From 31084ba52856329b972b84299b438bf5b73be292 Mon Sep 17 00:00:00 2001 From: Antoine Pitrou Date: Thu, 19 Mar 2015 23:29:36 +0100 Subject: [PATCH] Issue #23632: Memoryviews now allow tuple indexing (including for multi-dimensional memoryviews). --- Doc/library/stdtypes.rst | 38 +++++----- Lib/test/test_buffer.py | 56 +++++++++++++-- Misc/NEWS | 3 + Objects/memoryobject.c | 149 +++++++++++++++++++++++++++++++-------- 4 files changed, 195 insertions(+), 51 deletions(-) diff --git a/Doc/library/stdtypes.rst b/Doc/library/stdtypes.rst index af0a85eb544..739e90d4b1e 100644 --- a/Doc/library/stdtypes.rst +++ b/Doc/library/stdtypes.rst @@ -3282,10 +3282,8 @@ copying. the view. The :class:`~memoryview.itemsize` attribute will give you the number of bytes in a single element. - A :class:`memoryview` supports slicing to expose its data. If - :class:`~memoryview.format` is one of the native format specifiers - from the :mod:`struct` module, indexing will return a single element - with the correct type. Full slicing will result in a subview:: + A :class:`memoryview` supports slicing and indexing to expose its data. + One-dimensional slicing will result in a subview:: >>> v = memoryview(b'abcefg') >>> v[1] @@ -3297,25 +3295,29 @@ copying. >>> bytes(v[1:4]) b'bce' - Other native formats:: + If :class:`~memoryview.format` is one of the native format specifiers + from the :mod:`struct` module, indexing with an integer or a tuple of + integers is also supported and returns a single *element* with + the correct type. One-dimensional memoryviews can be indexed + with an integer or a one-integer tuple. Multi-dimensional memoryviews + can be indexed with tuples of exactly *ndim* integers where *ndim* is + the number of dimensions. Zero-dimensional memoryviews can be indexed + with the empty tuple. + + Here is an example with a non-byte format:: >>> import array >>> a = array.array('l', [-11111111, 22222222, -33333333, 44444444]) - >>> a[0] + >>> m = memoryview(a) + >>> m[0] -11111111 - >>> a[-1] + >>> m[-1] 44444444 - >>> a[2:3].tolist() - [-33333333] - >>> a[::2].tolist() + >>> m[::2].tolist() [-11111111, -33333333] - >>> a[::-1].tolist() - [44444444, -33333333, 22222222, -11111111] - .. versionadded:: 3.3 - - If the underlying object is writable, the memoryview supports slice - assignment. Resizing is not allowed:: + If the underlying object is writable, the memoryview supports + one-dimensional slice assignment. Resizing is not allowed:: >>> data = bytearray(b'abcefg') >>> v = memoryview(data) @@ -3348,12 +3350,16 @@ copying. True .. versionchanged:: 3.3 + One-dimensional memoryviews can now be sliced. One-dimensional memoryviews with formats 'B', 'b' or 'c' are now hashable. .. versionchanged:: 3.4 memoryview is now registered automatically with :class:`collections.abc.Sequence` + .. versionchanged:: 3.5 + memoryviews can now be indexed with tuple of integers. + :class:`memoryview` has several methods: .. method:: __eq__(exporter) diff --git a/Lib/test/test_buffer.py b/Lib/test/test_buffer.py index d921a5aacc9..0d672dd2ac2 100644 --- a/Lib/test/test_buffer.py +++ b/Lib/test/test_buffer.py @@ -11,6 +11,7 @@ # memoryview tests is now in this module. # +import contextlib import unittest from test import support from itertools import permutations, product @@ -2825,6 +2826,13 @@ class TestBufferProtocol(unittest.TestCase): m = memoryview(ex) self.assertRaises(TypeError, eval, "9.0 in m", locals()) + @contextlib.contextmanager + def assert_out_of_bounds_error(self, dim): + with self.assertRaises(IndexError) as cm: + yield + self.assertEqual(str(cm.exception), + "index out of bounds on dimension %d" % (dim,)) + def test_memoryview_index(self): # ndim = 0 @@ -2851,12 +2859,31 @@ class TestBufferProtocol(unittest.TestCase): self.assertRaises(IndexError, m.