Different behaviour of regexp in 3.6.0b2

[Serhiy Storchaka]

On 14.10.16 18:40, Lele Gaifax wrote:
> Hi all,
>
> trying out pgcli with Python 3.6.0b2 I got an error related to what seem a
> different behaviour, or even a bug, of re.sub().
>
> The original intent is to replace spaces within a string with the regular expression
> \s+ (see https://github.com/dbcli/pgcli/blob/master/pgcli/packages/prioritization.py#L11,
> ignore the fact that the re.sub() call seem underoptimal).
>
> With Python 3.5.2 is straightforward:
>
>   $ python3.5
>   Python 3.5.2+ (default, Sep 22 2016, 12:18:14)
>   [GCC 6.2.0 20160927] on linux
>   Type "help", "copyright", "credits" or "license" for more information.
>   >>> import re
>   >>> re.sub(r'\s+', r'\s+', 'foo bar')
>   'foo\\s+bar'
>
> While Python 3.6.0b2 gives:
>
>   $ python3.6
>   Python 3.6.0b2+ (default, Oct 11 2016, 08:30:05)
>   [GCC 6.2.0 20160927] on linux
>   Type "help", "copyright", "credits" or "license" for more information.
>   >>> import re
>   >>> re.sub(r'\s+', r'\s+', 'foo bar')
>   Traceback (most recent call last):
>     File "/usr/local/python3.6/lib/python3.6/sre_parse.py", line 945, in parse_template
>       this = chr(ESCAPES[this][1])
>   KeyError: '\\s'
>
>   During handling of the above exception, another exception occurred:
>
>   Traceback (most recent call last):
>     File "<stdin>", line 1, in <module>
>     File "/usr/local/python3.6/lib/python3.6/re.py", line 191, in sub
>       return _compile(pattern, flags).sub(repl, string, count)
>     File "/usr/local/python3.6/lib/python3.6/re.py", line 326, in _subx
>       template = _compile_repl(template, pattern)
>     File "/usr/local/python3.6/lib/python3.6/re.py", line 317, in _compile_repl
>       return sre_parse.parse_template(repl, pattern)
>     File "/usr/local/python3.6/lib/python3.6/sre_parse.py", line 948, in parse_template
>       raise s.error('bad escape %s' % this, len(this))
>   sre_constants.error: bad escape \s at position 0
>
> Accordingly to the documentation (https://docs.python.org/3.6/library/re.html#re.sub)
> “unknown escapes [in the repl argument] such as \& are left alone”.
>
> Am I missing something, or is this a regression?

Unknown escapes consisting of "\" following by ASCII letter are errors 
in 3.6 (and warnings in 3.5). Seems the documentation is not accurate. 
Could you file a report on https://bugs.python.org/ ?