Consider taking writers by value #20
Comments
HadrienG2
changed the title
|
This seems very plausible! I'll check it out. |
|
Investigation suggests that while this is fine for This seems like it would confound structural recursion through tree-structured data, e.g. struct Tree {
children: Vec<Tree>,
} |
|
Let's not do this then, as I said it's not a big issue ;) |
|
Wait. Given that &mut W implements Write since W implements Write, what prevents you from sending around an &mut writer while traversing a tree ? |
|
I think that is what happens now, isn't it? We could allow |
|
To clarify, I meant that because &mut W implements Write just like owned writers, this code is legal: Therefore, there is no problem with accepting "owned" writers, and using them multiple times. |
|
Yup, I agree. However, pub fn takes_write_impl(mut w: impl std::io::Write) {
takes_write_impl(&mut w)
}
fn main() {
let mut bytes = Vec::new();
takes_write_impl(&mut bytes);
}trips a recursion limit This is the problem I think exists, as we seem to need to use |
|
Oh, I see. This could cause problems with deep object trees indeed, especially as it also implies that the compiler is generating O(N) copies of take_write_impl when using this pattern, which is a compilation time liability. Then I agree with you: |
&mut TwhereT: WriteimplementsWrite, so there is no reason to ask for a borrowed writer in the API. But this is unlikely to be a problem in practice because owned writers are rarely used.The text was updated successfully, but these errors were encountered: