New example for mcap using RangedPlayback by ericmlujan · Pull Request #859 · foxglove/foxglove-sdk

ericmlujan · 2026-02-04T19:06:48Z

Changelog

Add Rust example for using the RangedPlayback capability

Docs

This is an example.

Description

Add a Rust example using the RangedPlayback websocket capability to load and play back an mcap file. This example introduces a documented PlaybackSource trait, which others can pattern match to when implementing their own loaders for custom data formats.

This PR supersedes and simplifies #772.

A Python version of this example exists at #862.
A C++ version of this example exists at #861.

Fixes: FLE-131

linear · 2026-02-04T19:06:51Z

FLE-131 Update mcap stream example to support RangedPlayback

claude · 2026-02-04T19:07:25Z

Claude finished @ericmlujan's task in 1m 29s —— View job

Summary

This PR adds a new Rust example demonstrating the RangedPlayback websocket capability for MCAP file playback. The implementation introduces a well-documented PlaybackSource trait that abstracts playback operations, making it reusable for other data formats. The code has been significantly simplified from the superseded PR #772, with good separation of concerns across three modules: main.rs, mcap_player.rs, and playback_source.rs.

The PR has gone through multiple rounds of review with substantive feedback from cursor, gasmith, and dave-foxglove, with most issues addressed in subsequent commits.

Blockers

None

Suggestions

1. Unanswered review question about playback speed in `log_next_message` doc comment

dave-foxglove asked in the review comments whether log_next_message should mention something about playback speed in its doc comment - i.e., whether the caller should divide the returned Duration by the playback speed, or if that's already accounted for. The current doc comment says "(accounting for the current playback speed)" which implies it's handled internally, but explicit confirmation or a sentence clarifying this would help future implementers.

rust/examples/ws_stream_mcap/src/playback_source.rs:56-65

2. Consider clarifying the "log" naming convention

dave-foxglove also asked why log_next_message has a "log_" prefix. In the context of foxglove SDK, "log" refers to logging messages to channels (as seen in channel.log_with_meta()), but this could be confusing for newcomers who might expect file logging. A brief note in the trait doc comment explaining that "log" in this context means "emit to Foxglove channels" might help.

3. Minor: PR title and description are clear 👍

The changelog, docs, and description sections follow the template well. The description explains the motivation (supersedes #772) and the key abstraction being introduced (PlaybackSource trait).

Review completed. The code looks well-structured and the significant issues from previous review rounds have been addressed. The two open questions from dave-foxglove about documentation are worth addressing but are non-blocking.

rust/foxglove/Cargo.toml

rust/examples/ws_stream_mcap/src/main.rs

rust/examples/ws_stream_mcap/src/mcap_player.rs

rust/examples/ws_stream_mcap/src/main.rs

Co-authored-by: Eric Lujan <ericmlujan@users.noreply.github.com>

rust/examples/ws_stream_mcap/src/main.rs

rust/examples/ws_stream_mcap/src/mcap_player.rs

- Add check in play() to prevent transitioning from Ended to Playing - Reset status to Paused when seeking from Ended state - Reorder listener to handle seek before play/pause for looping support Co-authored-by: Eric Lujan <ericmlujan@users.noreply.github.com>

Move current_time assignment after fallible reader creation to avoid inconsistent state if IndexedReader::new_with_options fails Co-authored-by: Eric Lujan <ericmlujan@users.noreply.github.com>

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

rust/examples/ws_stream_mcap/src/mcap_player.rs

Prevent pause() from transitioning out of Ended state, which would bypass the guard in play() and create a confusing state cycle Co-authored-by: Eric Lujan <ericmlujan@users.noreply.github.com>

dave-foxglove · 2026-02-09T19:18:59Z

rust/examples/ws_stream_mcap/src/mcap_player.rs

+        self.status
+    }
+
+    fn current_time(&self) -> u64 {


For all references to a u64 as the player time, can we define and use a named type instead of a raw u64.

Maybe a "newtype" like McapTimestamp and McapDuration (similar to rust standard timestamps and durations).

