Re: Switch to multi-inserts for pg_depend

I agree, this version looks much better, thanks.  Two very minor things:

On 2020-Sep-03, Michael Paquier wrote:

> @@ -76,11 +77,23 @@ recordMultipleDependencies(const ObjectAddress *depender,
>  
>  	dependDesc = table_open(DependRelationId, RowExclusiveLock);
>  
> +	/*
> +	 * Allocate the slots to use, but delay initialization until we know that
> +	 * they will be used.  The slot initialization is the costly part, and the
> +	 * exact number of dependencies inserted cannot be known in advance as it
> +	 * depends on what is pinned by the system.
> +	 */

I'm not sure you need the second sentence in this comment; keeping the
"delay initialization until ..." part seems sufficient.  If you really
want to highlight that initialization is costly, maybe just say "delay
costly initialization".

> +		/*
> +		 * Record the Dependency.  Note we don't bother to check for duplicate
> +		 * dependencies; there's no harm in them.
> +		 */

No need to uppercase "dependency".  (I know this is carried forward from
prior comment, but it was equally unnecessary there.)

>  	/*
>  	 * Allocate the slots to use, but delay initialization until we know that
> -	 * they will be used.
> +	 * they will be used.  A full scan of pg_shdepend is done to find all the
> +	 * dependencies from the template database to copy.  Their number is not
> +	 * known in advance and the slot initialization is the costly part.
>  	 */

As above, this change is not needed.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services