata 'struct test17' owning 8 bytes>"
assert s.a1 == 40
assert s.a2 == 40.0 * 40.0
+ #
+ BStruct17Ptr = new_pointer_type(BStruct17)
+ BFunc18 = new_function_type((BStruct17Ptr,), BInt)
+ f = cast(BFunc18, _testfunc(18))
+ x = f([[40, 2.5]])
+ assert x == 42
+ x = f([{'a2': 43.1}])
+ assert x == 43
def test_cast_with_functionptr():
BFunc = new_function_type((), new_void_type())
@@ -1458,8 +1476,7 @@
return len(unicode(p))
BFunc = new_function_type((BWCharP,), BInt, False)
f = callback(BFunc, cb, -42)
- #assert f(u'a\u1234b') == 3 -- not implemented
- py.test.raises(NotImplementedError, f, u'a\u1234b')
+ assert f(u'a\u1234b') == 3
#
if wchar4 and not pyuni4:
# try out-of-range wchar_t values
diff --git a/doc/source/index.rst b/doc/source/index.rst
--- a/doc/source/index.rst
+++ b/doc/source/index.rst
@@ -658,6 +658,7 @@
ffi.cdef("""
int main_like(int argv, char *argv[]);
""")
+ lib = ffi.dlopen("some_library.so")
Now, everything is simple, except, how do we create the ``char**`` argument
here?
@@ -665,20 +666,34 @@
.. code-block:: python
- argv = ffi.new("char *[]", ["arg0", "arg1"])
+ lib.main_like(2, ["arg0", "arg1"])
-Does not work, because the initializer receives python ``str`` instead of
-``char*``. Now, the following would almost work:
+does not work, because the initializer receives two Python ``str`` objects
+where it was expecting ``<cdata 'char *'>`` objects. You need to use
+``ffi.new()`` explicitly to make these objects:
.. code-block:: python
+ lib.main_like(2, [ffi.new("char[]", "arg0"),
+ ffi.new("char[]", "arg1")])
+
+Note that the two ``<cdata 'char[]'>`` objects are kept alive for the
+duration of the call: they are only freed when the list itself is freed,
+and the list is only freed when the call returns.
+
+If you want instead to build an "argv" variable that you want to reuse,
+then more care is needed:
+
+.. code-block:: python
+
+ # DOES NOT WORK!
argv = ffi.new("char *[]", [ffi.new("char[]", "arg0"),
ffi.new("char[]", "arg1")])
-However, the two ``char[]`` objects will not be automatically kept alive.
-To keep them alive, one solution is to make sure that the list is stored
-somewhere for long enough.
-For example:
+In the above example, the inner "arg0" string is deallocated as soon
+as "argv" is built. You have to make sure that you keep a reference
+to the inner "char[]" objects, either directly or by keeping the list
+alive like this:
.. code-block:: python
@@ -686,7 +701,6 @@
ffi.new("char[]", "arg1")]
argv = ffi.new("char *[]", argv_keepalive)
-will work.
Function calls
--------------
@@ -895,23 +909,21 @@
| | any pointer or array | | |
| | type | | |
+---------------+------------------------+ +----------------+
-| ``char *`` | another <cdata> with | | ``[]``, |
-| | any pointer or array | | ``+``, ``-``, |
-| | type, or | | str() |
-| | a Python string when | | |
-| | passed as func argument| | |
+| ``char *`` | same as pointers (*) | | ``[]``, |
+| | | | ``+``, ``-``, |
+| | | | str() |
+---------------+------------------------+ +----------------+
-| ``wchar_t *`` | same as pointers | | ``[]``, |
-| | (passing a unicode as | | ``+``, ``-``, |
-| | func argument is not | | unicode() |
-| | implemented) | | |
+| ``wchar_t *`` | same as pointers (*) | | ``[]``, |
+| | | | ``+``, ``-``, |
+| | | | unicode() |
+| | | | |
+---------------+------------------------+ +----------------+
-| pointers to | same as pointers | | ``[]``, |
+| pointers to | same as pointers (*) | | ``[]``, |
| structure or | | | ``+``, ``-``, |
| union | | | and read/write |
| | | | struct fields |
-+---------------+ | +----------------+
-| function | | | call |
++---------------+------------------------+ +----------------+
+| function | same as pointers | | call |
| pointers | | | |
+---------------+------------------------+------------------+----------------+
| arrays | a list or tuple of | a <cdata> | len(), iter(), |
@@ -941,6 +953,19 @@
| | | if out of range | |
+---------------+------------------------+------------------+----------------+
+(*) Note that when calling a function, as per C, a ``item *`` argument
+is identical to a ``item[]`` argument. So you can pass an argument that
+is accepted by either C type, like for example passing a Python string
+to a ``char *`` argument or a list of integers to a ``int *`` argument.
+Note that even if you want to pass a single ``item``, you need to specify
+it in a list of length 1; for example, a ``struct foo *`` argument might
+be passed as ``[[field1, field2...]]``.
+
+As an optimization, the CPython version of CFFI assumes that a function
+with a ``char *`` argument to which you pass a Python string will not
+actually modify the array of characters passed in, and so passes directly
+a pointer inside the Python string object.
+
Reference: verifier
-------------------
diff --git a/testing/backend_tests.py b/testing/backend_tests.py
--- a/testing/backend_tests.py
+++ b/testing/backend_tests.py
@@ -802,6 +802,28 @@
res = a(1) # and the error reported to stderr
assert res == 42
+ def test_structptr_argument(self):
+ ffi = FFI(backend=self.Backend())
+ ffi.cdef("struct foo_s { int a, b; };")
+ def cb(p):
+ return p[0].a * 1000 + p[0].b * 100 + p[1].a * 10 + p[1].b
+ a = ffi.callback("int(*)(struct foo_s[])", cb)
+ res = a([[5, 6], {'a': 7, 'b': 8}])
+ assert res == 5678
+ res = a([[5], {'b': 8}])
+ assert res == 5008
+
+ def test_array_argument_as_list(self):
+ ffi = FFI(backend=self.Backend())
+ ffi.cdef("struct foo_s { int a, b; };")
+ seen = []
+ def cb(argv):
+ seen.append(str(argv[0]))
+ seen.append(str(argv[1]))
+ a = ffi.callback("void(*)(char *[])", cb)
+ a([ffi.new("char[]", "foobar"), ffi.new("char[]", "baz")])
+ assert seen == ["foobar", "baz"]
+
def test_cast_float(self):
ffi = FFI(backend=self.Backend())
a = ffi.cast("float", 12)
diff --git a/testing/test_ctypes.py b/testing/test_ctypes.py
--- a/testing/test_ctypes.py
+++ b/testing/test_ctypes.py
@@ -13,3 +13,11 @@
def test_array_of_func_ptr(self):
py.test.skip("ctypes backend: not supported: "
"initializers for function pointers")
+
+ def test_structptr_argument(self):
+ py.test.skip("ctypes backend: not supported: passing a list "
+ "for a pointer argument")
+
+ def test_array_argument_as_list(self):
+ py.test.skip("ctypes backend: not supported: passing a list "
+ "for a pointer argument")
_______________________________________________
pypy-commit mailing list
pypy-commit@python.org
http://mail.python.org/mailman/listinfo/pypy-commit
