sdl2.ext.array - Tools for Working with ctypes Arrays¶
This module provides a number of utilites for accessing data in ctypes
arrays and converting ctypes
arrays into common Python formats.
Warning
These functions are primarily meant for internal use in PySDL2, and should be treated as experimental.
Providing read-write access for sequential data¶
Two classes allow you to access sequential data in different ways. The
CTypesView
provides byte-wise access to iterable objects and allows
you to convert the object representation to matching byte-widths for
ctypes
or other modules.
Depending on the the underlying object and the chosen size of each particular
item of the object, the CTypesView
allows you to operate directly
on different representations of the object’s contents.
>>> text = bytearray("Hello, I am a simple ASCII string!")
>>> ctview = CTypesView(text, itemsize=1)
>>> ctview.view[0] = 0x61
>>> print(text)
aello, I am a simple ASCII string!"
>>> ctview.to_uint16()[3] = 0x6554
>>> print(text)
aello,Te am a simple ASCII string!"
The snippet above provides a single-byte sized view on a bytearray()
object. Afterwards, the first item of the view is changed, which causes a
change on the bytearray()
, on the first item as well, since both, the
CTypesView
and the bytearray()
provide a byte-wise access to
the contents.
By using CTypesView.to_uint16()
, we change the access representation to
a 2-byte unsigned integer ctypes
pointer and change the fourth 2-byte
value, I to something else.
>>> text = bytearray("Hello, I am a simple ASCII string!")
>>> ctview = CTypesView(text, itemsize=2)
>>> ctview.view[0] = 0x61
>>> print(text)
aello, I am a simple ASCII string!"
>>> ctview.to_uint16()[3] = 0x6554
>>> print(text) aello,Te am a simple ASCII string!"
If the encapsuled object does not provide a (writable) buffer()
interface, but is iterable, the CTypesView
will create an
internal copy of the object data using Python’s array
module and
perform all operations on that copy.
>>> mylist = [18, 52, 86, 120, 154, 188, 222, 240]
>>> ctview = CTypesView(mylist, itemsize=1, docopy=True)
>>> print(ctview.object)
array('B', [18, 52, 86, 120, 154, 188, 222, 240])
>>> ctview.view[3] = 0xFF
>>> print(mylist)
[18, 52, 86, 120, 154, 188, 222, 240]
>>> print(ctview.object)
array('B', [18, 52, 86, 255, 154, 188, 222, 240])
As for directly accessible objects, you can define your own itemsize to be used. If the iterable does not provide a direct byte access to their contents, this won’t have any effect except for resizing the item widths.
>>> mylist = [18, 52, 86, 120, 154, 188, 222, 240]
>>> ctview = CTypesView(mylist, itemsize=4, docopy=True)
>>> print(ctview.object)
array('I', [18L, 52L, 86L, 120L, 154L, 188L, 222L, 240L])
Accessing data over multiple dimensions¶
The second class, MemoryView
provides an interface to access
data over multiple dimensions. You can layout and access a simple
byte stream over e.g. two or more axes, providing a greater flexibility
for functional operations and complex data.
Let’s assume, we are reading image data from a file stream into some buffer object and want to access and manipulate the image data. Images feature two axes, one being the width, the other being the height, defining a rectangular graphics area.
When we read all data from the file, we have an one-dimensional view of the
image graphics. The MemoryView
allows us to define a
two-dimensional view over the image graphics, so that we can operate on
both, rows and columns of the image.
>>> imagedata = bytearray("some 1-byte graphics data")
>>> view = MemoryView(imagedata, 1, (5, 5))
>>> print(view)
[[s, o, m, e, ], [1, -, b, y, t], [e, , g, r, a], [p, h, i, c, s], [ , d, a, t, a]]
>>> for row in view:
... print(row)
...
[s, o, m, e, ]
[1, -, b, y, t]
[e, , g, r, a]
[p, h, i, c, s]
[ , d, a, t, a]
>>> for row in view:
... row[1] = "X"
... print(row)
...
[s, X, m, e, ]
[1, X, b, y, t]
[e, X, g, r, a]
[p, X, i, c, s]
[ , X, a, t, a]
>>> print(imagedata)
sXme 1XbyteXgrapXics Xata
On accessing a particular dimension of a MemoryView
, a new
MemoryView
is created, if it does not access a single
element.
>>> firstrow = view[0]
>>> type(firstrow)
<class 'sdl2.ext.array.MemoryView'>
>>> type(firstrow[0])
<type 'bytearray'>
A MemoryView
features, similar to Python’s builtin
memoryview
, dimensions and strides, accessible via the
MemoryView.ndim
and MemoryView.strides
attributes.
>>> view.ndim
2
>>> view.strides
(5, 5)
The MemoryView.strides
, which have to be passed on creating a
new MemoryView
, define the layout of the data over different
dimensions. In the example above, we created a 5x5 two-dimensional view
to the image graphics.
>>> twobytes = MemoryView(imagedata, 2, (5, 1))
>>> print(twobytes)
[[sX, me, 1, Xb, yt], [eX, gr, ap, Xi, cs]]
Array API¶
-
sdl2.ext.array.
to_ctypes
(dataseq, dtype, mcount=0)[source]¶ Converts an arbitrary sequence to a
ctypes
array of a given type.Parameters: - dataseq – A sequence to convert to a
ctypes
array. - dtype (class) – The
ctypes
data type to use for the array (e.g.ctypes.c_uint8
). - mcount (int, optional) – The number of elements in
dataseq
. If not specified, this will be inferred automatically.
Returns: A tuple in the form
(valset, count)
, wherevalset
is the convertedctypes
array andcount
is the number of elements in the array.Return type: Raises: TypeError
– if any elements in the passed sequence do not match the specified type.- dataseq – A sequence to convert to a
-
sdl2.ext.array.
create_array
(obj, itemsize)[source]¶ Creates an
array.array
copy of a given object.Parameters: - obj – The object from which the array will be created.
- itemsize – The size (in bytes) of each item in the given object.
Returns: The array created from the object.
Return type:
-
class
sdl2.ext.array.
CTypesView
(obj, itemsize=1, docopy=False, objsize=None)[source]¶ A proxy for accessing byte-wise
ctypes
data types.This class provides read-write access for arbitrary
ctypes
objects that are iterable. In case the object does not provide abuffer()
interface for direct access, aCTypesView
can copy the object’s contents into an internal buffer from which data can be retrieved.Depending on the item type stored in the iterable object, you might need to manually specify the object’s item size (in bytes). Additionally, you may need to manually specify the number of items in the iterable if it does not properly return its length using the
len
function.For certain types, such as the bytearray, the original object must not be reassigned after being encapsuled and used in ctypes bindings if the contents are not copied.
Parameters: - obj – An arbitrary iterable
ctypes
object to access. - itemsize (int, optional) – The size (in bytes) of each item of the
iterable object. Defaults to
1
. - docopy (bool, optional) – If True, the view will be created for a copy
of the object and will not modify the original
ctypes
instance. Defalts to False (original object will be modified). - objsize (int, optional) – The number of items in the iterable object.
If not specified, the number of items will try to be inferred
automatically. Defaults to
None
(automatic inferrence).
-
to_bytes
()[source]¶ Casts the object to an array of bytes.
If the view was created with
docopy = False
(the default), the returned object provides direct read-write access to the object data. Otherwise, the returned object will only modify a copy of the data.Returns: A pointer to a ctypes.c_uint8
array.Return type: ctypes._Pointer
-
to_uint16
()[source]¶ Casts the object to an array of 16-bit unsigned ints.
If the view was created with
docopy = False
(the default), the returned object provides direct read-write access to the object data. Otherwise, the returned object will only modify a copy of the data.Returns: A pointer to a ctypes.c_uint16
array.Return type: ctypes._Pointer
-
to_uint32
()[source]¶ Casts the object to an array of 32-bit unsigned ints.
If the view was created with
docopy = False
(the default), the returned object provides direct read-write access to the object data. Otherwise, the returned object will only modify a copy of the data.Returns: A pointer to a ctypes.c_uint32
array.Return type: ctypes._Pointer
-
to_uint64
()[source]¶ Casts the object to an array of 64-bit unsigned ints.
If the view was created with
docopy = False
(the default), the returned object provides direct read-write access to the object data. Otherwise, the returned object will only modify a copy of the data.Returns: A pointer to a ctypes.c_uint64
array.Return type: ctypes._Pointer
-
view
¶ Provides a read/write-aware
ctypes
view of the encapsuled object.
Whether any modifications to the view will also modify the underlying object.
Type: bool
-
object
¶ The underlying object.
- obj – An arbitrary iterable
-
class
sdl2.ext.array.
MemoryView
(source, itemsize, strides, getfunc=None, setfunc=None, srcsize=None)[source]¶ A class that provides read-write access to indexable
ctypes
objects.Note
MemoryView
makes heavy use of recursion for multi-dimensional access, making it slow for many use-cases. For better performance, thenumpy
library can be used for fast access to many array-like data types, but it may support fewer types of arbitrary objects than theMemoryView
class.Parameters: - source – An arbitrary indexable
ctypes
object to access. - itemsize (int) – The size (in bytes) of each item of the indexable object.
- strides (tuple) – The length of each dimension in the object (e.g.
(4, 3)
for a 4 x 3 2D array). - getfunc (function, optional) – Deprecated, do not use.
- setfunc (function, optional) – Deprecated, do not use.
- srcsize (int, optional) – The size (in bytes) of the input object.
If
len(source)
returns the size of the object in bytes this will be inferred automatically, otherwise it must be manually specified.
-
source
¶ The underlying data source.
- source – An arbitrary indexable