no struct.pack for unicode strings?

[Martin v. Loewis]

Radovan Garabik <spam at melkor.dnp.fmph.uniba.sk> writes:

> it should pack them as raw Py_UNICODE data. At least that is what
> I'd need

It happens that this is probably not what most people would expect:
Instead, they would expect to get an array of wchar_t (which may or
may not be the same thing as Py_UNICODE): why would any C API know
what Py_UNICODE is?

> This is exactly what I am trying to avoid, since I need to quickly loop
> over the strings (it is a dictionary index) written in the file - hence
> the C extension module.
> I am afraid that using PyUnicode_DecodeUTF8 (or anything that creates a 
> PyObject) would impose a big speed penalty.

Did you actually measure this? I very much doubt you can demonstrate
any speed penalty.

In any case, to pack this as Py_UNICODE, you need to
.encode("unicode-internal") your string, then just add it to whatever
else you are packing.

Regards,
Martin