On Mon, Jan 23, 2023 at 11:34 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I spent some more time staring at this and came up with what seems like
> a workable patch, based on the idea that what we want to indent is
> specifically initialization expressions. pg_bsd_indent does have some
> understanding of that: ps.block_init is true within such an expression,
> and then ps.block_init_level is the brace nesting depth inside it.
> If you just enable ind_stmt based on block_init then you get a bunch
> of unwanted additional indentation inside struct initializers, but
> it seems to work okay if you restrict it to not happen inside braces.
> More importantly, it doesn't change anything we don't want changed.
Nice! LGTM now that I know about block_init.
