Re: a few question about named reference

[Bart Schaefer <schaefer@brasslantern.com>]

On Wed, Jul 12, 2023 at 9:06 PM Jun T <takimoto-j@kba.biglobe.ne.jp> wrote:
>
> [1] what does 'typeset -n ref' do when a variable 'ref' already exists?

As of workers/51402, it does the same thing as
  ref=xx
  typeset -n ref

So ...

> (1a) removing a named reference
>
>        ... so  to  remove  a
>        named reference, use either `unset -n pname' or one of:
>               typeset -n pname
>               typeset +n pname
>        followed by
>               unset pname
>
> But:
> % typeset -n pname     # this does nothing
> % typeset -p pname
> typeset -n pname=i     # still points to i
[...]
>
> Just fix the document?

Yep, documentation dates from the previous patches.  Need to update.

> (1b) converting a scalar into named reference
>
> In the first post by Bart about the named reference (worker/51360⁩):
>
> > One difference from ksh93 is that you can't convert scalars into
> > namerefs
>
> But it seems we can convert a scalar into nameref:

That also changed in 51402, from the commit log:

    * Ksh-style "foo=bar; typeset -n foo" creates foo=bar reference

> [2] readonly named reference
>
> So we can't remove the readonly attribute (and will not be able
> to remove/unset it). Is this intentional?

Yes, that's how it works in ksh.

> [3] can named reference be in 'unset' state?
>
> It seems resolve_nameref() in params.c resets the PM_UNSET bit.
> I feel it would be simpler to ignore typesettounset when creating
> a nameref.

That will need to be looked at.  Thanks for noticing.