Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Add an Actor custom derive macro #163

Merged
merged 11 commits into from
Aug 12, 2022
Merged

Add an Actor custom derive macro #163

merged 11 commits into from
Aug 12, 2022

Conversation

thomaseizinger
Copy link
Collaborator

Fixes #127.

@thomaseizinger thomaseizinger changed the title Add an Actor custom derive Add an Actor custom derive macro Aug 6, 2022
Restioson
Restioson reviewed Aug 10, 2022
View reviewed changes
Restioson
Restioson reviewed Aug 10, 2022
View reviewed changes
Restioson
Restioson approved these changes Aug 10, 2022
View reviewed changes
Copy link
Owner

@Restioson Restioson left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I like this change in concept, but I do wonder if this is the most optimal directory structure/layout. I see that https://github.com/tokio-rs/tokio has a separate directory for the main tokio crate, whereas this has the crate root as .. This also doesn't use a workspace, I think.

@thomaseizinger
Copy link
Collaborator Author

I like this change in concept, but I do wonder if this is the most optimal directory structure/layout. I see that https://github.com/tokio-rs/tokio has a separate directory for the main tokio crate, whereas this has the crate root as .. This also doesn't use a workspace, I think.

We do use a workspace here, it is defined in the main crate!

I don't care much about the structure but if we want to change it, I'd suggest a different PR because it will be quite the diff noise :)

thomaseizinger
thomaseizinger commented Aug 10, 2022
View reviewed changes
@Restioson
Copy link
Owner

We do use a workspace here, it is defined in the main crate!

Oh, I hadn't seen that!

@Restioson Restioson merged commit 1a859b6 into master Aug 12, 2022
@thomaseizinger thomaseizinger mentioned this pull request Aug 16, 2022
Closed
thomaseizinger added a commit that referenced this pull request Aug 31, 2022
@thomaseizinger
Squashed commit of the following:
5721ef9 
commit 9cb6544
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 31 20:09:45 2022 +0200

    Simplify visibility

    Mailbox is not yet a public type so we don't need to say pub(super).

commit f7debd3
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 31 20:09:27 2022 +0200

    Merge impl blocks

commit ed9f4d1
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 31 20:06:53 2022 +0200

    Squashed commit of the following:

    commit a0f1dfe
    Merge: d63a549 100195d
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 30 16:45:39 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit d63a549
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 30 15:58:53 2022 +0200

        Fix invalid doc link

    commit 0eca513
    Merge: 1ea9670 89ac7bb
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 30 15:57:20 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit 1ea9670
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 30 15:54:04 2022 +0200

        Express conversion to `Either` via `Into` bound

        This avoids an unreachable clause where `Rc` could not be converted
        into an `Either` type.

    commit 128e655
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Wed Aug 17 17:54:15 2022 +0200

        Make impl resilient against moves

    commit 4c87c84
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Wed Aug 17 17:52:13 2022 +0200

        Hide impl details of pointer debug output

    commit 6ddb280
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Wed Aug 17 17:45:29 2022 +0200

        Fix build error

    commit 36921d4
    Merge: 4449c33 f169e6d
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Wed Aug 17 17:44:53 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit 4449c33
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 16 11:44:13 2022 +0200

        Undo renaming

        We are calling this on a generic Rc but the return type is not generic
        enough.

    commit 2bd2be1
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 16 11:40:38 2022 +0200

        Reduce diff to master

    commit 40fd855
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 16 11:29:27 2022 +0200

        Introduce `RefCounter::into_either`

        This allows us to access `Address::to_either` when abstracting over
        strong/weak reference counts.

    commit 5cd6c29
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Mon Aug 15 22:03:59 2022 +0200

        Seal `RefCounter` trait

        This allows us to export it without being worried about users
        depending on its actual design.

    commit 5547c9f
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Sat Aug 6 00:17:30 2022 +0200

        Add more docs

    commit 73ef566
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Sat Aug 6 00:05:50 2022 +0200

        Fix docs

    commit db08146
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Sat Aug 6 00:00:10 2022 +0200

        Promote `inbox::rx` to `Mailbox` type

        This type is a counter-part to `Address` and will be public at some
        point in the future.

    commit 9ed521b
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:56:46 2022 +0200

        Use `ChanPtr` inside `ReceiveFuture`

    commit 47d6092
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:42:41 2022 +0200

        Re-order functions to be in meaningful order

    commit f6e8365
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:36:00 2022 +0200

        Use short initializer

    commit dcea698
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:34:29 2022 +0200

        Add docs to `RefCounter` trait

    commit 3b4725e
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:27:11 2022 +0200

        Encapsulate reference counting policy logic on `Chan`

    commit ed46b76
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 22:59:37 2022 +0200

        Add docs

    commit 517fd16
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 22:57:12 2022 +0200

        Don't export `RefCounter`

        Not exporting it allows us to avoid sealing the trait because if
        users cannot name the type, the cannot call functions or implement
        it on types.

    commit 45941be
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 22:47:47 2022 +0200

