-
-
Notifications
You must be signed in to change notification settings - Fork 2.5k
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.
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
task: add task IDs #4630
task: add task IDs #4630
Conversation
These IDs aren't necessarily guaranteed to be unique since the same address can be reused after being deallocated. I remember this being considered important for the standard library thread IDs for some reason. |
Yeah, I added a note in the docs stating that uniqueness is not guaranteed after a task has completed: tokio/tokio/src/runtime/task/mod.rs Lines 190 to 195 in a332b7a
If we're concerned about this, we could alternatively switch to sequential IDs, but I thought that the guarantees offered by using an address would be sufficient for implementing a I could change the implementation, though, if we don't think the uniqueness guarantees of addresses are strong enough. |
(i agree that a 64-bit task counter is probably not going to wrap in most reasonable programs, but...idk) |
I don't think you could get it to wrap even in unreasonable programs. |
@Darksonn I've changed this to assign task IDs based on a sequential counter, so they should remain unique after a task is deallocated. My main concern (more than the potential for IDs to wrap) was not wanting to put an additional field in the task header, but I realized this wasn't actually necessary and the ID can just go in the |
/// [unstable]: crate#unstable-features | ||
#[cfg_attr(docsrs, doc(cfg(all(feature = "rt", tokio_unstable))))] | ||
#[cfg_attr(not(tokio_unstable), allow(unreachable_pub))] | ||
// TODO(eliza): there's almost certainly no reason not to make this `Copy` as well... |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we should probably just derive Copy
as well, but i didn't add it out of an abundance of caution for forwards-compatibility. anyone have a problem with deriving Copy
for these?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Doesn't std
have Id
actually hold an Arc
? While i don't think we want to do this, we may want to wait hold off on Copy
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
But, I also don't care that much, if y'all think Copy
is good, do it.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Doesn't
std
haveId
actually hold anArc
? While i don't think we want to do this, we may want to wait hold off onCopy
.
Yeah, this is why I didn't add Copy
to this, so that we could reserve the right to change the repr however we want to later. I just think we are probably not going to actually do that. I'll hold off on making that commitment for now.
@@ -120,6 +123,7 @@ impl<T: Future, S: Schedule> Cell<T, S> { | |||
stage: CoreStage { | |||
stage: UnsafeCell::new(Stage::Running(future)), | |||
}, | |||
task_id, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
if we store this, it should probably be in the trailer no?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The header needs to stay under a cache line, but as long as CI passes, it should be good to go: https://github.com/tokio-rs/tokio/blob/master/tokio/src/runtime/task/core.rs#L266
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It has to be stored somewhere in the task structure because it's passed to poll_future
and cancel_task
in order to populate a JoinError
:
tokio/tokio/src/runtime/task/harness.rs
Lines 444 to 451 in a201b91
match res { | |
Ok(()) => { | |
stage.store_output(Err(JoinError::cancelled(id))); | |
} | |
Err(panic) => { | |
stage.store_output(Err(JoinError::panic(id, panic))); | |
} | |
} |
and
tokio/tokio/src/runtime/task/harness.rs
Line 479 in a201b91
Err(panic) => Err(JoinError::panic(id, panic)), |
It could go in the Core
or in the Trailer
--- my sense was that the Core
contained data that's used when actually polling the task, and the trailer contains data that's only used when the task completes. Technically the Id
is consumed only on completion, but inside of the functions that are called on every poll, so we would need to dereference the trailer in those functions to get the ID, and we don't currently dereference the trailer on polls.
I agree that it shouldn't go in the header; it currently doesn't.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, I had a few suggestions, but none are blocking at this point.
