feat: ensure the output order of metric items aligns with their definition order

[niebayes]

What problem does the new feature solve?

Let's take the MergeScanMetric as an example. In the code, we have defined four metric items in the order of ready time, first consume time, finish time, and output rows. However, the display order of these items in the output of the explain analyze SQL query doesn't match the definition order. Ensuring their alignment can enhance the clarity of the explain analyze output in certain cases.

What does the feature do?

Ensure the output order of metric items aligns with their definition order.

Implementation challenges

No response

[NiwakaDev]

I'm interested in this issue, but how do we add some tests for this issue?

I guess that explain analyze tests depends on time.

[niebayes]

I have not investigated the implementation details yet. You are encouraged to try any solution you deem appropriate. Niwaka ***@***.***> 于2023年9月16日周六 13:15写道：

…

I'm interested in this issue, but how do we add some tests for this issue? I guess that explain analyze tests depends on time. — Reply to this email directly, view it on GitHub <#2222 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGBF34PTFKMMSYB4G5J63A3X2UYXFANCNFSM6AAAAAA3ZI3YHQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[NiwakaDev]

@waynexia

This issue happens due to HashMap in the aggregate_by_partition method (datafusion) since elements order in HashMap are non-deterministic:
https://github.com/waynexia/arrow-datafusion/blob/57a66d871fccb58f51588a4f1d30fae8964dc206/datafusion/core/src/physical_plan/metrics/mod.rs#L249-L280

Once I implement a patch, could I submit a PR for this issue to your repository?

[niebayes]

@waynexia PTAL

[waynexia]

Once I implement a patch, could I submit a PR for this issue to your repository?

Please submit to apache/arrow-datafusion. I'll replace my fork with upstream later