'str' to 'wxString' converstation is broken, when emoji is inside string

[timrid]

Operating system: Windows 10
wxPython version & source: wxPython 4.2.1 from pypi
Python version & source: Python 3.11

Description of the problem:
Whem some special emojis like 🔄 are inside a python string, that is converted to an wxString (internally in wxPython), the last character of the string is lost. It is also not shown in the gui.

The changes in 29520a1 creates the problem. I compiled it with and without the changes and only with the changes the error occures. I guess that PyUnicode_GET_LENGTH is not an 1:1 replacement for PyUnicode_GET_SIZE.

I created the following small example to illustrate the problem.

import wx

app = wx.App(False)
txt = wx.Button()

def test_string(input):
    txt.SetLabel(input)
    output = txt.GetLabel()
    print(f"{input=}, {output=}")

test_string("asdf")  # Output is: "asdf" -> ok
test_string("⟳ asdf")  # Output is: "⟳ asdf" -> ok
test_string("🔄 asdf")  # Output is: "🔄 asd" -> error: last char is missing
test_string("🔄🔄 asdf")  # Output is: "🔄🔄 as" -> error: last two chars are missing
test_string("🔄")  # Output is: "\ud83d" -> error

[swt2c]

Thanks @timrid for the bug report and analysis. I had feared that PyUnicode_GET_LENGTH was not an 1:1 replacement for PyUnicode_GET_SIZE. Unfortunately, the documentation/examples for the Python C APIs is not very extensive.

I'll try to take a look.

[timrid]

I also looked at the Python C API. The docs are not so useful. But when i looked in the cpython code i saw this comment for PyUnicode_AsWideChar.

/* Convert a Unicode object to a wide character string.

   - If w is NULL: return the number of wide characters (including the null
     character) required to convert the unicode object. Ignore size argument.

   - Otherwise: return the number of wide characters (excluding the null
     character) written into w. Write at most size wide characters (including
     the null character). */
Py_ssize_t
PyUnicode_AsWideChar(PyObject *unicode,
                     wchar_t *w,
                     Py_ssize_t size)
{
    ...

So i guess that

size_t len = PyUnicode_GET_SIZE(uni);

can be replaced with

size_t len = PyUnicode_AsWideChar(uni, NULL, 0);