82787d5 also adds the runtime generated task IDs to the task |
## Motivation PR #4538 adds a prototype implementation of a `JoinMap` API in `tokio::task`. In [this comment][1] on that PR, @carllerche pointed out that a much simpler `JoinMap` type could be implemented outside of `tokio` (either in `tokio-util` or in user code) if we just modified `JoinSet` to return a task ID type when spawning new tasks, and when tasks complete. This seems like a better approach for the following reasons: * A `JoinMap`-like type need not become a permanent part of `tokio`'s stable API * Task IDs seem like something that could be generally useful outside of a `JoinMap` implementation ## Solution This branch adds a `tokio::task::Id` type that uniquely identifies a task relative to all currently spawned tasks. The ID is internally represented as a `NonZeroUsize` that's based on the address of the task's header. I thought that it was better to use addresses to generate IDs than assigning sequential IDs to tasks, because a sequential ID would mean adding an additional usize field to the task data somewhere, making it a word bigger. I've added methods to `JoinHandle` and `AbortHandle` for returning a task's ID. In addition, I modified `JoinSet` to add a `join_with_id` method that behaves identically to `join_one` but also returns an ID. This can be used to implement a `JoinMap` type. Note that because `join_with_id` must return a task ID regardless of whether the task completes successfully or returns a `JoinError` (in order to ensure that dead tasks are always cleaned up from a map), it inverts the ordering of the `Option` and `Result` returned by `join_one` --- which we've bikeshedded about a bit [here][2]. This makes the method's return type inconsistent with the existing `join_one` method, which feels not great. As I see it, there are three possible solutions to this: * change the existing `join_one` method to also swap the `Option` and `Result` nesting for consistency. * change `join_with_id` to return `Result<Option<(Id, T)>, (Id, JoinError)>>`, which feels gross... * add a task ID to `JoinError` as well. [1]: #4538 (comment) [2]: #4335 (comment)
Signed-off-by: Eliza Weisman <eliza@buoyant.io>
Signed-off-by: Eliza Weisman <eliza@buoyant.io>
This commit adds the runtime generated task IDs to the task `tracing` spans, when tracing is enabled. Unfortunately, this means the IDs have to be generated much higher up the callstack so they can be added to the tracing spans, and then passed down to the actual task machinery. But, I think this is worth it to be able to record that data. Signed-off-by: Eliza Weisman <eliza@buoyant.io>
Signed-off-by: Eliza Weisman <eliza@buoyant.io>
Signed-off-by: Eliza Weisman <eliza@buoyant.io>
Signed-off-by: Eliza Weisman <eliza@buoyant.io>
Signed-off-by: Eliza Weisman <eliza@buoyant.io>
Signed-off-by: Eliza Weisman <eliza@buoyant.io>
Does this expose an API to get the id of the current task? That's a common feature request. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM 👍 thanks
It does not currently, but it ought to be possible to add that in a subsequent PR. |
Signed-off-by: Eliza Weisman <eliza@buoyant.io>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM!