__getitem__, -8) self.assertRaises(IndexError, m.__getitem__, 8) - # Not implemented: multidimensional sub-views + # multi-dimensional ex = ndarray(list(range(12)), shape=[3,4], flags=ND_WRITABLE) m = memoryview(ex) - self.assertRaises(NotImplementedError, m.__getitem__, 0) - self.assertRaises(NotImplementedError, m.__setitem__, 0, 9) + self.assertEqual(m[0, 0], 0) + self.assertEqual(m[2, 0], 8) + self.assertEqual(m[2, 3], 11) + self.assertEqual(m[-1, -1], 11) + self.assertEqual(m[-3, -4], 0) + + # out of bounds + for index in (3, -4): + with self.assert_out_of_bounds_error(dim=1): + m[index, 0] + for index in (4, -5): + with self.assert_out_of_bounds_error(dim=2): + m[0, index] + self.assertRaises(IndexError, m.__getitem__, (2**64, 0)) + self.assertRaises(IndexError, m.__getitem__, (0, 2**64)) + + self.assertRaises(TypeError, m.__getitem__, (0, 0, 0)) + self.assertRaises(TypeError, m.__getitem__, (0.0, 0.0)) + + # Not implemented: multidimensional sub-views + self.assertRaises(NotImplementedError, m.__getitem__, ()) self.assertRaises(NotImplementedError, m.__getitem__, 0) def test_memoryview_assign(self): @@ -2945,10 +2972,27 @@ class TestBufferProtocol(unittest.TestCase): m = memoryview(ex) self.assertRaises(NotImplementedError, m.__setitem__, 0, 1) - # Not implemented: multidimensional sub-views + # multi-dimensional ex = ndarray(list(range(12)), shape=[3,4], flags=ND_WRITABLE) m = memoryview(ex) + m[0,1] = 42 + self.assertEqual(ex[0][1], 42) + m[-1,-1] = 43 + self.assertEqual(ex[2][3], 43) + # errors + for index in (3, -4): + with self.assert_out_of_bounds_error(dim=1): + m[index, 0] = 0 + for index in (4, -5): + with self.assert_out_of_bounds_error(dim=2): + m[0, index] = 0 + self.assertRaises(IndexError, m.__setitem__, (2**64, 0), 0) + self.assertRaises(IndexError, m.__setitem__, (0, 2**64), 0) + self.assertRaises(TypeError, m.__setitem__, (0, 0, 0), 0) + self.assertRaises(TypeError, m.__setitem__, (0.0, 0.0), 0) + + # Not implemented: multidimensional sub-views self.assertRaises(NotImplementedError, m.__setitem__, 0, [2, 3]) def test_memoryview_slice(self): @@ -2961,8 +3005,8 @@ class TestBufferProtocol(unittest.TestCase): self.assertRaises(ValueError, m.__setitem__, slice(0,2,0), bytearray([1,2])) - # invalid slice key - self.assertRaises(TypeError, m.__getitem__, ()) + # 0-dim slicing (identity function) + self.assertRaises(NotImplementedError, m.__getitem__, ()) # multidimensional slices ex = ndarray(list(range(12)), shape=[12], flags=ND_WRITABLE) diff --git a/Misc/NEWS b/Misc/NEWS index 4dd5c446497..c99a1c95d34 100644 --- a/Misc/NEWS +++ b/Misc/NEWS @@ -10,6 +10,9 @@ Release date: 2015-03-28 Core and Builtins ----------------- +- Issue #23632: Memoryviews now allow tuple indexing (including for + multi-dimensional memoryviews). + - Issue #23192: Fixed generator lambdas. Patch by Bruno Cauet. - Issue #23629: Fix the default __sizeof__ implementation for variable-sized diff --git a/Objects/memoryobject.c b/Objects/memoryobject.c index 6142b4cf739..b6951ab88e0 100644 --- a/Objects/memoryobject.c +++ b/Objects/memoryobject.c @@ -192,10 +192,10 @@ PyTypeObject _PyManagedBuffer_Type = { #define VIEW_ADDR(mv) (&((PyMemoryViewObject *)mv)->view) /* Check for the presence of suboffsets in the first dimension. */ -#define HAVE_PTR(suboffsets) (suboffsets && suboffsets[0] >= 0) +#define HAVE_PTR(suboffsets, dim) (suboffsets && suboffsets[dim] >= 0) /* Adjust ptr if suboffsets are present. */ -#define ADJUST_PTR(ptr, suboffsets) \ - (HAVE_PTR(suboffsets) ? *((char**)ptr) + suboffsets[0] : ptr) +#define ADJUST_PTR(ptr, suboffsets, dim) \ + (HAVE_PTR(suboffsets, dim) ? *((char**)ptr) + suboffsets[dim] : ptr) /* Memoryview buffer properties */ #define