On Mon, 12 Aug 2019 at 11:24, Christian Tismer <tismer@stackless.com> wrote:
On 12.08.19 10:52, Ivan Levkivskyi wrote:
> On Thu, 8 Aug 2019 at 17:17, Christian Tismer <tismer@stackless.com
> <mailto:tismer@stackless.com>> wrote:
>
>     Yes, that's what I mean.
>     Probably retval or whatever people prefer would be adequate,
>     with a special rule if that name is taken.
>
>     I think btw. that using StructSequence(somename=sometype, ..., ) that
>     does a dict lookup is quite appealing. It returns a class like
>     stat_result, but the function call shows its arguments (see answer to
>     Ronald).
>
>     Ciao -- Chris
>
>
> Just a little comment: there is a (vague) plan to add a feature (key
> types) to the type system that will allow user-defined constructs
> similar to NamedTuple (for example StructSequence you mention).
> However, taking into account current schedule it is unlikely it will be
> added before mid-2020, so your best bet is indeed using ad-hoc named tuples.

Interesting! Can I read something more about this? I'm curious ;-)

Some preliminary ideas were discussed at one of recent typing summits, see https://paper.dropbox.com/doc/Type-system-improvements--AfxnxPOd_hhYvtamiI9nOdZ9Ag-HHOkniMG9WcCgS0LzXZAe

--
Ivan