        Go back to `Rc` naming for type parameter

    commit 08242a5
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 22:42:49 2022 +0200

        Remove double increment

    commit e7252c3
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 17:47:18 2022 +0200

        Rename `RefCountPolicy` to `RefCounter`

        The old name was good enough and we don't need to cause unnecessary
        diff.

    commit d29c37b
    Merge: e1ae74e 17cfd3a
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 17:46:20 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit e1ae74e
    Merge: 7e275b3 b8cb4c1
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 17:36:10 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit 7e275b3
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 23:33:23 2022 +0200

        Add test for width of pointer

    commit 78f5e4f
    Merge: faa48f8 8a85729
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 23:26:11 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit faa48f8
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 23:09:46 2022 +0200

        Remove `Sender`

        This is an unnecessary layer now.

    commit 4cb448d
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 23:04:07 2022 +0200

        Implement fmt::Debug for ChanPtr

    commit b653f8d
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 22:56:20 2022 +0200

        Make `chan_ptr` module private

    commit 862e9fa
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 20:31:19 2022 +0200

        Fix clippy

    commit fbf5e5a
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 20:15:19 2022 +0200

        Temporarily fix tests

    commit 8781f15
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Mon Jul 25 12:05:20 2022 +0200

        PoC for generic ref counter

commit 1423520
Merge: d33b741 100195d
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 16:50:05 2022 +0200

    Merge branch 'master' into unify-recv-future

commit 100195d
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 16:42:31 2022 +0200

    Simplify conditional code around instrumentation feature (#172)

    By splitting things into modules with identical APIs, we can remove a lot of `cfg(feature)` attributes.

commit d33b741
Merge: 158a1c8 89ac7bb
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 16:09:11 2022 +0200

    Merge branch 'master' into unify-recv-future

commit 89ac7bb
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 15:46:06 2022 +0200

    Differentiate single and broadcast message on type-system level (#166)

    Prior to this patch-set, we had constructs like `MessageType` and `SentMessage`. These were used to write code that is generic over a particular message type. In reality however, we don't actually need to differentiate between these cases because the codepaths for each type are statically known in all cases.

    Unfortunately, introducing an actual split introduced a bit more code (net ~ 70 lines) but in exchange, we remove a few `unreachable` error clauses and make it overall easier to follow the dataflow through xtra's channel implementation.

commit e6941e2
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 13:53:08 2022 +0200

    Update feature list (#178)

commit bcfd20c
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Thu Aug 18 02:15:33 2022 +1000

    Remove irrelevant changelog entry

commit f169e6d
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 17 10:51:29 2022 +0200

    Fix `scoped_actor_task` example (#175)

    Because of an unknown feature, this example was not compiled as
    part of our CI run and we missed it while moving to the new spawn
    API.

commit 358480c
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 17 10:51:14 2022 +0200

    Remove `global_spawner_ext` example (#177)

    This is the same as the `basic_smol` example.

commit cf613fa
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 17 10:33:09 2022 +0200

    Enforce grouping of imports per module (#174)

    Most of our codebase seems to be formatted this way, we can enforce
    this with a rustfmt setting.

commit 41bc8db
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 17 10:32:36 2022 +0200

    Simplify example in README (#173)

commit aa923d0
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 16 13:32:59 2022 +0200

    Add test for public API of `Address<A, Rc>` (#171)

commit 4940e9a
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 16 11:28:45 2022 +0200

    Add `public_api` test (#169)

commit 1a859b6
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Aug 12 10:54:05 2022 +0200

    Add an `Actor` custom derive macro (#163)

    * Add basic implementation of `Actor` custom derive

    This is located within in the xtra-macros subcrate. The top level crate also is now a workspace.

commit 4d4d4c1
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 10 14:34:17 2022 +0200

    Refactor instrumentation tests (#165)

    * Re-order elements in `instrumentation` test by priority

    1. Tests come first
    2. Actors second
    3. Test infrastructure last

    * Remove unnecessary `pub` qualifier

    * Rename `MockWriter` to `BufferWriter`

    Given that we write to a buffer, this is a more appropiate name.

    * Introduce `Buffer` to allow for direct comparison with `[&str, N]`

    * Construct `BufferWriter` in `get_subscriber`

    This reduced some duplication in the tests.

commit 6ca3382
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 10 14:19:28 2022 +0200

    Simplify control-flow in `ChanInner::pop_` functions (#167)

    Previously, we first checked whether the corresponding queue WILL
    have space if we now pop a message.

    This is functionally equivalent to _trying_ to pop a message, exiting
    early if we can't (using `?`) and then adding another message to the
    queue if we now have space.