## Motivation In many cases, it is desirable to spawn a set of tasks associated with keys, with the ability to cancel them by key. As an example use case for this sort of thing, see Tower's [`ReadyCache` type][1]. Now that PR #4530 adds a way of cancelling tasks in a `tokio::task::JoinSet`, we can implement a map-like API based on the same `IdleNotifiedSet` primitive. ## Solution This PR adds an implementation of a `JoinMap` type to `tokio_util::task`, using the `JoinSet` type from `tokio::task`, the `AbortHandle` type added in #4530, and the new task IDs added in #4630. Individual tasks can be aborted by key using the `JoinMap::abort` method, and a set of tasks whose key match a given predicate can be aborted using `JoinMap::abort_matching`. When tasks complete, `JoinMap::join_one` returns their associated key alongside the output from the spawned future, or the key and the `JoinError` if the task did not complete successfully. Overall, I think the way this works is pretty straightforward; much of this PR is just API boilerplate to implement the union of applicable APIs from `JoinSet` and `HashMap`. Unlike previous iterations on the `JoinMap` API (e.g. #4538), this version is implemented entirely in `tokio_util`, using only public APIs from the `tokio` crate. Currently, the required `tokio` APIs are unstable, but implementing `JoinMap` in `tokio-util` means we will never have to make stability commitments for the `JoinMap` API itself. [1]: https://github.com/tower-rs/tower/blob/master/tower/src/ready_cache/cache.rs Signed-off-by: Eliza Weisman <eliza@buoyant.io>
# 1.18.0 (April 27, 2022) This release adds a number of new APIs in `tokio::net`, `tokio::signal`, and `tokio::sync`. In addition, it adds new unstable APIs to `tokio::task` (`Id`s for uniquely identifying a task, and `AbortHandle` for remotely cancelling a task), as well as a number of bugfixes. ### Fixed - blocking: add missing `#[track_caller]` for `spawn_blocking` ([#4616]) - macros: fix `select` macro to process 64 branches ([#4519]) - net: fix `try_io` methods not calling Mio's `try_io` internally ([#4582]) - runtime: recover when OS fails to spawn a new thread ([#4485]) ### Added - macros: support setting a custom crate name for `#[tokio::main]` and `#[tokio::test]` ([#4613]) - net: add `UdpSocket::peer_addr` ([#4611]) - net: add `try_read_buf` method for named pipes ([#4626]) - signal: add `SignalKind` `Hash`/`Eq` impls and `c_int` conversion ([#4540]) - signal: add support for signals up to `SIGRTMAX` ([#4555]) - sync: add `watch::Sender::send_modify` method ([#4310]) - sync: add `broadcast::Receiver::len` method ([#4542]) - sync: add `watch::Receiver::same_channel` method ([#4581]) - sync: implement `Clone` for `RecvError` types ([#4560]) ### Changed - update `nix` to 0.24, limit features ([#4631]) - update `mio` to 0.8.1 ([#4582]) - macros: rename `tokio::select!`'s internal `util` module ([#4543]) - runtime: use `Vec::with_capacity` when building runtime ([#4553]) ### Documented - improve docs for `tokio_unstable` ([#4524]) - runtime: include more documentation for thread_pool/worker ([#4511]) - runtime: update `Handle::current`'s docs to mention `EnterGuard` ([#4567]) - time: clarify platform specific timer resolution ([#4474]) - signal: document that `Signal::recv` is cancel-safe ([#4634]) - sync: `UnboundedReceiver` close docs ([#4548]) ### Unstable The following changes only apply when building with `--cfg tokio_unstable`: - task: add `task::Id` type ([#4630]) - task: add `AbortHandle` type for cancelling tasks in a `JoinSet` ([#4530], [#4640]) - task: fix missing `doc(cfg(...))