P.S. I'm also up for a "type alias" like type McapNanos = u64 but it doesn't offer the whoops-protection a newtype does.

dave-foxglove · 2026-02-09T19:19:27Z

rust/examples/ws_stream_mcap/src/mcap_player.rs

+        self.status
+    }
+
+    fn current_time(&self) -> u64 {


For all references to a u64 as the player time, can we define and use a named type instead of a raw u64.

Maybe a "newtype" like McapTimestamp and McapDuration (similar to rust standard timestamps and durations).

P.S. I'm also up for a "type alias" like type McapNanos = u64 but it doesn't offer the whoops-protection a newtype does.

Addressed in 25abf13. I decided against introducing a full tuple struct for nanoseconds and duration; I found it added a bit too much boilerplate in the context of a code example.

dave-foxglove · 2026-02-09T19:25:37Z

rust/examples/ws_stream_mcap/src/playback_source.rs

+    /// Used to send a PlaybackState to Foxglove
+    fn playback_speed(&self) -> f32;
+
+    /// Logs the next message for playback if it's ready, or returns a duration to wait.


Sorry if I'm missing something, but could/should this mention something about playback speed? E.e. if the playback speed is 10x should I sleep for 1/10 the returned Duration (or is it already accounting for that?)

Addressed in 22218a6. It is already accounting for that

dave-foxglove · 2026-02-09T19:27:41Z

rust/examples/ws_stream_mcap/src/playback_source.rs

+use foxglove::{websocket::PlaybackStatus, WebSocketServerHandle};
+
+/// A data source that supports ranged playback with play/pause, seek, and variable speed.
+///


It looks like log_next_message is expected to be called in a loop(?). If true, let's mention that in the overall doc comment.

Addressed in 22218a6

dave-foxglove · 2026-02-09T19:30:40Z

rust/examples/ws_stream_mcap/src/playback_source.rs

+    ///
+    /// The caller should sleep outside of any lock to allow control requests to be processed.
+    /// This method also broadcasts time updates via `server.broadcast_time()`.
+    fn log_next_message(&mut self, server: &WebSocketServerHandle) -> Result<Option<Duration>>;


I'm likely missing some context in the wider code base, but in just this interface/api - I'm not clear on why this functions has a "log_*" prefix on it. If it's just pushing a message to a log file, then it's perfectly named, but I have an impression that it's doing something more or different(?)

Good question. "Logging" is the general term for sending your robotics data over to Foxglove. We have the notion of a "channel", which you can think of as a topic that data messages can be logged to/sent over, and a channel can be backed by one or more sinks, which handle actually doing something with that data (i.e. saving to an MCAP file or sending over a WebSocket). In this case, this function calls a Channel::log() on a bunch of channels that are set up to publish the data over a WebSocket; hence the name.

dave-foxglove · 2026-02-09T19:32:08Z

rust/examples/ws_stream_mcap/src/mcap_player.rs

+    pending_message: Option<(mcap::records::MessageHeader, Vec<u8>)>,
+}
+
+impl McapPlayer {


Can we describe what an McapPlayer is and what it is used for (asking for doc comments), either here, or on the struct (line 19) or at the top of the file (i.e. //! comments).

Addressed in 22218a6

ericmlujan added 18 commits January 22, 2026 17:05

First version of replay extended to be live

418ca12

Support did_seek field

b0b1ff7

Migrate from atomic to mutex

57ae303

Update to take start time as arg

32cedd2

Simplify time tracking

243de5f

Working commit

97d1c12

More work

863e7a8

Move files to different modules

989e9cd

Slight restructure of main loop

873609f

Document trait

96f34b1

Reset changes

a986a64

Manual preliminary setup

cce0621

Working manual changes

65dd479

Time range implementation

86f4897

Making progress

40ed4db

Working

7a1d42a

ckpt

edf77da

Working and fixed

c016e5f

ericmlujan requested a review from gasmith February 4, 2026 19:06

github-actions bot deployed to foxglove-sdk-api-docs (Preview) February 4, 2026 19:09 View deployment

foxglovebot deployed to foxglove-sdk-playground (Preview) February 4, 2026 19:10 View deployment

