add impl-ordering-except-fn-names config field for arbitrary_source_item_ordering

[relaxcn]

I think it makes sense to set up a list of function names to exclude certain functions in the impl block from being checked.

Some functions, such as new, are typically used as constructors—either as the first item or in other positions—so we need a configuration to exclude them.

I added a config filed impl-ordering-except-fn-names(Vec<String>) to define the list of function names that should be excluded.

This modification does not change the existing behavior of Clippy; it only takes effect after being explicitly configured.

changelog: [arbitrary_source_item_ordering]: add impl-ordering-except-fn-names config field to exclude some functions in impl blocks.

Close: #14739

[rustbot]

r? @Jarcho

rustbot has assigned @Jarcho.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[Jarcho]

The config for this lint is currently quite awkward and in need of being redone and I'd rather not add more to it until that's done. I have a few ideas as to how I would change it to make it easier to use and I'll see if I can get to it soon.

If I don't have anything within a week or so can you ping me and I can merge this first instead?

[Jarcho]

This happens too early as type sorting hasn't happened yet.

You'll have to move the kind order check on line 353 to happen before the name check at line 351 and then insert this in between.

[relaxcn]

This change is only to skip the function names that we don't want to sort.
I don't want to care about if it is one of the two compared objects, I just skip it.
If I change the code to this:

                for item in trait_impl.items {
                    if item.span.in_external_macro(cx.sess().source_map()) {
                        continue;
                    }
                    if let Some(cur_t) = cur_t {
                        let item_kind = convert_assoc_item_kind(item.kind);
                        let item_kind_index = self.assoc_types_order.index_of(&item_kind);

                        let cur_t_kind = convert_assoc_item_kind(cur_t.kind);
                        let cur_t_kind_index = self.assoc_types_order.index_of(&cur_t_kind);

                        // If there are exception function names, skip it.
                        if !self.impl_ordering_except_fn_names.is_empty()
                            && convert_assoc_item_kind(item.kind) == SourceItemOrderingTraitAssocItemKind::Fn
                            && self
                                .impl_ordering_except_fn_names
                                .contains(&item.ident.name.to_string())
                        {
                            continue;
                        }
                        
                        if cur_t_kind == item_kind && cur_t.ident.name.as_str() > item.ident.name.as_str() {
                            Self::lint_member_name(cx, &item.ident, &cur_t.ident);
                        } else if cur_t_kind_index > item_kind_index {
                            self.lint_impl_item(cx, item, cur_t);
                        }
                    }
                    cur_t = Some(item);
                }

Then you will find a uitest error when you run command TESTNAME="ordering_exceptions" cargo uitest.

impl BasicStruct1 {
    fn new() -> Self {
        Self {}
    }
    fn a() {}
    fn b() {}
}
error: there were 1 unmatched diagnostics
  --> tests/ui-toml/arbitrary_source_item_ordering/ordering_exceptions.rs:14:8
   |
14 |     fn a() {}
   |        ^ Error[clippy::arbitrary_source_item_ordering]: incorrect ordering of items (must be alphabetically ordered)
   |

The first function will always participate in the sorting.

[relaxcn]

The config for this lint is currently quite awkward and in need of being redone and I'd rather not add more to it until that's done. I have a few ideas as to how I would change it to make it easier to use and I'll see if I can get to it soon.

If I don't have anything within a week or so can you ping me and I can merge this first instead?

Ok, you are right.

[relaxcn]

Just a kind ping @Jarcho.

I think you are right. If you are ready to rewrite this lint, I could close this PR.