Re: Add IS_INDEX macro to brin and gist index

On 1/14/26 2:56 AM, Japin Li wrote:
>
> Hi, hackers,
>
> While working on pageinspect [0], I noticed that brin_page_items() and
> gist_page_items() only checked the access method (IS_BRIN/IS_GIST) but did
> not verify that the passed relation is actually an index relation.
>
> To make the check more robust and consistent with other pageinspect index
> functions (like btreefuncs.c, hashfuncs.c, etc.), the attached patch:
>
> 1. Defines a local helper macro IS_INDEX(r) in both brinfuncs.c and gistfuncs.c.
>
> 2. Updates the error check to require both: the relation must be an index and
> use the expected access method.
>
> The change is very small, low-risk, and only affects two functions in
> contrib/pageinspect.

Since the two functions you touch are gist_page_items() and
brin_page_items() is there actually any harm from being able to use the
index definition from a partitioned when parsing the page? Seems
unnecessary to prevent people from doing so unless I am missing
something. It is not like we have any way to prevent the wrong index
from being used when parsing page the page so why prevent partitioned
indexes specifically?

Andreas