    Checking for space is equivalent to the queue not being full and we
    already have helper functions for this.

commit 17cfd3a
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Aug 5 13:01:58 2022 +0200

    Have `Sender` deref to `Chan` (#161)

    Essentially, `Sender` is a smart-pointer to `Chan` with custom reference
    counting logic and thus it makes sense for it to implement `Deref`.
    Most of the functionality accessed by other modules lives on `Chan`.

commit 158a1c8
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 20:27:18 2022 +0200

    Don't implement `Deref` on `Receiver`

    If we have to wrap it to provide functionality, it is not a
    smart-pointer.

commit 3dd7075
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 20:11:58 2022 +0200

    Have `Receiver` have a public API again

commit aad40ff
Merge: 422956f b8cb4c1
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 19:58:52 2022 +0200

    Merge branch 'master' into unify-recv-future

commit b8cb4c1
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 11:51:50 2022 +0200

    Remove generic ref-counting from `Receiver` (#160)

    * Remove `RxWeak`

    We never have a reference to a weak mailbox of the actor.

    * Remove ref-counter concept from `Receiver`

    We don't need to delegate to anything if there is only one way of
    ref-counting.

    * Inline shutdown of waiting senders into decrement function

    This applies the "Tell; don't ask" principle.

commit f8b61bd
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 11:49:51 2022 +0200

    Remove duplicated code for popping from broadcast channel (#155)

    * Extract utility function for longest broadcast queue

    This brings the `try_advance_broadcast_tail` fn more in line with
    the other `pop_` function in terms of internal structure.

    * Introduce `pop_broadcast` on `ChanInner`

    This creates consistency with `pop_ordered` and `pop_priority`.

    * Move locking of `inner` to top of function

    To achieve similar APIs between the different mailboxes, we lock
    the broadcast mailbox multiple times in the process of dealing
    with it.

    To still avoid race conditions, we need to lock `ChanInner` as
    early as possible because the broadcast mailbox is also only
    modified with `ChanInner` being locked.

commit 422956f
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:22:33 2022 +0200

    Remove references to private futures

    They don't seem to resolve unfortunately :(

commit 360f022
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:15:53 2022 +0200

    Revert to actually mentioning self type

commit 7c1aa3c
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:12:19 2022 +0200

    Inline `next_broadcast_message`

commit d271eae
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:10:37 2022 +0200

    Remove unnecessary `Rc` type parameter

commit aeebf60
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:09:34 2022 +0200

    Remove unnecessary `Rc` argument from `Waiting`

commit b9a338a
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:07:56 2022 +0200

    Add more docs to internal futures

commit 64518d6
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:06:49 2022 +0200

    Align variant name with `Sending` future

commit 727c4ec
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:06:05 2022 +0200

    Move `ReceiveFuture` from `rx` into `recv_future` module

commit dee03f7
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 13:56:26 2022 +0200

    Move priority to dedicated module

commit 452db5e
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 13:38:21 2022 +0200

    Move `ReceiveFuture` into separate top-level module

commit 8a85729
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Jul 26 14:46:15 2022 +0100

    Redesign xtra's spawn API (#156)

    * Redesign xtra's spawn API

    This removes `Actor::create`, `ActorManager` and the various runtime
    extension traits in favor of top-level free functions for spawning
    an actor instance into a specific runtime.

    This is deemed to be more ergonomic because:

    - Users do not need to import an extension trait for their runtime
    - Users do not need to import the `Actor` trait to run it which was
    otherwise required to call `create`.
    - The implementation has less indirection, thus making it easier for
    users to understand and perhaps adapt / write their own `spawn` fn
    depending on their needs.

    The implementation is not strictly functionally equivalent to what
    we had before. In particular, we no longer support spawning in a
    runtime handle of the tokio and smol runtimes. We accept this
    regression because we assume that it is a niche usecase and the now
    easier implementation should showcase users, how they can easily
    write themselves a helper function that spawns an actor using a
    runtime handle in a single line of code. In other words, the `spawn`
    API of xtra is purposely design to make 80% of usecases work. For
    everything else, users should write their own `spawn` function.

    Closes #136.

commit 285b3e9
Author: Restioson <restiosondev@gmail.com>
Date:   Fri Jul 22 16:03:28 2022 +0200

    Use or_current when creating spans (#154)

    * Use or_current when creating spans

    This makes sure that if the xtra_actor_request is disabled, the other spans and child-spans of xtra_message_handler will be children of the would-be parent of xtra_actor_request, rather than being parentless, top-level spans.

    * Add test for or_current

commit 3d063ff
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 22 11:50:53 2022 +0100

    Move capacity from `Chan` to `ChanInner` (#150)

    This allows us to avoid some back and forth passing of data which
    makes a few functions have less arguments and overall better
    ergonomics.

commit 4b5f8af
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Thu Jul 21 23:27:05 2022 +0100

    Encapsulate more functionality in `WaitingReceiver`  (#149)

    * Make `WakeReason` an implementation detail of `WaitingReceiver`

    We can achieve this by having more specific functions on
    `WaitingReceiver` itself.

    * Introduce `FulfillHandle` for `WaitingReceiver`

    Previously, the references to `WaitingReceiver` were stored "naked"
    within the channel and the `Future` implementation. This made it hard
    to understand that there are only ever two references.

    This patch introduces those two references as proper types and adds
    documentation on how the `WaitingReceiver` works.

    * Store cancellation state in boolean flag

    This avoids an "unreachable" branch when polling the receiver.

