Why I cannot coerce python string to char*?

[Simo Salminen]

I have following code:

cdef class Test:
def get_stuff(self):
cdef str id = "test"
cdef char* id2 = id <- error

This produces error:
'str' objects do not support coercion to C types (use 'bytes'?)

How this should work? From document http://docs.cython.org/src/tutorial/strings.html
I though this should work.

[Damien Churchill]

2009/12/20 Simo Salminen <ssal...@gmail.com>:

bytes is just a byte string. I think it's following Py3 now, so str ==
unicode, and bytes == str

[Stefan Behnel]

Simo Salminen, 20.12.2009 22:06:

> I have following code:
>
> cdef class Test:
> def get_stuff(self):
> cdef str id = "test"
> cdef char* id2 = id <- error
>
> This produces error:
> 'str' objects do not support coercion to C types (use 'bytes'?)

'str' is a special type that is bytes in Py2 and unicode in Py3. It is
therefore not possible to assign it to a char* without prior treatment.

If you want a byte string, use bytes. If you use unicode, encode it
appropriately.

Note that there is some discussion going on amongst the core developers
regarding better ways to deal with this. As of now, nothing has been decided.

> How this should work? From document http://docs.cython.org/src/tutorial/strings.html
> I though this should work.

I can't find any reference to "cdef str" in that document.

Stefan

[Simo Salminen]

Ok thanks.

> > How this should work? From documenthttp://docs.cython.org/src/tutorial/strings.html

> > I though this should work.
>
> I can't find any reference to "cdef str" in that document.
>

I misunderstood that type str equals to python string (document says
this: cdef char* other_c_string = py_string).

[Cass]

So was there no easy fix to take a Python 2.6 str object, and pass it
into an external C function as a char *?

How do you check what version of Cython you are using, btw?

[Cass]

To be more specific, I am trying to take a Python 2.6 Unicode str, and
trying to convert it into a char * so that I can pass it into an
existing C function.
I can modify the Python wrapper layer (i.e. I don't have to use a
Python str) but cannot modify the C code. The input just consists of
alphabetical words of varying lengths.
Any suggestions? :)

Thanks!

Cass

[Stefan Behnel]

Cass, 20.07.2010 22:28:

> To be more specific, I am trying to take a Python 2.6 Unicode str, and
> trying to convert it into a char * so that I can pass it into an
> existing C function.
> I can modify the Python wrapper layer (i.e. I don't have to use a
> Python str) but cannot modify the C code. The input just consists of
> alphabetical words of varying lengths.
> Any suggestions? :)

Well, yes, read the extensive docs on string handling.

Admittedly, the most up-to-date tutorial for Cython 0.13 is not on the web
site yet, but here it is in text form:

http://hg.cython.org/cython-docs/raw-file/tip/src/tutorial/strings.rst

In short: switch to using bytes objects internally and convert unicode
string input to bytes when receiving it from Python space by encoding it in
a suitable encoding.

Stefan

[Cass]

Thanks!! It worked. Just used:

py_byte_string = py_unicode_string.encode('UTF-8')
cdef char* c_string = py_byte_string

However, now I'm confused. After my last post, I read at (http://
farmdev.com/talks/unicode/) that the default encoding for Python 2 is
ASCII. Was that information wrong?

[Stefan Behnel]

Hi, please avoid top-posting.

Cass, 20.07.2010 23:01:

> Thanks!! It worked. Just used:
>
> py_byte_string = py_unicode_string.encode('UTF-8')
> cdef char* c_string = py_byte_string

Looks good.

> However, now I'm confused. After my last post, I read at (http://
> farmdev.com/talks/unicode/) that the default encoding for Python 2 is
> ASCII. Was that information wrong?

Yes. The default encoding in Python 2 is configurable and platform
specific. Don't rely on it being anything.

Stefan