MV_C_CONTIGUOUS(flags) (flags&(_Py_MEMORYVIEW_SCALAR|_Py_MEMORYVIEW_C)) @@ -332,11 +332,11 @@ copy_base(const Py_ssize_t *shape, Py_ssize_t itemsize, char *p; Py_ssize_t i; for (i=0, p=mem; i < shape[0]; p+=itemsize, sptr+=sstrides[0], i++) { - char *xsptr = ADJUST_PTR(sptr, ssuboffsets); + char *xsptr = ADJUST_PTR(sptr, ssuboffsets, 0); memcpy(p, xsptr, itemsize); } for (i=0, p=mem; i < shape[0]; p+=itemsize, dptr+=dstrides[0], i++) { - char *xdptr = ADJUST_PTR(dptr, dsuboffsets); + char *xdptr = ADJUST_PTR(dptr, dsuboffsets, 0); memcpy(xdptr, p, itemsize); } } @@ -364,8 +364,8 @@ copy_rec(const Py_ssize_t *shape, Py_ssize_t ndim, Py_ssize_t itemsize, } for (i = 0; i < shape[0]; dptr+=dstrides[0], sptr+=sstrides[0], i++) { - char *xdptr = ADJUST_PTR(dptr, dsuboffsets); - char *xsptr = ADJUST_PTR(sptr, ssuboffsets); + char *xdptr = ADJUST_PTR(dptr, dsuboffsets, 0); + char *xsptr = ADJUST_PTR(sptr, ssuboffsets, 0); copy_rec(shape+1, ndim-1, itemsize, xdptr, dstrides+1, dsuboffsets ? dsuboffsets+1 : NULL, @@ -2057,7 +2057,7 @@ tolist_base(const char *ptr, const Py_ssize_t *shape, return NULL; for (i = 0; i < shape[0]; ptr+=strides[0], i++) { - const char *xptr = ADJUST_PTR(ptr, suboffsets); + const char *xptr = ADJUST_PTR(ptr, suboffsets, 0); item = unpack_single(xptr, fmt); if (item == NULL) { Py_DECREF(lst); @@ -2091,7 +2091,7 @@ tolist_rec(const char *ptr, Py_ssize_t ndim, const Py_ssize_t *shape, return NULL; for (i = 0; i < shape[0]; ptr+=strides[0], i++) { - const char *xptr = ADJUST_PTR(ptr, suboffsets); + const char *xptr = ADJUST_PTR(ptr, suboffsets, 0); item = tolist_rec(xptr, ndim-1, shape+1, strides+1, suboffsets ? suboffsets+1 : NULL, fmt); @@ -2171,33 +2171,66 @@ memory_repr(PyMemoryViewObject *self) /* Indexing and slicing */ /**************************************************************************/ -/* Get the pointer to the item at index. */ static char * -ptr_from_index(Py_buffer *view, Py_ssize_t index) +lookup_dimension(Py_buffer *view, char *ptr, int dim, Py_ssize_t index) { - char *ptr; - Py_ssize_t nitems; /* items in the first dimension */ + Py_ssize_t nitems; /* items in the given dimension */ assert(view->shape); assert(view->strides); - nitems = view->shape[0]; + nitems = view->shape[dim]; if (index < 0) { index += nitems; } if (index < 0 || index >= nitems) { - PyErr_SetString(PyExc_IndexError, "index out of bounds"); + PyErr_Format(PyExc_IndexError, + "index out of bounds on dimension %d", dim + 1); return NULL; } - ptr = (char *)view->buf; - ptr += view->strides[0] * index; + ptr += view->strides[dim] * index; - ptr = ADJUST_PTR(ptr, view->suboffsets); + ptr = ADJUST_PTR(ptr, view->suboffsets, dim); return ptr; } +/* Get the pointer to the item at index. */ +static char * +ptr_from_index(Py_buffer *view, Py_ssize_t index) +{ + char *ptr = (char *)view->buf; + return lookup_dimension(view, ptr, 0, index); +} + +/* Get the pointer to the item at tuple. */ +static char * +ptr_from_tuple(Py_buffer *view, PyObject *tup) +{ + char *ptr = (char *)view->buf; + Py_ssize_t dim, nindices = PyTuple_GET_SIZE(tup); + + if (nindices > view->ndim) { + PyErr_Format(PyExc_TypeError, + "cannot index %zd-dimension view with %zd-element tuple", + view->ndim, nindices); + return NULL; + } + + for (dim = 0; dim < nindices; dim++) { + Py_ssize_t index; + index = PyNumber_AsSsize_t(PyTuple_GET_ITEM(tup, dim), + PyExc_IndexError); + if (index == -1 && PyErr_Occurred()) + return NULL; + ptr = lookup_dimension(view, ptr, dim, index); + if (ptr == NULL) + return NULL; + } + return ptr; +} + /* Return the item at index. In a one-dimensional view, this is an object with the type specified by view->format. Otherwise, the item is a sub-view. The function is used in memory_subscript() and memory_as_sequence. */ @@ -2229,6 +2262,32 @@ memory_item(PyMemoryViewObject *self, Py_ssize_t index) return NULL; } +/* Return the item at position *key* (a tuple of indices). */ +static PyObject * +memory_item_multi(PyMemoryViewObject *self, PyObject *tup) +{ + Py_buffer *view = &(self->view); + const char *fmt; + Py_ssize_t nindices = PyTuple_GET_SIZE(tup); + char *ptr; + + CHECK_RELEASED(self); + + fmt = adjust_fmt(view); + if (fmt == NULL) + return NULL; + + if (nindices < view->ndim) { + PyErr_SetString(PyExc_NotImplementedError, + "sub-views are not implemented"); + return NULL; + } + ptr = ptr_from_tuple(view, tup); + if (ptr == NULL) + return NULL; + return unpack_single(ptr, fmt); +} + Py_LOCAL_INLINE(int) init_slice(Py_buffer *base, PyObject *key, int dim) { @@ -2277,6 +2336,22 @@ is_multislice(PyObject *key) return 1; } +static Py_ssize_t +is_multiindex(PyObject *key) +{ + Py_ssize_t size, i; + + if (!PyTuple_Check(key)) + return 0; + size = PyTuple_GET_SIZE(key); + for (i = 0; i < size; i++) { + PyObject *x = PyTuple_GET_ITEM(key, i); + if (!PyIndex_Check(x)) + return 0; + } + return 1; +} + /* mv[obj] returns an object holding the data for one element if obj fully indexes the memoryview or another memoryview object if it does not. @@ -2332,6 +2407,9 @@ memory_subscript(PyMemoryViewObject *self, PyObject *key) return (PyObject *)sliced; } + else if (is_multiindex(key)) { + return memory_item_multi(self, key); + } else if (is_multislice(key)) { PyErr_SetString(PyExc_NotImplementedError, "multi-dimensional slicing is not implemented"); @@ -2376,14 +2454,15 @@ memory_ass_sub(PyMemoryViewObject *self, PyObject *key, PyObject *value) return -1; } } - if (view->ndim != 1) { - PyErr_SetString(PyExc_NotImplementedError, - "memoryview assignments are currently restricted to ndim = 1"); - return -1; - } if (PyIndex_Check(key)) { - Py_ssize_t index = PyNumber_AsSsize_t(key, PyExc_IndexError); + Py_ssize_t index; + if (1 < view->ndim) { + PyErr_SetString(PyExc_NotImplementedError, + "sub-views are not implemented"); + return -1; + } + index = PyNumber_AsSsize_t(key, PyExc_IndexError); if (index == -1 && PyErr_Occurred()) return -1; ptr = ptr_from_index(view, index); @@ -2418,7 +2497,19 @@ memory_ass_sub(PyMemoryViewObject *self, PyObject *key, PyObject *value) PyBuffer_Release(&src); return ret; } - else if (PySlice_Check(key) || is_multislice(key)) { + if (is_multiindex(key)) { + char *ptr; + if (PyTuple_GET_SIZE(key) < view->ndim) { + PyErr_SetString(PyExc_NotImplementedError, + "sub-views are not implemented"); + return -1; + } + ptr = ptr_from_tuple(view, key); + if (ptr == NULL) + return -1; + return pack_single(ptr, value, fmt); + } + if (PySlice_Check(key) || is_multislice(key)) { /* Call memory_subscript() to produce a sliced lvalue, then copy rvalue into lvalue. This is already implemented in _testbuffer.c. */ PyErr_SetString(PyExc_NotImplementedError, @@ -2591,8 +2682,8 @@ cmp_base(const char *p, const char *q, const Py_ssize_t *shape, int equal; for (i = 0; i < shape[0]; p+=pstrides[0], q+=qstrides[0], i++) { - const char *xp = ADJUST_PTR(p, psuboffsets); - const char *xq = ADJUST_PTR(q, qsuboffsets); + const char *xp = ADJUST_PTR(p, psuboffsets, 0); + const char *xq = ADJUST_PTR(q, qsuboffsets, 0); equal = unpack_cmp(xp, xq, fmt, unpack_p, unpack_q); if (equal <= 0) return equal; @@ -2626,8 +2717,8 @@ cmp_rec(const char *p, const char *q, } for (i = 0; i < shape[0]; p+=pstrides[0], q+=qstrides[0], i++) { - const char *xp = ADJUST_PTR(p, psuboffsets); - const char *xq = ADJUST_PTR(q, qsuboffsets); + const char *xp = ADJUST_PTR(p, psuboffsets, 0); + const char *xq = ADJUST_PTR(q, qsuboffsets, 0); equal = cmp_rec(xp, xq, ndim-1, shape+1, pstrides+1, psuboffsets ? psuboffsets+1 : NULL, qstrides+1, qsuboffsets ? qsuboffsets+1 : NULL,