` attributes for `JoinSet` ([#4531]) - task: fix broken link in `AbortHandle` RustDoc ([#4545]) - metrics: add initial IO driver metrics ([#4507]) [#4616]: #4616 [#4519]: #4519 [#4582]: #4582 [#4485]: #4485 [#4613]: #4613 [#4611]: #4611 [#4626]: #4626 [#4540]: #4540 [#4555]: #4555 [#4310]: #4310 [#4542]: #4542 [#4581]: #4581 [#4560]: #4560 [#4631]: #4631 [#4582]: #4582 [#4543]: #4543 [#4553]: #4553 [#4524]: #4524 [#4511]: #4511 [#4567]: #4567 [#4474]: #4474 [#4634]: #4634 [#4548]: #4548 [#4630]: #4630 [#4530]: #4530 [#4640]: #4640 [#4531]: #4531 [#4545]: #4545 [#4507]: #4507
# 1.18.0 (April 27, 2022) This release adds a number of new APIs in `tokio::net`, `tokio::signal`, and `tokio::sync`. In addition, it adds new unstable APIs to `tokio::task` (`Id`s for uniquely identifying a task, and `AbortHandle` for remotely cancelling a task), as well as a number of bugfixes. ### Fixed - blocking: add missing `#[track_caller]` for `spawn_blocking` ([#4616]) - macros: fix `select` macro to process 64 branches ([#4519]) - net: fix `try_io` methods not calling Mio's `try_io` internally ([#4582]) - runtime: recover when OS fails to spawn a new thread ([#4485]) ### Added - macros: support setting a custom crate name for `#[tokio::main]` and `#[tokio::test]` ([#4613]) - net: add `UdpSocket::peer_addr` ([#4611]) - net: add `try_read_buf` method for named pipes ([#4626]) - signal: add `SignalKind` `Hash`/`Eq` impls and `c_int` conversion ([#4540]) - signal: add support for signals up to `SIGRTMAX` ([#4555]) - sync: add `watch::Sender::send_modify` method ([#4310]) - sync: add `broadcast::Receiver::len` method ([#4542]) - sync: add `watch::Receiver::same_channel` method ([#4581]) - sync: implement `Clone` for `RecvError` types ([#4560]) ### Changed - update `nix` to 0.24, limit features ([#4631]) - update `mio` to 0.8.1 ([#4582]) - macros: rename `tokio::select!`'s internal `util` module ([#4543]) - runtime: use `Vec::with_capacity` when building runtime ([#4553]) ### Documented - improve docs for `tokio_unstable` ([#4524]) - runtime: include more documentation for thread_pool/worker ([#4511]) - runtime: update `Handle::current`'s docs to mention `EnterGuard` ([#4567]) - time: clarify platform specific timer resolution ([#4474]) - signal: document that `Signal::recv` is cancel-safe ([#4634]) - sync: `UnboundedReceiver` close docs ([#4548]) ### Unstable The following changes only apply when building with `--cfg tokio_unstable`: - task: add `task::Id` type ([#4630]) - task: add `AbortHandle` type for cancelling tasks in a `JoinSet` ([#4530], [#4640]) - task: fix missing `doc(cfg(...))` attributes for `JoinSet` ([#4531]) - task: fix broken link in `AbortHandle` RustDoc ([#4545]) - metrics: add initial IO driver metrics ([#4507]) [#4616]: #4616 [#4519]: #4519 [#4582]: #4582 [#4485]: #4485 [#4613]: #4613 [#4611]: #4611 [#4626]: #4626 [#4540]: #4540 [#4555]: #4555 [#4310]: #4310 [#4542]: #4542 [#4581]: #4581 [#4560]: #4560 [#4631]: #4631 [#4582]: #4582 [#4543]: #4543 [#4553]: #4553 [#4524]: #4524 [#4511]: #4511 [#4567]: #4567 [#4474]: #4474 [#4634]: #4634 [#4548]: #4548 [#4630]: #4630 [#4530]: #4530 [#4640]: #4640 [#4531]: #4531 [#4545]: #4545 [#4507]: #4507
# 1.18.0 (April 27, 2022) This release adds a number of new APIs in `tokio::net`, `tokio::signal`, and `tokio::sync`. In addition, it adds new unstable APIs to `tokio::task` (`Id`s for uniquely identifying a task, and `AbortHandle` for remotely cancelling a task), as well as a number of bugfixes. ### Fixed - blocking: add missing `#[track_caller]` for `spawn_blocking` ([#4616]) - macros: fix `select` macro to process 64 branches ([#4519]) - net: fix `try_io` methods not calling Mio's `try_io` internally ([#4582]) - runtime: recover when OS fails to spawn a new thread ([#4485]) ### Added - macros: support setting a custom crate name for `#[tokio::main]` and `#[tokio::test]` ([#4613]) - net: add `UdpSocket::peer_addr` ([#4611]) - net: add `try_read_buf` method for named pipes ([#4626]) - signal: add `SignalKind` `Hash`/`Eq` impls and `c_int` conversion ([#4540]) - signal: add support for signals up to `SIGRTMAX` ([#4555]) - sync: add `watch::Sender::send_modify` method ([#4310]) - sync: add `broadcast::Receiver::len` method ([#4542]) - sync: add `watch::Receiver::same_channel` method ([#4581]) - sync: implement `Clone` for `RecvError` types ([#4560]) ### Changed - update `nix` to 0.24, limit features ([#4631]) - update `mio` to 0.8.1 ([#4582]) - macros: rename `tokio::select!