The C code often uses %.<number><format> in PyUnicode_FromFormat(). %.200s protects from unlimited output when broken pointer points on random non-null-terminated data. %.200R is used to limit the size of human-readable messages.

In all these case formatted string can look well-formed with short data, but mis-formed (not closed quote, truncated backslash escaping or � decoded from truncated UTF-8 sequence) with long data.

I propose to make truncating in PyUnicode_FromFormat() more smart.

1. Truncated %R should keep at least one end character (the quote or ">").
2. Truncated output should include "..." or "[...]" as truncating sign.
3. \c, \OOO, \xXX, \uXXXX, and \UXXXXXXXX should not be truncated. It is better to omit these sequences at all (cut the string before them) that output them truncated.
4. Doesn't truncate UTF-8 sequence inside a character for %s.

See my old issue #10833 which proposed to *remove* the arbitrary limit
on strings. It was rejected.

Could we make this feature available at the Python level too? It sounds
really useful.

--Guido (mobile)
On Jan 12, 2016 2:01 AM, "STINNER Victor" <report@bugs.python.org> wrote:

>
> STINNER Victor added the comment:
>
> See my old issue #10833 which proposed to *remove* the arbitrary limit
> on strings. It was rejected.
>
> ----------
>
> _______________________________________
> Python tracker <report@bugs.python.org>
> <http://bugs.python.org/issue26090>
> _______________________________________
>

I think we can make this feature available with classic formatting '%.100r', but with new formatting '{0:.100!r}' (especially with f-strings) this can be not so easy.

Well it seems a little odd to spend effort on a corner case of the C-level
error messages if we can't even replicate it in pure Python.