Is it possible to get keys in the same order?

[xphoniex]

File 1:

version: 1
config:
  name: app

File 2:

version: 999

Result:

config:
  name: app
version: 999

whereas I expected it to keep the order.

[qtfkwk]

As you can tell, this crate hasn't had any attention in some time. Sorry about that. Working on it a bit now.

Was able to reproduce via the following failing slightly modified copies of the merge_multiple_string_string_with_conflict test:

#[test]
fn merge_multiple_string_string_with_conflict_2() {
    let mut hash = MergeYamlHash::new();
    let yaml1 = "apple: 1\nbanana: 2".to_string();
    let yaml2 = "apple: 3".to_string();
    let result = "apple: 3\nbanana: 2";
    hash.merge_vec(vec![yaml1, yaml2]);
    assert_eq!(hash.to_string(), result);
}

---- tests::merge_multiple_string_string_with_conflict_2 stdout ----
thread 'tests::merge_multiple_string_string_with_conflict_2' panicked at src/tests.rs:69:5:
assertion `left == right` failed
  left: "banana: 2\napple: 3"
 right: "apple: 3\nbanana: 2"

#[test]
fn merge_multiple_string_string_with_conflict_3() {
    let mut hash = MergeYamlHash::new();
    let yaml1 = "apple: 1\nbanana: 2\ncherry: 3".to_string();
    let yaml2 = "banana: 4".to_string();
    let result = "apple: 1\nbanana: 4\ncherry: 3";
    hash.merge_vec(vec![yaml1, yaml2]);
    assert_eq!(hash.to_string(), result);
}

---- tests::merge_multiple_string_string_with_conflict_3 stdout ----
thread 'tests::merge_multiple_string_string_with_conflict_3' panicked at src/tests.rs:79:5:
assertion `left == right` failed
  left: "apple: 1\ncherry: 3\nbanana: 4"
 right: "apple: 1\nbanana: 4\ncherry: 3"

It seems the readme is right when it claims that "insertion order is maintained" only when there is no conflict or the conflict in file2 is the last item in file1, which is to say it's mostly wrong actually.

This crate as of version 0.3.0 has been upgraded to edition 2021, dependencies updated, code/comments/etc revamped, the merge_hashes method has been moved to a function due to clippy's only_used_in_recursion lint, and its Yaml::Hash.insert approach has been modified to use the entry().and_modify().or_insert_with() pattern which should resolve this issue.

@xphoniex If you're still around and able to test this and confirm it resolves, please do and lmk!