    * Make `WakeReason` an implementation detail of `WaitingReceiver`

    We can achieve this by making a custom `poll` function on
    `WaitingReceiver` and passing `Receiver` as a reference in.

    * Make `WakeReason` private

    This is possible, now that we made it an implementation detail of
    `WaitingReceiver`.

    * Move `WaitingReceiver` to its own module

    This component houses a good chunk of functionality now and is
    better understood in isolation.

    Additionally, this allows us to better enforce visibility of types.

    * Introduce dedicated `Waiting` sub-state

    This allows for a more focused `Drop` implementation.

    * Use `catty` to implement `WaitingReceiver` (#151)

    Now that the waker management is encapsulated in catty, `WakeReason`
    is no longer an accurate name. `CtrlMsg` seems better.

    Additionally, instead of "fulfilling" the waiting receiver, the
    language of "notifying" feels more natural.

    * Remove dependency on catty patch (#153)

    * Remove dependency on catty patch

    Co-authored-by: Restioson <restiosondev@gmail.com>

    * Update src/inbox.rs

    Co-authored-by: Restioson <restiosondev@gmail.com>

commit eab4691
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Thu Jul 21 22:24:42 2022 +0100

    Update `with-tracing-0_1` feature docs to `instrumentation` feature (#152)

    * Remove mention of outdated feature

    * Mention `instrumentation` feature

commit eafe175
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Jul 19 13:09:14 2022 +0100

    Unify message sending into one `SendFuture` (#147)

    * Have `SendFuture` handle every case of sending messages

    This one future now handles:

    - Sending a single message with the actor type being named
    - Sending a single message with the actor type being erased
    - Sending a broadcast message with the actor type being named

    * Add SendFuture::broadcast_erased

    Using this function takes a few more changes but it is added here
    because it will definitely be useful later.

    * Employ "double result" pattern for `try_send`

    Using a double result is useful in cases where the errors returned
    by a function fall into two categories: recoverable and non-recoverable.

    A disconnected channel is not recoverable and we can thus directly
    forward it with `?`. The mailbox being full is something we need to
    deal with in `SendFuture` which is why it is within another layer
    of `Result`.

commit ff782ad
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 18:52:27 2022 +0100

    Don't pass message name into envelope (#148)

    * Don't pass message name into envelope

    An envelope knows which message it is carrying, no need to pass
    it in.

    * Remove `MessageKind`

    Without the additional field for the message's name, this layer
    does not serve any purpose.

commit 910f3fb
Author: Restioson <restiosondev@gmail.com>
Date:   Mon Jul 18 17:35:52 2022 +0200

    Strip span fields from waiting and handler spans (#145)

    Currently, these are always children of the request span, so adding them as fields in them too is redundant.

    It would be a simplification to move this just to xtra_message_handler, but this is unfortunately not possible in the case of a disconnected actor which will never handle the message - this would miss all actor and message type information.

commit a0c1ad6
Author: Restioson <restiosondev@gmail.com>
Date:   Mon Jul 18 14:05:29 2022 +0200

    Improve span variables in tracing spans (#143)

    * Rename message -> message_type

    Weird behaviour is exhibited by some subscribers with `message` as a field name, probably since it is reserved, so `message_type` is used instead. "actor" was also renamed "actor_type" for consistency.

    Also, the display implementation of type name strings are used to avoid unnecessary quotes in the output.

commit a2bedd5
Author: Restioson <restiosondev@gmail.com>
Date:   Mon Jul 18 13:05:57 2022 +0200

    Create spans late (#120)

    Create the actor request span after first poll of `send` in order to ensure `send(...).instrument(...)` works as expected.

commit 7b0b6ea
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 09:53:35 2022 +0100

    Always store message priority in envelope (#142)

    * Always store message priority in envelope

    Previously, we only stored the priority of broadcast messages in
    the envelope. For regular messages, the priority was stored
    outside in a wrapper struct.

    This disparity is hard to understand as there is no apparent reason
    why the priority could not be stored in the message envelope.

    With this patch, we move the priority into the message envelope
    and also expose a `set_priority` function for each envelope type
    which allows us to consolidate the way we set priority from the
    `SendFuture`. To allow this, message envelopes need to be employ
    interior mutability (they are either box'ed or arc'ed). We use
    a spin lock which should basically be free. A `SendFuture` cannot
    be cloned and thus will never be accessed from two different
    threads, meaning the lock will never be congested.

commit fd3c282
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 09:41:54 2022 +0100

    Remove `ReceiveFuture::cancel` in favor of cancellation-safety docs (#138)

    Remove `ReceiveFuture::cancel` in favor of cancellation-safety docs, which fixes #134. Tests were added to check behaviour of resending messages on drop.

commit 07e4664
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 09:10:48 2022 +0100

    Introduce optional `sink` feature (#140)

    * Introduce optional `sink` feature

    Not every user may need the `sink` feature and we can minimise
    our dependency tree in those cases.

