=E2=80=8BWorking as documented.=E2=80=8B
https://www.postgresql.org/docs/9.5/static/functions-matching.html#POSIX-MA=
TCHING-RULES
Specifically, this implementation considers greediness at a level higher
than just the atom/expression - and in a mixed "branch" if there is a
non-greedy quantifier in a branch the entire branch is non-greedy and can
in many situations cause greedy atoms to behave non-greedily.
In might help to consider that there aren't really any explicit "greedy"
operators like other engines have (i.e., ??, ?, ?+) but rather non-greedy
(lazy) and default. The default inherits the non-greedy trait from its
parent if applicable otherwise is behaves greedily.
On Mon, Aug 15, 2016 at 7:53 AM, Foster, Russell <Russell.Foster@crl.com>
wrote:
> Hello,
>
>
>
> For the following query:
>
>
>
> select substring('>772' from '.*?[0-9]+')
>
=E2=80=8BThe pattern itself is non-greedy=E2=80=8B due to their only being =
a single branch
and it having a non-greedy quantifier within it.
.*? matches ">" and [0-9]+ only needs a single character to generate a
non-greedy match conforming match
>
> I would expect the output to be =E2=80=98>772=E2=80=99, but it is =E2=80=
=98>7=E2=80=99. You can also see
> the expected result on https://regex101.com/, although I am aware not all
> regex processors work the same.
>
>
>
> The following queries:
>
>
>
> select substring('>772' from '^.*?[0-9]+$')
>
=E2=80=8BThis is treated exactly the same as the above but because of the ^=
$ the
shortest possible output string is the entire string=E2=80=8B
>
> and:
>
>
>
> select substring('>772' from '[0-9]+')
>
>
>
> both return =E2=80=98>772=E2=80=99, which is expected. Could the less gr=
eedy operator on
> the left (.*?) be affecting the more greedy right one (+)?
>
>
>
Typo here? I'm not fluent with substring(regex).
Anyway, the entire RE (single branch) is now greedy so the greedy [0-9]+
atom matches as many numbers as possible.
David J.