[ctypes-users] comments on this ctypes static variable idiom?

[Marcus G. <mag...@gm...>]

I implemented a strange (to me) pythonic/ctypes equivalent of a static
variable for a function that iteratively calls a DLL to update a structure
(which contains buffers).  It appears that my buffer (allocated by python)
goes out-of-scope between calls, even though the structure I iterate on
contains a pointer to the buffer memory.  This seems reasonable, but I
wonder: is there a better idiom?

Here is a snippet which I hope illustrates the item.  Comments and
suggestions for improvement appreciated!
Marcus



def foo(strData, aStruct=None):
    # 1: This function is called iteratively; we must track pointer changes
    if aStruct is None:
        # create our processing structure
        myStructure = THE_STRUCTURE()

        # set the initial buffer pointer to NULL; indicates 1st buffer
update
        myStructure.pOldBuffer = None

        # create the buffer; use ctypes.create_string_buffer for mutable
memory blocks
        tbuf = create_string_buffer('', 100)

        myStructure.sReturnObject.pBuffer = addressof(tbuf)
        myStructure.sReturnObject.uiBufferSize = 100

        # static variable here...  this seems fragile.  Is there a better
method?  If I don't
        # do this, tbuf goes out-of-scope between function calls and I get
exceptions
        foo.tbuf = tbuf
    else:
        # we are iterating; use the existing struct and set the old buffer
accordingly
        myStructure = aStruct
        myStructure.pOldBuffer = myStructure.sReturnObject.pBuffer

    # ... some processing ...

   # the C call
   cdll.theLibraryCall(byref(myStructure))

   # return the structure for analysis, and potentially a future call
   return myStructure