    * Extend README docs

    * Fix docs

commit bc7808c
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 07:07:45 2022 +0100

    Move channel logic into `Chan` (#122)

    Move the channel logic (e.g try_send and try_recv) into Chan itself.

    The goal was is follows:
        - Have Chan house the core channel implementation with descriptive function names for key operations
        - Have Sender and Receiver be thin layers on top that combine Chan with reference counting and provide a Future-based APIs for sending and receiving

commit e54b413
Author: Restioson <restiosondev@gmail.com>
Date:   Sun Jul 17 20:52:20 2022 +0200

    Annotate TickFuture with must_use (#139)

    This is since it is a future
thomaseizinger added a commit that referenced this pull request Aug 31, 2022
@thomaseizinger
Squashed commit of the following:
ae95920 
commit 9cb6544
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 31 20:09:45 2022 +0200

    Simplify visibility

    Mailbox is not yet a public type so we don't need to say pub(super).

commit f7debd3
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 31 20:09:27 2022 +0200

    Merge impl blocks

commit ed9f4d1
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 31 20:06:53 2022 +0200

    Squashed commit of the following:

    commit a0f1dfe
    Merge: d63a549 100195d
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 30 16:45:39 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit d63a549
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 30 15:58:53 2022 +0200

        Fix invalid doc link

    commit 0eca513
    Merge: 1ea9670 89ac7bb
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 30 15:57:20 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit 1ea9670
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 30 15:54:04 2022 +0200

        Express conversion to `Either` via `Into` bound

        This avoids an unreachable clause where `Rc` could not be converted
        into an `Either` type.

    commit 128e655
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Wed Aug 17 17:54:15 2022 +0200

        Make impl resilient against moves

    commit 4c87c84
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Wed Aug 17 17:52:13 2022 +0200

        Hide impl details of pointer debug output

    commit 6ddb280
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Wed Aug 17 17:45:29 2022 +0200

        Fix build error

    commit 36921d4
    Merge: 4449c33 f169e6d
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Wed Aug 17 17:44:53 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit 4449c33
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 16 11:44:13 2022 +0200

        Undo renaming

        We are calling this on a generic Rc but the return type is not generic
        enough.

    commit 2bd2be1
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 16 11:40:38 2022 +0200

        Reduce diff to master

    commit 40fd855
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Tue Aug 16 11:29:27 2022 +0200

        Introduce `RefCounter::into_either`

        This allows us to access `Address::to_either` when abstracting over
        strong/weak reference counts.

    commit 5cd6c29
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Mon Aug 15 22:03:59 2022 +0200

        Seal `RefCounter` trait

        This allows us to export it without being worried about users
        depending on its actual design.

    commit 5547c9f
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Sat Aug 6 00:17:30 2022 +0200

        Add more docs

    commit 73ef566
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Sat Aug 6 00:05:50 2022 +0200

        Fix docs

    commit db08146
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Sat Aug 6 00:00:10 2022 +0200

        Promote `inbox::rx` to `Mailbox` type

        This type is a counter-part to `Address` and will be public at some
        point in the future.

    commit 9ed521b
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:56:46 2022 +0200

        Use `ChanPtr` inside `ReceiveFuture`

    commit 47d6092
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:42:41 2022 +0200

        Re-order functions to be in meaningful order

    commit f6e8365
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:36:00 2022 +0200

        Use short initializer

    commit dcea698
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:34:29 2022 +0200

        Add docs to `RefCounter` trait

    commit 3b4725e
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 23:27:11 2022 +0200

        Encapsulate reference counting policy logic on `Chan`

    commit ed46b76
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 22:59:37 2022 +0200

        Add docs

    commit 517fd16
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 22:57:12 2022 +0200

        Don't export `RefCounter`

        Not exporting it allows us to avoid sealing the trait because if
        users cannot name the type, the cannot call functions or implement
        it on types.

    commit 45941be
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 22:47:47 2022 +0200

        Go back to `Rc` naming for type parameter

    commit 08242a5
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 22:42:49 2022 +0200

        Remove double increment

    commit e7252c3
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 17:47:18 2022 +0200

        Rename `RefCountPolicy` to `RefCounter`

        The old name was good enough and we don't need to cause unnecessary
        diff.

    commit d29c37b
    Merge: e1ae74e 17cfd3a
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 17:46:20 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit e1ae74e
    Merge: 7e275b3 b8cb4c1
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Fri Aug 5 17:36:10 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit 7e275b3
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 23:33:23 2022 +0200

        Add test for width of pointer

    commit 78f5e4f
    Merge: faa48f8 8a85729
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 23:26:11 2022 +0200

        Merge branch 'master' into generic-refcounter

    commit faa48f8
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 23:09:46 2022 +0200

        Remove `Sender`

        This is an unnecessary layer now.

    commit 4cb448d
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 23:04:07 2022 +0200

        Implement fmt::Debug for ChanPtr

    commit b653f8d
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 22:56:20 2022 +0200

        Make `chan_ptr` module private

    commit 862e9fa
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 20:31:19 2022 +0200

        Fix clippy

    commit fbf5e5a
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Thu Jul 28 20:15:19 2022 +0200

        Temporarily fix tests

    commit 8781f15
    Author: Thomas Eizinger <thomas@eizinger.io>
    Date:   Mon Jul 25 12:05:20 2022 +0200

        PoC for generic ref counter

commit 1423520
Merge: d33b741 100195d
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 16:50:05 2022 +0200

    Merge branch 'master' into unify-recv-future

commit 100195d
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 16:42:31 2022 +0200

    Simplify conditional code around instrumentation feature (#172)

    By splitting things into modules with identical APIs, we can remove a lot of `cfg(feature)` attributes.

commit d33b741
Merge: 158a1c8 89ac7bb
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 16:09:11 2022 +0200

    Merge branch 'master' into unify-recv-future

commit 89ac7bb
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 15:46:06 2022 +0200

    Differentiate single and broadcast message on type-system level (#166)

    Prior to this patch-set, we had constructs like `MessageType` and `SentMessage`. These were used to write code that is generic over a particular message type. In reality however, we don't actually need to differentiate between these cases because the codepaths for each type are statically known in all cases.

    Unfortunately, introducing an actual split introduced a bit more code (net ~ 70 lines) but in exchange, we remove a few `unreachable` error clauses and make it overall easier to follow the dataflow through xtra's channel implementation.

commit e6941e2
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 30 13:53:08 2022 +0200

    Update feature list (#178)

commit bcfd20c
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Thu Aug 18 02:15:33 2022 +1000

    Remove irrelevant changelog entry

commit f169e6d
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 17 10:51:29 2022 +0200

    Fix `scoped_actor_task` example (#175)

    Because of an unknown feature, this example was not compiled as
    part of our CI run and we missed it while moving to the new spawn
    API.

commit 358480c
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 17 10:51:14 2022 +0200

    Remove `global_spawner_ext` example (#177)

    This is the same as the `basic_smol` example.

commit cf613fa
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 17 10:33:09 2022 +0200

    Enforce grouping of imports per module (#174)

    Most of our codebase seems to be formatted this way, we can enforce
    this with a rustfmt setting.

commit 41bc8db
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 17 10:32:36 2022 +0200

    Simplify example in README (#173)

commit aa923d0
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 16 13:32:59 2022 +0200

    Add test for public API of `Address<A, Rc>` (#171)

commit 4940e9a
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 16 11:28:45 2022 +0200

    Add `public_api` test (#169)

commit 1a859b6
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Aug 12 10:54:05 2022 +0200

    Add an `Actor` custom derive macro (#163)

    * Add basic implementation of `Actor` custom derive

    This is located within in the xtra-macros subcrate. The top level crate also is now a workspace.

commit 4d4d4c1
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 10 14:34:17 2022 +0200

    Refactor instrumentation tests (#165)

    * Re-order elements in `instrumentation` test by priority

    1. Tests come first
    2. Actors second
    3. Test infrastructure last

    * Remove unnecessary `pub` qualifier

    * Rename `MockWriter` to `BufferWriter`

    Given that we write to a buffer, this is a more appropiate name.

    * Introduce `Buffer` to allow for direct comparison with `[&str, N]`

    * Construct `BufferWriter` in `get_subscriber`

    This reduced some duplication in the tests.

commit 6ca3382
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Wed Aug 10 14:19:28 2022 +0200

    Simplify control-flow in `ChanInner::pop_` functions (#167)

    Previously, we first checked whether the corresponding queue WILL
    have space if we now pop a message.

    This is functionally equivalent to _trying_ to pop a message, exiting
    early if we can't (using `?`) and then adding another message to the
    queue if we now have space.

    Checking for space is equivalent to the queue not being full and we
    already have helper functions for this.

commit 17cfd3a
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Aug 5 13:01:58 2022 +0200

    Have `Sender` deref to `Chan` (#161)

    Essentially, `Sender` is a smart-pointer to `Chan` with custom reference
    counting logic and thus it makes sense for it to implement `Deref`.
    Most of the functionality accessed by other modules lives on `Chan`.

commit 158a1c8
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 20:27:18 2022 +0200

    Don't implement `Deref` on `Receiver`

    If we have to wrap it to provide functionality, it is not a
    smart-pointer.

commit 3dd7075
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 20:11:58 2022 +0200

    Have `Receiver` have a public API again

commit aad40ff
Merge: 422956f b8cb4c1
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 19:58:52 2022 +0200

    Merge branch 'master' into unify-recv-future

commit b8cb4c1
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 11:51:50 2022 +0200

    Remove generic ref-counting from `Receiver` (#160)

    * Remove `RxWeak`

    We never have a reference to a weak mailbox of the actor.

    * Remove ref-counter concept from `Receiver`

    We don't need to delegate to anything if there is only one way of
    ref-counting.

    * Inline shutdown of waiting senders into decrement function

    This applies the "Tell; don't ask" principle.

commit f8b61bd
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Aug 2 11:49:51 2022 +0200

    Remove duplicated code for popping from broadcast channel (#155)

    * Extract utility function for longest broadcast queue

    This brings the `try_advance_broadcast_tail` fn more in line with
    the other `pop_` function in terms of internal structure.

