Re: Should we use MemSet or {0} for struct initialization?

> On Thu, Aug 31, 2023 at 5:34 PM Richard Guo <guofenglinux@gmail.com> wrote:
> >
> > While working on a bug in expandRecordVariable() I noticed that in the
> > switch statement for case RTE_SUBQUERY we initialize struct ParseState
> > with {0} while for case RTE_CTE we do that with MemSet.  I understand
> > that there is nothing wrong with this, just cannot get away with the
> > inconsistency inside the same function (sorry for the nitpicking).
> >
> > Do we have a preference for how to initialize structures?  From 9fd45870
> > it seems that we prefer to {0}.  So here is a trivial patch doing that.

It seems to have been deliberately left that way in the wake of that commit, see:

https://www.postgresql.org/message-id/87d2e5f8-3c37-d185-4bbc-1de163ac4b10%40enterprisedb.com

(If so, it deserves a comment to keep people from trying to change it...)

> > And with a rough scan the MemSet calls in pg_stat_get_backend_subxact()
> > can also be replaced with {0}, so include that in the patch too.

I _believe_ that's harmless to change.

On Thu, Aug 31, 2023 at 4:57 PM Junwang Zhao <zhjwpku@gmail.com> wrote:

> If the struct has padding or aligned, {0} only guarantee the struct
> members initialized to 0, while memset sets the alignment/padding
> to 0 as well, but since we will not access the alignment/padding, so
> they give the same effect.

See above -- if it's used as a hash key, for example, you must clear everything.

> I bet {0} should be faster since there is no function call, but I'm not
> 100% sure ;)

Neither has a function call. MemSet is a PG macro. You're thinking of memset, the libc library function, but a decent compiler can easily turn that into something else for fixed-size inputs.

--
John Naylor
EDB: http://www.enterprisedb.com