[gusthoff]

• Return the stored empty value through an expression function.
• Add a runnable demonstration and remove the older GNAT pin.

Co-Authored-By: Codex noreply@openai.com

[swbaird]

Looks good. There is always more one could add - in this case, we could also talk about the case where the Empty function takes a parameter (in addition to talking about the parameterless case). But this change is fine as it stands.

[gusthoff]

Looks good. There is always more one could add - in this case, we could also talk about the case where the Empty function takes a parameter (in addition to talking about the parameterless case). But this change is fine as it stands.

Thanks for the suggestion! I'll put this on my list and work on it separately.

[sttaft]

@gusthoff @swbaird I noticed on line 78 of the ".rst" file that you say that you cannot specify both Add_Named and Add_Unnamed in the Aggregate aspect. However, AI22-0134-1 (Aggregate-named-and-unnamed) changed that, and now allows both to be specified, with Add_Unnamed being used any place where it is legal, and Add_Named used when you give an explicit "use Key => XXX".

I imagine we don't describe this new feature since it is an approved amendment, but one that is not part of Ada 2022 (and not part of the first Ada 2022 amendment being standardized as we speak). How are we handling such approved amendments? Do we mention them for future reference, ignore them completely, or pretend they are already part of the language?