    * Introduce `pop_broadcast` on `ChanInner`

    This creates consistency with `pop_ordered` and `pop_priority`.

    * Move locking of `inner` to top of function

    To achieve similar APIs between the different mailboxes, we lock
    the broadcast mailbox multiple times in the process of dealing
    with it.

    To still avoid race conditions, we need to lock `ChanInner` as
    early as possible because the broadcast mailbox is also only
    modified with `ChanInner` being locked.

commit 422956f
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:22:33 2022 +0200

    Remove references to private futures

    They don't seem to resolve unfortunately :(

commit 360f022
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:15:53 2022 +0200

    Revert to actually mentioning self type

commit 7c1aa3c
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:12:19 2022 +0200

    Inline `next_broadcast_message`

commit d271eae
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:10:37 2022 +0200

    Remove unnecessary `Rc` type parameter

commit aeebf60
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:09:34 2022 +0200

    Remove unnecessary `Rc` argument from `Waiting`

commit b9a338a
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:07:56 2022 +0200

    Add more docs to internal futures

commit 64518d6
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:06:49 2022 +0200

    Align variant name with `Sending` future

commit 727c4ec
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 14:06:05 2022 +0200

    Move `ReceiveFuture` from `rx` into `recv_future` module

commit dee03f7
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 13:56:26 2022 +0200

    Move priority to dedicated module

commit 452db5e
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 29 13:38:21 2022 +0200

    Move `ReceiveFuture` into separate top-level module

commit 8a85729
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Jul 26 14:46:15 2022 +0100

    Redesign xtra's spawn API (#156)

    * Redesign xtra's spawn API

    This removes `Actor::create`, `ActorManager` and the various runtime
    extension traits in favor of top-level free functions for spawning
    an actor instance into a specific runtime.

    This is deemed to be more ergonomic because:

    - Users do not need to import an extension trait for their runtime
    - Users do not need to import the `Actor` trait to run it which was
    otherwise required to call `create`.
    - The implementation has less indirection, thus making it easier for
    users to understand and perhaps adapt / write their own `spawn` fn
    depending on their needs.

    The implementation is not strictly functionally equivalent to what
    we had before. In particular, we no longer support spawning in a
    runtime handle of the tokio and smol runtimes. We accept this
    regression because we assume that it is a niche usecase and the now
    easier implementation should showcase users, how they can easily
    write themselves a helper function that spawns an actor using a
    runtime handle in a single line of code. In other words, the `spawn`
    API of xtra is purposely design to make 80% of usecases work. For
    everything else, users should write their own `spawn` function.

    Closes #136.

commit 285b3e9
Author: Restioson <restiosondev@gmail.com>
Date:   Fri Jul 22 16:03:28 2022 +0200

    Use or_current when creating spans (#154)

    * Use or_current when creating spans

    This makes sure that if the xtra_actor_request is disabled, the other spans and child-spans of xtra_message_handler will be children of the would-be parent of xtra_actor_request, rather than being parentless, top-level spans.

    * Add test for or_current

commit 3d063ff
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Fri Jul 22 11:50:53 2022 +0100

    Move capacity from `Chan` to `ChanInner` (#150)

    This allows us to avoid some back and forth passing of data which
    makes a few functions have less arguments and overall better
    ergonomics.

commit 4b5f8af
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Thu Jul 21 23:27:05 2022 +0100

    Encapsulate more functionality in `WaitingReceiver`  (#149)

    * Make `WakeReason` an implementation detail of `WaitingReceiver`

    We can achieve this by having more specific functions on
    `WaitingReceiver` itself.

    * Introduce `FulfillHandle` for `WaitingReceiver`

    Previously, the references to `WaitingReceiver` were stored "naked"
    within the channel and the `Future` implementation. This made it hard
    to understand that there are only ever two references.

    This patch introduces those two references as proper types and adds
    documentation on how the `WaitingReceiver` works.

    * Store cancellation state in boolean flag

    This avoids an "unreachable" branch when polling the receiver.

    * Make `WakeReason` an implementation detail of `WaitingReceiver`

    We can achieve this by making a custom `poll` function on
    `WaitingReceiver` and passing `Receiver` as a reference in.

    * Make `WakeReason` private

    This is possible, now that we made it an implementation detail of
    `WaitingReceiver`.

    * Move `WaitingReceiver` to its own module

    This component houses a good chunk of functionality now and is
    better understood in isolation.

    Additionally, this allows us to better enforce visibility of types.

    * Introduce dedicated `Waiting` sub-state

    This allows for a more focused `Drop` implementation.

    * Use `catty` to implement `WaitingReceiver` (#151)

    Now that the waker management is encapsulated in catty, `WakeReason`
    is no longer an accurate name. `CtrlMsg` seems better.

    Additionally, instead of "fulfilling" the waiting receiver, the
    language of "notifying" feels more natural.

