New recording sink for an in-memory binary stream #6241
Assignees
Labels
Comments
5 tasks
jleibs
added a commit
that referenced
this issue
May 7, 2024
…ded bytes (#6242) ### What - Resolves: #6241 This effectively combines the guts of FileSink and MemorySink. Like the FileSink, this runs file-encoding on its own thread so that all we need to do when we're ready to read is move out the bytes. The downside relative to memory_sink is the stream is a singular RRD. You have no way of accessing information about individual messages, draining from the backlog, etc. This is designed to be as easy as possible to wire into a gradio output stream. Example usage in a Gradio Component: ```python @rr.thread_local_stream("rerun_example_live") def start_stream_direct(state): state['streaming'] = True cap = cv2.VideoCapture(0) frame_nr = 0 stream = rr.binary_stream() while state.get('streaming', False): ret, img = cap.read() frame_time_ms = cap.get(cv2.CAP_PROP_POS_MSEC) if frame_time_ms != 0: rr.set_time_nanos("frame_time", int(frame_time_ms * 1_000_000)) rr.set_time_sequence("frame_nr", frame_nr) frame_nr += 1 print("Processing frame", frame_nr) img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) # Log the original image rr.log("image/rgb", rr.Image(img).compress()) # Convert to grayscale gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) rr.log("image/gray", rr.Image(gray).compress()) # Run the canny edge detector canny = cv2.Canny(gray, 50, 200) rr.log("image/canny", rr.Image(canny).compress()) yield stream.read() ``` ### Checklist * [x] I have read and agree to [Contributor Guide](https://github.com/rerun-io/rerun/blob/main/CONTRIBUTING.md) and the [Code of Conduct](https://github.com/rerun-io/rerun/blob/main/CODE_OF_CONDUCT.md) * [x] I've included a screenshot or gif (if applicable) * [x] I have tested the web demo (if applicable): * Using examples from latest `main` build: [rerun.io/viewer](https://rerun.io/viewer/pr/6242?manifest_url=https://app.rerun.io/version/main/examples_manifest.json) * Using full set of examples from `nightly` build: [rerun.io/viewer](https://rerun.io/viewer/pr/6242?manifest_url=https://app.rerun.io/version/nightly/examples_manifest.json) * [x] The PR title and labels are set such as to maximize their usefulness for the next release's CHANGELOG * [x] If applicable, add a new check to the [release checklist](https://github.com/rerun-io/rerun/blob/main/tests/python/release_checklist)! - [PR Build Summary](https://build.rerun.io/pr/6242) - [Recent benchmark results](https://build.rerun.io/graphs/crates.html) - [Wasm size tracking](https://build.rerun.io/graphs/sizes.html) To run all checks from `main`, comment on the PR with `@rerun-bot full-check`.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
memory_recording()has adrainAPI, but the way it's structured it makes more sense for it to return self-contained RRD files.Other usages (in rust), also take advantage of the fact that it exposes APIs to directly consume buffered messages, rather than a raw stream of bytes.
The call to drain also incurs non-trivial overhead since we lazily do the encode at this step rather than eagerly encoding in a dedicated thread as with the File/Tcp sinks.
The text was updated successfully, but these errors were encountered: