[ python-Bugs-1221424 ] buffer() object broken.

[SourceForge.net]

Bugs item #1221424, was opened at 2005-06-15 18:17
Message generated for change (Settings changed) made by nascheme
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1221424&group_id=5470

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Python Interpreter Core
Group: Python 2.4
>Status: Closed
>Resolution: Wont Fix
Priority: 5
Submitted By: Ray Schumacher (rjs_org)
Assigned to: Neil Schemenauer (nascheme)
Summary: buffer() object broken.

Initial Comment:
Same as:
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1033720&group_id=5470:

Still broken in 2.4.1...!

C:\projects\temp\fftw>python
ActivePython 2.4.1 Build 245 (ActiveState Corp.) based on
Python 2.4.1 (#65, Mar 30 2005, 09:33:37) [MSC v.1310
32 bit (Intel)] on win32
Type "help", "copyright", "credits" or "license" for
more information.
>>> import Numeric
>>> i = Numeric.zeros(8)
>>> repr(buffer(i))
'<read-only buffer for 0x008D7200, size -1, offset 0 at
0x008D1B20>'
>>> repr(buffer(""))
'<read-only buffer for 0x0086A038, size -1, offset 0 at
0x008D1B20>'

(Numeric v23.8)

----------------------------------------------------------------------

Comment By: Ray Schumacher (rjs_org)
Date: 2005-06-20 18:11

Message:
Logged In: YES 
user_id=287034

No, since we have now used the PyObject_AsReadBuffer(). 

We use ctypes to transfer data returned from an C-API A/D
driver directly to Numeric arrays via address.

----------------------------------------------------------------------

Comment By: Neil Schemenauer (nascheme)
Date: 2005-06-18 17:26

Message:
Logged In: YES 
user_id=35752

Calling PyObject_AsReadBuffer() is the proper way to get the
address.  It should work with older versions of Python as
well.  The buffer() object was changed in 2.4 to no longer
keep a pointer to buffer memory, only the underlying
PyObject.  It would be possible to make the repr look the
same as pre-2.4 Python.  Do you still care?

----------------------------------------------------------------------

Comment By: Ray Schumacher (rjs_org)
Date: 2005-06-17 19:04

Message:
Logged In: YES 
user_id=287034

In 2.3, the actual adress of the first element of an array
could be parsed for doing a memmove(), and now, not, which
breaks some older code. If this is a "more Pythonic"
feature, then OK; the previous resolution said "Fixed in
bufferobject.c 2.25." without remark. I assumed that it had
been restored to the previous.
I now have:
    ptr = tpcnv[a.typecode()]()
    pythonapi.PyObject_AsReadBuffer(py_object(a), 
                                                       
byref(ptr), 
                                                       
byref(c_int()))
    ptr.refed_array = a
instead of:
    split(repr(buffer(a)))[5][:-1]
to get/use an address with ctypes...

----------------------------------------------------------------------

Comment By: Neil Schemenauer (nascheme)
Date: 2005-06-17 17:52

Message:
Logged In: YES 
user_id=35752

The repr of a buffer() object has indeed changed.  I don't
see how that's a bug though.

----------------------------------------------------------------------

Comment By: Ray Schumacher (rjs_org)
Date: 2005-06-17 17:31

Message:
Logged In: YES 
user_id=287034

repr(buffer("")) used to return:
<read-only buffer for 0x0086A038, ptr 0x008D0520, size 1 at 
0x008D1B20>

now returns:
>>> repr(buffer(""))
'<read-only buffer for 0x0086A038, size -1, offset 0 at
0x008D1B20>'

same as:
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1033720&group_id=5470


----------------------------------------------------------------------

Comment By: Neil Schemenauer (nascheme)
Date: 2005-06-16 18:49

Message:
Logged In: YES 
user_id=35752

Please explain exactly what you think is broken about the
buffer object.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105470&aid=1221424&group_id=5470