Remove parse_vec_expr helper

[findepi]

Behavior of parse_vec_expr and parse_exprs is almost similar -- both take a collection of expressions to parse. The only difference is that parse_vec_expr returns Option::None when collections is empty, but this difference in behavior does not correspond to difference in function names. Since the function is used once only, remove it instead of coming up with a fancy name.

Prep for #12177

[findepi]

this pattern is already used here

datafusion/datafusion/proto/src/logical_plan/mod.rs

Lines 817 to 819 in 117ab1b

	let sort_expr = match distinct_on.sort_expr.len() {
	0 => None,
	_ => Some(

[crepererum]

I wonder if in both places, we could use another new helper like

fn optimize_empty_vec<T>(v: Vec<T>) -> Option<Vec<T>> {
    (!v.empty()).then_some(v)
}

so the combination would look like this:

optimize_empty_vec(parse_exprs(&pb.order_by, registry, codec)?)

However, that's a personal preference, mostly based on the fact that naming the input (pb.order_by) twice, which can easily introduce copy&paste errors.

[findepi]

empty_vec_to_none? vec_if_non_empty?

in this case, however, my preference would be to remove Option and keep just Vec.
Empty collection perfectly describes lack of sorting, so no need to wrap if with optional value.

[crepererum]

Empty collection perfectly describes lack of sorting, so no need to wrap if with optional value.

Let's try to do that then. I agree that having an Option<Vec<_>> in new_udf doesn't make much sense.

[findepi]

good idea. i would rather have this as a follow-up though.
this work would conflict with #12177 considerably.

[findepi]

filed #12195 issue for this

[alamb]

Seems like this is a reasonable change to me and we are tracking futher improvements. Let's keep the code flowing

Thank you @findepi and @crepererum for the review