`'s internal `util` module ([#4543]) - runtime: use `Vec::with_capacity` when building runtime ([#4553]) ### Documented - improve docs for `tokio_unstable` ([#4524]) - runtime: include more documentation for thread_pool/worker ([#4511]) - runtime: update `Handle::current`'s docs to mention `EnterGuard` ([#4567]) - time: clarify platform specific timer resolution ([#4474]) - signal: document that `Signal::recv` is cancel-safe ([#4634]) - sync: `UnboundedReceiver` close docs ([#4548]) ### Unstable The following changes only apply when building with `--cfg tokio_unstable`: - task: add `task::Id` type ([#4630]) - task: add `AbortHandle` type for cancelling tasks in a `JoinSet` ([#4530], [#4640]) - task: fix missing `doc(cfg(...))` attributes for `JoinSet` ([#4531]) - task: fix broken link in `AbortHandle` RustDoc ([#4545]) - metrics: add initial IO driver metrics ([#4507]) [#4616]: #4616 [#4519]: #4519 [#4582]: #4582 [#4485]: #4485 [#4613]: #4613 [#4611]: #4611 [#4626]: #4626 [#4540]: #4540 [#4555]: #4555 [#4310]: #4310 [#4542]: #4542 [#4581]: #4581 [#4560]: #4560 [#4631]: #4631 [#4582]: #4582 [#4543]: #4543 [#4553]: #4553 [#4524]: #4524 [#4511]: #4511 [#4567]: #4567 [#4474]: #4474 [#4634]: #4634 [#4548]: #4548 [#4630]: #4630 [#4530]: #4530 [#4640]: #4640 [#4531]: #4531 [#4545]: #4545 [#4507]: #4507 Signed-off-by: Eliza Weisman <eliza@buoyant.io>
This PR contains the following updates: | Package | Type | Update | Change | |---|---|---|---| | [tokio](https://tokio.rs) ([source](https://github.com/tokio-rs/tokio)) | dependencies | minor | `1.17.0` -> `1.18.0` | | [tokio](https://tokio.rs) ([source](https://github.com/tokio-rs/tokio)) | dev-dependencies | minor | `1.17.0` -> `1.18.0` | --- ### Release Notes <details> <summary>tokio-rs/tokio</summary> ### [`v1.18.0`](https://github.com/tokio-rs/tokio/releases/tokio-1.18.0) [Compare Source](tokio-rs/tokio@tokio-1.17.0...tokio-1.18.0) ##### 1.18.0 (April 27, 2022) This release adds a number of new APIs in `tokio::net`, `tokio::signal`, and `tokio::sync`. In addition, it adds new unstable APIs to `tokio::task` (`Id`s for uniquely identifying a task, and `AbortHandle` for remotely cancelling a task), as well as a number of bugfixes. ##### Fixed - blocking: add missing `#[track_caller]` for `spawn_blocking` ([#​4616](tokio-rs/tokio#4616)) - macros: fix `select` macro to process 64 branches ([#​4519](tokio-rs/tokio#4519)) - net: fix `try_io` methods not calling Mio's `try_io` internally ([#​4582](tokio-rs/tokio#4582)) - runtime: recover when OS fails to spawn a new thread ([#​4485](tokio-rs/tokio#4485)) ##### Added - macros: support setting a custom crate name for `#[tokio::main]` and `#[tokio::test]` ([#​4613](tokio-rs/tokio#4613)) - net: add `UdpSocket::peer_addr` ([#​4611](tokio-rs/tokio#4611)) - net: add `try_read_buf` method for named pipes ([#​4626](tokio-rs/tokio#4626)) - signal: add `SignalKind` `Hash`/`Eq` impls and `c_int` conversion ([#​4540](tokio-rs/tokio#4540)) - signal: add support for signals up to `SIGRTMAX` ([#​4555](tokio-rs/tokio#4555)) - sync: add `watch::Sender::send_modify` method ([#​4310](tokio-rs/tokio#4310)) - sync: add `broadcast::Receiver::len` method ([#​4542](tokio-rs/tokio#4542)) - sync: add `watch::Receiver::same_channel` method ([#​4581](tokio-rs/tokio#4581)) - sync: implement `Clone` for `RecvError` types ([#​4560](tokio-rs/tokio#4560)) ##### Changed - update `mio` to 0.8.1 ([#​4582](tokio-rs/tokio#4582)) - macros: rename `tokio::select!