    * Remove dependency on catty patch (#153)

    * Remove dependency on catty patch

    Co-authored-by: Restioson <restiosondev@gmail.com>

    * Update src/inbox.rs

    Co-authored-by: Restioson <restiosondev@gmail.com>

commit eab4691
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Thu Jul 21 22:24:42 2022 +0100

    Update `with-tracing-0_1` feature docs to `instrumentation` feature (#152)

    * Remove mention of outdated feature

    * Mention `instrumentation` feature

commit eafe175
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Tue Jul 19 13:09:14 2022 +0100

    Unify message sending into one `SendFuture` (#147)

    * Have `SendFuture` handle every case of sending messages

    This one future now handles:

    - Sending a single message with the actor type being named
    - Sending a single message with the actor type being erased
    - Sending a broadcast message with the actor type being named

    * Add SendFuture::broadcast_erased

    Using this function takes a few more changes but it is added here
    because it will definitely be useful later.

    * Employ "double result" pattern for `try_send`

    Using a double result is useful in cases where the errors returned
    by a function fall into two categories: recoverable and non-recoverable.

    A disconnected channel is not recoverable and we can thus directly
    forward it with `?`. The mailbox being full is something we need to
    deal with in `SendFuture` which is why it is within another layer
    of `Result`.

commit ff782ad
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 18:52:27 2022 +0100

    Don't pass message name into envelope (#148)

    * Don't pass message name into envelope

    An envelope knows which message it is carrying, no need to pass
    it in.

    * Remove `MessageKind`

    Without the additional field for the message's name, this layer
    does not serve any purpose.

commit 910f3fb
Author: Restioson <restiosondev@gmail.com>
Date:   Mon Jul 18 17:35:52 2022 +0200

    Strip span fields from waiting and handler spans (#145)

    Currently, these are always children of the request span, so adding them as fields in them too is redundant.

    It would be a simplification to move this just to xtra_message_handler, but this is unfortunately not possible in the case of a disconnected actor which will never handle the message - this would miss all actor and message type information.

commit a0c1ad6
Author: Restioson <restiosondev@gmail.com>
Date:   Mon Jul 18 14:05:29 2022 +0200

    Improve span variables in tracing spans (#143)

    * Rename message -> message_type

    Weird behaviour is exhibited by some subscribers with `message` as a field name, probably since it is reserved, so `message_type` is used instead. "actor" was also renamed "actor_type" for consistency.

    Also, the display implementation of type name strings are used to avoid unnecessary quotes in the output.

commit a2bedd5
Author: Restioson <restiosondev@gmail.com>
Date:   Mon Jul 18 13:05:57 2022 +0200

    Create spans late (#120)

    Create the actor request span after first poll of `send` in order to ensure `send(...).instrument(...)` works as expected.

commit 7b0b6ea
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 09:53:35 2022 +0100

    Always store message priority in envelope (#142)

    * Always store message priority in envelope

    Previously, we only stored the priority of broadcast messages in
    the envelope. For regular messages, the priority was stored
    outside in a wrapper struct.

    This disparity is hard to understand as there is no apparent reason
    why the priority could not be stored in the message envelope.

    With this patch, we move the priority into the message envelope
    and also expose a `set_priority` function for each envelope type
    which allows us to consolidate the way we set priority from the
    `SendFuture`. To allow this, message envelopes need to be employ
    interior mutability (they are either box'ed or arc'ed). We use
    a spin lock which should basically be free. A `SendFuture` cannot
    be cloned and thus will never be accessed from two different
    threads, meaning the lock will never be congested.

commit fd3c282
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 09:41:54 2022 +0100

    Remove `ReceiveFuture::cancel` in favor of cancellation-safety docs (#138)

    Remove `ReceiveFuture::cancel` in favor of cancellation-safety docs, which fixes #134. Tests were added to check behaviour of resending messages on drop.

commit 07e4664
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 09:10:48 2022 +0100

    Introduce optional `sink` feature (#140)

    * Introduce optional `sink` feature

    Not every user may need the `sink` feature and we can minimise
    our dependency tree in those cases.

    * Extend README docs

    * Fix docs

commit bc7808c
Author: Thomas Eizinger <thomas@eizinger.io>
Date:   Mon Jul 18 07:07:45 2022 +0100

    Move channel logic into `Chan` (#122)

    Move the channel logic (e.g try_send and try_recv) into Chan itself.

    The goal was is follows:
        - Have Chan house the core channel implementation with descriptive function names for key operations
        - Have Sender and Receiver be thin layers on top that combine Chan with reference counting and provide a Future-based APIs for sending and receiving

commit e54b413
Author: Restioson <restiosondev@gmail.com>
Date:   Sun Jul 17 20:52:20 2022 +0200

    Annotate TickFuture with must_use (#139)

    This is since it is a future
@thomaseizinger thomaseizinger deleted the derive-macro branch September 15, 2022 03:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@Restioson Restioson Awaiting requested review from Restioson

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

Add a custom-derive for Actor
2 participants
@thomaseizinger @Restioson