Re: pgindent && weirdness

On Wed, Jan 15, 2020 at 11:30 AM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> writes:
> > I just ran pgindent over some patch, and noticed that this hunk ended up
> > in my working tree:
>
> > - if (IsA(leftop, Var) && IsA(rightop, Const))
> > + if (IsA(leftop, Var) &&IsA(rightop, Const))
>
> Yeah, it's been doing that for decades. I think the triggering
> factor is the typedef name (Var, here) preceding the &&.
>
> It'd be nice to fix properly, but I've tended to take the path
> of least resistance by breaking such lines to avoid the ugliness:
>
> if (IsA(leftop, Var) &&
> IsA(rightop, Const))

I am on vacation away from the Internet this week but somehow saw this
on my phone and couldn't stop myself from peeking at pg_bsd_ident
again. Yeah, "(Var)" (where Var is a known typename) causes it to
think that any following operator must be unary.

One way to fix that in the cases Alvaro is referring to is to tell
override the setting so that && (and likewise ||) are never considered
to be unary, though I haven't tested this much and there are surely
other ways to achieve this:

diff --git a/lexi.c b/lexi.c
index d43723c..6de3227 100644
--- a/lexi.c
+++ b/lexi.c
@@ -655,6 +655,12 @@ stop_lit:
unary_delim = state->last_u_d;
break;
}
+
+ /* && and || are never unary */
+ if ((token[0] == '&' && *buf_ptr == '&') ||
+ (token[0] == '|' && *buf_ptr == '|'))
+ state->last_u_d = false;
+
while (*(e_token - 1) == *buf_ptr || *buf_ptr == '=') {
/*
* handle ||, &&, etc, and also things as in int *****i

The problem with that is that && sometimes *should* be formatted like
a unary operator: when it's part of the nonstandard GCC computed goto
syntax.