Re: [PATCH] Add pg_get_subscription_ddl() function

Hi Alvaro,

Thanks for your input.

On Thu, Nov 6, 2025 at 9:18 PM Álvaro Herrera <alvherre(at)kurilemu(dot)de> wrote:

> Hello Vaibhav,
>
> I wonder why is Subscription->publications a list of String rather than
> a list of C strings. That's something you'd see in a Node structure,
> but Subscription is not a node, so this seems wasteful and pointless.
>

I looked more into this and came to know that we can't make
Subscription->publications a list of C strings because input publications
list is also in the list of String from the parser:

CreateSubscriptionStmt:
CREATE SUBSCRIPTION name CONNECTION Sconst PUBLICATION name_list
opt_definition
{
CreateSubscriptionStmt *n =
makeNode(CreateSubscriptionStmt);
n->subname = $3;
n->conninfo = $5;
n->publication = $7;
n->options = $8;
$$ = (Node *) n;
};

name_list: name
{ $$ = list_make1(makeString($1)); }
| name_list ',' name
{ $$ = lappend($1, makeString($3)); };

Oh, we also have textarray_to_strvaluelist() which is essentially
> identical, but also static. If we're making one of them non-static,
> then for sure let's remove the other one. But maybe what we really need
> is a third one to use in ruleutils, and expose neither? (I think if we
> get rid of the String around Subscription->publications, that's likely
> what I'd do, since they'd be mostly trivial wrappers around
> deconstruct_array_builtin.)
>

I think we really need a third one to use in ruleutils, and expose neither.
Find the attached v3 patch which does the same.