cursor bot reviewed Feb 4, 2026

View reviewed changes

rust/foxglove/Cargo.toml Outdated Show resolved Hide resolved

rust/examples/ws_stream_mcap/src/main.rs Show resolved Hide resolved

Bugbot

b920237

github-actions bot deployed to foxglove-sdk-api-docs (Preview) February 4, 2026 19:17 View deployment

foxglovebot deployed to foxglove-sdk-playground (Preview) February 4, 2026 19:19 View deployment

cursor bot reviewed Feb 4, 2026

View reviewed changes

rust/examples/ws_stream_mcap/src/mcap_player.rs Show resolved Hide resolved

ericmlujan commented Feb 4, 2026

View reviewed changes

rust/examples/ws_stream_mcap/src/main.rs Outdated Show resolved Hide resolved

Combine lock acquisitions when broadcasting ended state

abe1adb

Co-authored-by: Eric Lujan <ericmlujan@users.noreply.github.com>

Address review feedback

0a3cee4

github-actions bot deployed to foxglove-sdk-api-docs (Preview) February 7, 2026 00:26 View deployment

foxglovebot deployed to foxglove-sdk-playground (Preview) February 7, 2026 00:28 View deployment

cursor bot reviewed Feb 7, 2026

View reviewed changes

rust/examples/ws_stream_mcap/src/main.rs Outdated Show resolved Hide resolved

rust/examples/ws_stream_mcap/src/mcap_player.rs Show resolved Hide resolved

Fix max

075e967

github-actions bot deployed to foxglove-sdk-api-docs (Preview) February 7, 2026 00:43 View deployment

foxglovebot deployed to foxglove-sdk-playground (Preview) February 7, 2026 00:45 View deployment

cursor bot reviewed Feb 7, 2026

View reviewed changes

rust/examples/ws_stream_mcap/src/mcap_player.rs Show resolved Hide resolved

github-actions bot deployed to foxglove-sdk-api-docs (Preview) February 7, 2026 01:04 View deployment

foxglovebot deployed to foxglove-sdk-playground (Preview) February 7, 2026 01:05 View deployment

Fix state consistency in reset_reader

091f4bb

Move current_time assignment after fallible reader creation to avoid inconsistent state if IndexedReader::new_with_options fails Co-authored-by: Eric Lujan <ericmlujan@users.noreply.github.com>

github-actions bot deployed to foxglove-sdk-api-docs (Preview) February 7, 2026 01:12 View deployment

foxglovebot deployed to foxglove-sdk-playground (Preview) February 7, 2026 01:14 View deployment

cursor bot reviewed Feb 7, 2026

View reviewed changes

rust/examples/ws_stream_mcap/src/mcap_player.rs Show resolved Hide resolved

Add Ended guard to pause() method

69145d5

Prevent pause() from transitioning out of Ended state, which would bypass the guard in play() and create a confusing state cycle Co-authored-by: Eric Lujan <ericmlujan@users.noreply.github.com>

github-actions bot deployed to foxglove-sdk-api-docs (Preview) February 7, 2026 01:48 View deployment

foxglovebot deployed to foxglove-sdk-playground (Preview) February 7, 2026 01:50 View deployment

This was referenced Feb 7, 2026

C++ example for mcap using RangedPlayback #861

Open

Python example for mcap using RangedPlayback #862

Open

dave-foxglove reviewed Feb 9, 2026

View reviewed changes

ericmlujan added 2 commits February 9, 2026 18:35

Update doc comments

22218a6

Use type alias for Nanoseconds

25abf13

github-actions bot deployed to foxglove-sdk-api-docs (Preview) February 10, 2026 03:18 View deployment

foxglovebot deployed to foxglove-sdk-playground (Preview) February 10, 2026 03:20 View deployment

Conversation

ericmlujan commented Feb 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changelog

Docs

Description

Uh oh!

linear bot commented Feb 4, 2026

Uh oh!

claude bot commented Feb 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Blockers

Suggestions

1. Unanswered review question about playback speed in log_next_message doc comment

2. Consider clarifying the "log" naming convention

3. Minor: PR title and description are clear 👍

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cursor bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

5 participants

ericmlujan commented Feb 4, 2026 •

edited

Loading

claude bot commented Feb 4, 2026 •

edited

Loading

1. Unanswered review question about playback speed in `log_next_message` doc comment