Version: Next

MCAP

MCAP is a container format for recorded robotics data. In ROS 2, it commonly stores CDR-encoded messages across many topics with metadata like timestamps and schemas. ReductROS can extract and transform ROS 2 messages from MCAP files stored in ReductStore (content type application/mcap).

License Information

This feature is available under a commercial license. For testing, you can either use a free demo server (extension included) or request a demo license for your own deployment.

Get Demo Server Request Demo License

Query format

A user can use the ext query parameter to activate the ros extension and define the parameters for extracting and transforming ROS data in the following format:

{
    "ext": {
        "ros": {
            "extract": {
               "topic": "string",      # ROS topic name to extract from
               "encode": "object",     # e.g., {"data":"jpeg"} for JPEG encoding
                "as_label": "object"   # e.g., {"label_name": "path_to_json"}
            },

            "transform": {
              "include": "string[]",   # e.g., ["/topic-.*"]
              "exclude": "string[]",   # e.g., ["/topic-b", "/ext-.*"]
              "duration": "string",    # e.g., "1m" (max episode duration)
              "size": "string"         # e.g., "100MB" (max content length)
            }
        }
    }
}

Data extraction

The extract property allows you to specify the ROS topic from which to extract messages and convert them to JSON format. The extension will read the MCAP file, decode the messages from the specified topic, and return each message as a JSON record with the timestamp from the message header (when present).

Parameter	Type	Mandatory	Description
`topic`	`string`	Yes	The name of the ROS topic to extract messages from. This should match the topic names used in the MCAP file.
`encode`	`object`	No	A dictionary specifying how to encode binary data in the messages. For example, `{"data": "jpeg"}` for JPEG encoding.
`encode.<field>`	`string`	No	The encoding format for binary fields in the message. Supported values are `base64` and `jpeg`. If not specified, the field will be returned as a JSON list.
`as_label`	`object`	No	An object that specifies dynamic computed labels that will be delivered to the client. Can be a basis for output filtering.

Encoding binary data

The encode property allows you to specify how to handle binary data in the extracted messages. You can choose to encode binary fields in the following formats:

Format	Description
`base64`	Encodes binary data as a base64 string, suitable for text-based formats like JSON.
`jpeg`	Encodes binary data as a JPEG image encoded to a base64 string, suitable for image data.

info

Currently, the extension supports encoding only sensor_msgs/msg/Image messages with encoding set to rgb8, bgr8, or mono8.

Data transformation

The transform property allows you to generate new MCAP episodes with topic filtering and size/duration limits.

Field	Type	Mandatory	Description
`include`	`string[]`	No	List of topics to include. Supports regular expressions. If omitted, all topics are considered.
`exclude`	`string[]`	No	List of topics to exclude. Applied after `include`.
`duration`	`string`	No	Maximum duration per episode (e.g., `5m`, `2h`, `1d`).
`size`	`string`	No	Maximum content length per episode (e.g., `100MB`, `1GB`).

info

The size parameter refers to the uncompressed size of the episode. The actual file size may be smaller due to MCAP's chunk compression.

Transforming MCAP

When transform is provided, the extension:

Reads the source MCAP files.
Applies topic filters (include first, then exclude).
Streams matching messages into a new MCAP file.
Starts a new episode whenever duration or size is reached.

Example

{
  "ext": {
    "ros": {
      "transform": {
        "include": ["/camera/.*", "^/imu/data$"],
        "exclude": ["/camera/debug"],
        "duration": "10m",
        "size": "500MB"
      }
    }
  }
}

Result

All topics starting with /camera/ are included (because of /camera/.*).
The topic /imu/data is included exactly (because of ^/imu/data$).
The topic /camera/debug is excluded, even though it matches the /camera/.* rule.
Output is a sequence of MCAP episodes, each ≤ 10 minutes or ≤ 500 MB (whichever comes first).

Examples

The following examples demonstrate how to use the ReductROS extension to extract and transform ROS messages. Although this example is written in Python, it can be run using any of the official SDKs.

Extracting messages as JSON

This example demonstrates how to use the ROS extension to extract a topic from an MCAP file stored in ReductStore, convert it to JSON format, and use as_label plus when for filtering.

Python

from time import time_ns
from pathlib import Path

from reduct import Client

HERE = Path(__file__).parent


async def main():
    async with Client("http://localhost:8383", api_token="my-token") as client:
        bucket = await client.create_bucket(
            "my-bucket",
            exist_ok=True,
        )
        # Write a mcap file with timestamps
        now = time_ns() // 1000

        data = b""
        with open(f"{HERE}/../data/file.mcap", "rb") as f:
            data = f.read()

        await bucket.write("mcap", data, content_length=len(data), timestamp=now, content_type="application/mcap")

        # Prepare the query with the 'ros' extension
        condition = {
            "#ext": {
                "ros": {  # name of the extension to use
                    "extract": {
                        "topic": "/test",  # Specify the topic to extract from the mcap file
                        "as_label": {
                            "data1": "data",
                        },
                    },
                },
                "when": {
                    "@data1": {"$eq": "hello"},
                },
            }
        }

