this post was submitted on 23 Sep 2024
0 points (NaN% liked)

Rust Programming

8118 readers
2 users here now

founded 5 years ago
MODERATORS
[email protected]
[email protected]
[email protected]
0
Which code style to initialize structs? (lemmy.ml)
submitted 1 month ago by [email protected] to c/[email protected]
20 comments fedilink hide all child comments
 

Which of these code styles do you find preferable?

First option using mut with constructor in the beginning:

  let mut post_form = PostInsertForm::new(
    data.name.trim().to_string(),
    local_user_view.person.id,
    data.community_id,
  );
  post_form.url = url.map(Into::into);
  post_form.body = body;
  post_form.alt_text = data.alt_text.clone();
  post_form.nsfw = data.nsfw;
  post_form.language_id = language_id;

Second option without mut and constructor at the end:

  let post_form = PostInsertForm {
    url: url.map(Into::into),
    body,
    alt_text: data.alt_text.clone(),
    nsfw: data.nsfw,
    language_id,
    ..PostInsertForm::new(
      data.name.trim().to_string(),
      local_user_view.person.id,
      data.community_id,
    )
  };

You can see the full PR here: https://github.com/LemmyNet/lemmy/pull/5037/files

top 20 comments
sorted by: hot top controversial new old
[–] [email protected] 0 points 1 month ago* (last edited 1 month ago) (1 children)

I prefer to encapsulate a mutable reference to the instance in a scope.

let post_form = {
    let mut post_form = PostInsertForm::new(
        // your constructor arguments
    );
    post_form.some_mutating_method(
        // mutation arguments
    );
    post_form
};

This way you're left with an immutable instance and you encapsulate all of the logic needed to setup the instance in one place.

[–] [email protected] 0 points 1 month ago

@livingcoder @nutomic that's a nice one. Had never thought of it. But I'd just use the builder pattern.

[–] [email protected] 0 points 1 month ago* (last edited 1 month ago) (1 children)

Second one if a constructor or a builder is not an option. 1 is out of the question.

Why are the Lemmy devs asking for this though?

[–] [email protected] 0 points 1 month ago

To decide if I should merge the linked PR or not (I did merge it).

[–] [email protected] 0 points 1 month ago

Neither.

  • make new() give you a fully valid and usable struct value.
  • or use a builder (you can call it something else like Partial/Incomplete/whatever) struct so you can't accidentally do anything without a fully initialized value.

Maybe you should also use substructs that hold some of the info.

[–] [email protected] 0 points 1 month ago

also adding my vote for the second one

[–] [email protected] 0 points 1 month ago (1 children)

@[email protected] @[email protected] @[email protected] Thanks for the feedback! Personally I prefer the first option, but based on your comments I will merge the PR with the second option.

[–] [email protected] 0 points 1 month ago (2 children)

If you're ever forced to do something the second way, you can also wrap it in braces, that way you end up with an immutable value again:

let app = {
  let mut app = ...
  ...
  app
};
[–] [email protected] 0 points 1 month ago (1 children)

Why not just a let app = app; line after the let mut app = ...; one?

[–] [email protected] 0 points 1 month ago (1 children)

A scope groups the initialization visually together, while adding the let app = app; feels like it just adds clutter - I'd probably just leave it mut in that case.

[–] [email protected] 0 points 1 month ago

Rebinding with and without mut is a known and encouraged pattern in rust. Leaving things as mut longer than necessary is not.

[–] [email protected] 0 points 1 month ago (1 children)

Thats even more verbose so the second option is better.

[–] [email protected] 0 points 1 month ago (1 children)

Yeah if you have the second option, use it, but if the struct has private fields it won't work.

[–] [email protected] 0 points 1 month ago (1 children)

The first one won't work either for private fields.

[–] [email protected] 0 points 1 month ago (1 children)

You can have setters that set private fields, there are also sometimes structs with mixed private and public fields

[–] [email protected] 0 points 1 month ago (1 children)

But why not use a proper builder pattern in that case?

[–] [email protected] 0 points 1 month ago

Because you don't control third party libraries

[–] [email protected] 0 points 1 month ago

Defo the second one, the first is weird imo

[–] [email protected] 0 points 1 month ago

100% the second one. It's the idiomatic way to do this in Rust, and it leaves you with an immutable object.

I personally like to move the short declarations together (i.e. body down with language_id (or both at the top)) but that's a minor quibble.

[–] [email protected] 0 points 1 month ago

Definitely the second one.

  1. It avoids Mut
  2. It makes clear that the initialization is over at the end of of the statement. The first option invites people to change some more properties hundreds of lines down where you won't see them.