The PyBytes (pep3137 buffer) .extend() method currently only accepts as
input something supporting the pep3118 buffer API.  It also needs to
accept an iterable of ints in the 0..255 range.

Here a patch that adds support for any iterable (or sequence) of
integers to bytearray.extend().

Made 2 minor enhancements to the patch:
   + Added the proper type-cast to PyMem_Realloc call.
   + Changed (len >> 1) to (len >> 1) + 1, just to be sure that the
buffer  doesn't overflow if __length_hint__ return 0 or 1 erroneously.

There is a reference leak in my previous patches. So, I updated (again)
the patch. There is still another possible leak if the PyMem_Realloc
return NULL (i.e., the system is out of memory), but I don't think it
worth fixing.

Here yet another revision of the patch. This one makes
bytearray.extend()   try to determine the length of its argument a bit
more aggressively -- i.e., also uses PyObject_Length().

"There is still another possible leak if the PyMem_Realloc
return NULL (i.e., the system is out of memory), but I don't think it
worth fixing."

Do it.  It looks easy: a Py_DECREF(it) before the return PyErr_NoMemory().

Done. Is there any other issue with the patch?

reading 5.patch over...

Any particular reason for using buf_size = 32 when the length isn't
known up front?  add a comment saying why (or that its just a magic
guess).  anyways it sounds like a fine starting value.  picking
anything "better" would require profiling.

perhaps use a list comprehension instead of map() in the unit test?
either works, its a style thing.

  (int(x) for x in orig*50)    [int(x) for x in orig*50]

also the uses of 5 and -5 in that test could be written using
len(orig) instead of 5.

add another assertRaises that tests to make sure a list with -1 in it
raises a ValueError.

While I dislike that this code makes a temporary copy of the data
first, doing otherwise is more complicated so the simplicity of this
one wins.  Leave that optimization for later.  Your code looks good.

-gps

On 12/2/07, Alexandre Vassalotti <report@bugs.python.org> wrote:
>
> Alexandre Vassalotti added the comment:
>
> Done. Is there any other issue with the patch?
>
> Added file: http://bugs.python.org/file8857/byte_extend-5.patch
>
> __________________________________
> Tracker <report@bugs.python.org>
> <http://bugs.python.org/issue1283>
> __________________________________
>

Thank you Gregory for the review!

Committed to r59314.