Is it bug or feature in sre?

[Roman Suzi]

On Mon, 5 Jan 2004, Skip Montanaro wrote:

>
>     >>> re.compile("(?P<gr>[A-Z]*)?")
>     Roman> Traceback (most recent call last):
>     ...
>     Roman> sre_constants.error: nothing to repeat
>
>     Roman> I wonder why I can't do this?
>
> Is there some reason why
>
>     "(?P<gr>[A-Z]*)"

- I want it to be optional, so this is not what I wanted

> or
>
>     "(?P<gr>[A-Z]+)?"

- this is right thing

> isn't sufficient?  You specified the zero-or-more repetition inside the
> group.  There's nothing else to have zero-or-one of.
>
>
>     Roman> I guess it's of terrible style, but is this behaviour intentional
>     Roman> or not?
>
> I suspect it's intentional, since Fredrik went to the trouble of raising an
> exception with a useful error message. <wink>

And I fully agree with it.

>     Roman> Probably it is good idea to mention this in Python update path
>     Roman> documents.  Otherwise some old programs will mysteriously fail.
>
> What were you converting from, an older version of sre, pre, regex, or
> (shudder) regexp?

Python 1.5.2
re

> Skip