`'s internal `util` module ([#​4543](tokio-rs/tokio#4543)) - runtime: use `Vec::with_capacity` when building runtime ([#​4553](tokio-rs/tokio#4553)) ##### Documented - improve docs for `tokio_unstable` ([#​4524](tokio-rs/tokio#4524)) - runtime: include more documentation for thread_pool/worker ([#​4511](tokio-rs/tokio#4511)) - runtime: update `Handle::current`'s docs to mention `EnterGuard` ([#​4567](tokio-rs/tokio#4567)) - time: clarify platform specific timer resolution ([#​4474](tokio-rs/tokio#4474)) - signal: document that `Signal::recv` is cancel-safe ([#​4634](tokio-rs/tokio#4634)) - sync: `UnboundedReceiver` close docs ([#​4548](tokio-rs/tokio#4548)) ##### Unstable The following changes only apply when building with `--cfg tokio_unstable`: - task: add `task::Id` type ([#​4630](tokio-rs/tokio#4630)) - task: add `AbortHandle` type for cancelling tasks in a `JoinSet` ([#​4530](tokio-rs/tokio#4530)], \[[#​4640](tokio-rs/tokio#4640)) - task: fix missing `doc(cfg(...))` attributes for `JoinSet` ([#​4531](tokio-rs/tokio#4531)) - task: fix broken link in `AbortHandle` RustDoc ([#​4545](tokio-rs/tokio#4545)) - metrics: add initial IO driver metrics ([#​4507](tokio-rs/tokio#4507)) </details> --- ### Configuration 📅 **Schedule**: At any time (no schedule defined). 🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied. ♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox. 🔕 **Ignore**: Close this PR and you won't be reminded about these updates again. --- - [ ] <!-- rebase-check -->If you want to rebase/retry this PR, click this checkbox. --- This PR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate). Co-authored-by: cabr2-bot <cabr2.help@gmail.com> Reviewed-on: https://codeberg.org/Calciumdibromid/CaBr2/pulls/1327 Reviewed-by: crapStone <crapstone@noreply.codeberg.org> Co-authored-by: Calciumdibromid Bot <cabr2_bot@noreply.codeberg.org> Co-committed-by: Calciumdibromid Bot <cabr2_bot@noreply.codeberg.org>
Motivation
PR #4538 adds a prototype implementation of a
JoinMap
API intokio::task
. In this comment on that PR, @carllerche pointed outthat a much simpler
JoinMap
type could be implemented outside oftokio
(either intokio-util
or in user code) if we just modifiedJoinSet
to return a task ID type when spawning new tasks, and whentasks complete. This seems like a better approach for the following
reasons:
JoinMap
-like type need not become a permanent part oftokio
'sstable API
a
JoinMap
implementationSolution
This branch adds a
tokio::task::Id
type that uniquely identifies atask relative to all other spawned tasks. Task IDs are assigned
sequentially based on an atomic
usize
counter of spawned tasks.In addition, I modified
JoinSet
to add ajoin_with_id
method thatbehaves identically to
join_one
but also returns an ID. This can beused to implement a
JoinMap
type.Note that because
join_with_id
must return a task ID regardless ofwhether the task completes successfully or returns a
JoinError
, I'vealso changed
JoinError
to carry the ID of the task that errored, andadded a
JoinError::id
method for accessing it. Alternatively, we couldhave done one of the following:
join_with_id
returnOption<(Id, Result<T, JoinError>)>
, whichwould be inconsistent with the return type of
join_one
(which we'vealready bikeshedded over once...)
join_with_id
returnResult<Option<(Id, T)>, (Id, JoinError)>>
,which just feels gross.
I thought adding the task ID to
JoinError
was the nicest option, andis potentially useful for other stuff as well, so it's probably a good API to
have anyway.
Closes #4538