The main bottleneck was the PyUnicode_READ() macro.

Patch written by Arnaud Calmettes.

Patch written by Arnaud Calmettes.

Use findchar() which is more optimized than a dummy loop using
PyUnicode_READ().  PyUnicode_READ() is a complex and slow macro.

bytes.find() and handle correctly OverflowError (raise the same ValueError than
the error for -1).

Patch written by Brice Berna.

Patch written by Brice Berna.

Patch written by Jérémy Anger.

Patch written by Jérémy Anger.

Thanks Jérémy Anger for the fix.

Farrell.

Search the invalid character using mbrtowc().

instead of surrogateescape

So it would be possible to support more error handlers later.

Add also FIXMEs in unicodeobject.c

 * Use PyUnicode_EncodeLocale() in time.strftime() if wcsftime() is not
   available
 * Document my last changes in Misc/NEWS

 * PyUnicode_DecodeLocaleAndSize() and PyUnicode_DecodeLocale() decode a string
   from the current locale encoding
 * _Py_char2wchar() writes an "error code" in the size argument to indicate
   if the function failed because of memory allocation failure or because of a
   decoding error. The function doesn't write the error message directly to
   stderr.
 * Fix time.strftime() (if wcsftime() is missing): decode strftime() result
   from the current locale encoding, not from the filesystem encoding.

line argument cannot be decoded

cannot be decoded

encoding and the surrogateescape error handler, rather than UTF-8.  Patch
by David Watson.

Original patch by Hallvard B Furuseth.

Original patch by Hallvard B Furuseth.

Initial patch by Matthias Troffaes.

Thanks Mark Shannon.