Add BoxCloneSyncService

[tthebst]

resolves #770

[jlizen]

Thanks for cutting this! This is useful in an issue i'm working on for reqwest, to expose the tower stack:
seanmonstar/reqwest#2490 (comment)

Will ping in the tokio/#tower-dev discord to try to help find a reviewer.

[jlizen]

@tthebst do you see a reason we couldn't add BoxCloneSyncServiceLayer similar to BoxCloneServiceLayer?

Seems fine out of box to me:

use crate::boxed_clone_sync::BoxCloneSyncService;
use std::{fmt, sync::Arc};
use tower_layer::{layer_fn, Layer};
use tower_service::Service;

pub struct BoxCloneSyncServiceLayer<In, T, U, E> {
    boxed: Arc<dyn Layer<In, Service = BoxCloneSyncService<T, U, E>> + Send + Sync + 'static>,
}

impl<In, T, U, E> BoxCloneSyncServiceLayer<In, T, U, E> {
    /// Create a new [`BoxCloneServiceLayer`].
    pub fn new<L>(inner_layer: L) -> Self
    where
        L: Layer<In> + Send + Sync + 'static,
        L::Service: Service<T, Response = U, Error = E> + Send + Sync + Clone + 'static,
        <L::Service as Service<T>>::Future: Send + 'static,
    {
        let layer = layer_fn(move |inner: In| {
            let out = inner_layer.layer(inner);
            BoxCloneSyncService::new(out)
        });

        Self {
            boxed: Arc::new(layer),
        }
    }
}

impl<In, T, U, E> Layer<In> for BoxCloneSyncServiceLayer<In, T, U, E> {
    type Service = BoxCloneSyncService<T, U, E>;

    fn layer(&self, inner: In) -> Self::Service {
        self.boxed.layer(inner)
    }
}

impl<In, T, U, E> Clone for BoxCloneSyncServiceLayer<In, T, U, E> {
    fn clone(&self) -> Self {
        Self {
            boxed: Arc::clone(&self.boxed),
        }
    }
}

impl<In, T, U, E> fmt::Debug for BoxCloneSyncServiceLayer<In, T, U, E> {
    fn fmt(&self, fmt: &mut fmt::Formatter) -> fmt::Result {
        fmt.debug_struct("BoxCloneServiceLayer").finish()
    }
}

[seanmonstar]

Excellent work, thank you!

[yyaroshevich]

Should this one also be updated to be considered for merging? #746

[yyaroshevich]

Does it make sense to extend ServiceExt to include helper method to construct new service wrapper? The same seem to be also relevant for an existing UnsyncBoxService wrapper.

[jlizen]

Does it make sense to extend ServiceExt to include helper method to construct new service wrapper? The same seem to be also relevant for an existing UnsyncBoxService wrapper.

Sure, I can cut that tomorrow unless you have time to do it now. Was meaning to but slipped my mind.

Regarding #746 , you could probably ping for a reviewer in the tokio discord's tower-dev channel.

[yyaroshevich]

Sure, I can cut that tomorrow unless you have time to do it now. Was meaning to but slipped my mind.

No rush, just was curious maybe there was a reason no to include.

Regarding #746 , you could probably ping for a reviewer in the tokio discord's tower-dev channel.

I'm not the author, just today wanted this to be presented in tower. I'll need to take closer look at existing MR, it look abandoned from the first glance, so maybe it's already need to be adjusted/re-done.