Convention around fn new() vs. impl Default

[killercup]

C-COMMON-TRAITS wants me to impl Default. The first example in C-CTOR is fn new() -> Example<T>.

It would be nice to have a convention around when a new constructor with no parameters makes sense, when we should just use Default::default, and when new should actually use default (or some other variations).

Previous discussions about this:

• new idiom: Default trait rust-unofficial/patterns#48
• https://deterministic.space/elegant-apis-in-rust.html#fn:new
• killercup/scribbles@c5863eb#commitcomment-19523203
• killercup/scribbles@628f531#commitcomment-19527945

cc @llogiq

[brson]

Clarify that if you can have a 'new' ctor it should exist, and do the same as 'default'. Add to C-CTOR.

[llogiq]

Note that a new ctor is easier to find in the docs (and to document specifics for) than a Default impl. So even if the latter exists, it makes sense to provide the former.

[killercup]

Note that a new ctor is easier to find in the docs

Will put this on my 'Rustdoc2 features wish list'! (It's getting longer every day.)

…

Am 01.07.2017 um 15:31 schrieb llogiq ***@***.***>: Note that a new ctor is easier to find in the docs (and to document specifics for) than a Default impl. So even if the latter exists, it makes sense to provide the former. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.