        # Query the data with the 'ros' extension
        async for record in bucket.query("mcap", start=now, when=condition):
            print(f"Record entry: {record.entry}")
            print(f"Record timestamp: {record.timestamp}")
            print(f"Record labels: {record.labels}")

            json = await record.read_all()
            print(json.decode("utf-8").strip())


# 5. Run the main function
if __name__ == "__main__":
    import asyncio

    asyncio.run(main())

Expected output

The expected output of the above code is as follows:

Record timestamp: 24
Record labels: {'@encoding': 'cdr', '@schema': 'std_msgs/String', '@topic': '/test', '@data1': 'hello'}
{"data":"hello"}

Explanation

The extension extracts ROS 2 messages from an .mcap file stored in ReductStore.
Only messages from the topic /test are selected using the topic filter in the ros.extract configuration.
The content of each message is CDR-encoded and decoded by the extension.
The decoded message is returned as JSON with the field data, matching the std_msgs/String schema.
Each record corresponds to one ROS 2 message and includes:
The decoded JSON payload, e.g., {"data":"hello"}
Message metadata as labels, including:
- topic: /test
- schema: std_msgs/String
- encoding: cdr
The as_label option adds @data1 based on the JSON path data.
The when filter selects only records where @data1 equals hello.

Extracting messages as JSON with JPEG encoding

This example demonstrates how to use the ROS extension to extract topic from an MCAP file stored in ReductStore and convert it to JSON format, while also encoding binary image data into JPEG format.

Python

import base64
import json
from time import time_ns
from pathlib import Path

from reduct import Client

HERE = Path(__file__).parent


async def main():
    async with Client("http://localhost:8383", api_token="my-token") as client:
        bucket = await client.create_bucket(
            "my-bucket",
            exist_ok=True,
        )
        # Write a mcap file with timestamps
        now = time_ns() // 1000

        data = b""
        with open(f"{HERE}/../data/camera_bag_0.mcap", "rb") as f:
            data = f.read()

        await bucket.write("mcap", data, content_length=len(data), timestamp=now, content_type="application/mcap")

        # Prepare the query with the 'ros' extension
        condition = {
            "#ext": {
                "ros": {  # name of the extension to use
                    "extract": {
                        "topic": "/image_raw",
                        # encode the data filed in http://docs.ros.org/en/noetic/api/sensor_msgs/html/msg/Image.html
                        "encode": {
                            "data": "jpeg",
                        },
                    },
                },
                "when": {  # optional filter to apply
                    "$limit": 1,  # return only one record
                }
            }
        }

        # Query the data with the 'ros' extension
        async for record in bucket.query("mcap", start=now, when=condition):
            print(f"Record entry: {record.entry}")
            print(f"Record timestamp: {record.timestamp}")
            print(f"Record labels: {record.labels}")
            content = await record.read_all()
            # Record content is a JSON object with metadata and base64-encoded data
            obj = json.loads(content)
            # Decode the base64-encoded data and save it as a JPEG file
            with open("output.jpg", "wb") as f:
                # Decode the base64-encoded data
                encoded = base64.decodebytes(obj["data"].encode("ascii"))
                f.write(encoded)

            # Print the image parameters without the data field
            del obj["data"]
            print(f"Image parameters: {obj}")


# 5. Run the main function
if __name__ == "__main__":
    import asyncio

    asyncio.run(main())

Expected output

The expected output of the above code is as follows:

Record timestamp: 1753341400522732
Record labels: {'@encoding': 'cdr', '@schema': 'sensor_msgs/Image', '@topic': '/image_raw'}
Image parameters: {'height': 720, 'width': 1280, 'is_bigendian': 0, 'encoding': 'rgb8', 'step': 3840, 'header': {'frame_id': 'camera', 'stamp': {'sec': 1753341400, 'nanosec': 522732248}}}

Explanation

The extension extracts ROS 2 messages from an .mcap file stored in ReductStore.
Only messages from the topic /image_raw are selected using the topic filter in the ros.extract configuration.
The content of each message is CDR-encoded and decoded by the extension.
The decoded message is returned as JSON with the field data, which contains the image data encoded in JPEG format.
Each record corresponds to one ROS 2 message and includes:
The decoded image parameters, such as height, width, encoding, and step.
The timestamp from the message header.
Message metadata as labels, including:
- topic: /image_raw
- schema: sensor_msgs/Image
- encoding: cdr

Transforming MCAP with splitting and topic filtering

This example demonstrates how to use the ROS extension to create new MCAP episodes while including and excluding topics using regular expressions.

Python

from time import time_ns
from pathlib import Path

from reduct import Client

HERE = Path(__file__).parent

from time import time_ns
from pathlib import Path

from reduct import Client

async def main():
    async with Client("http://localhost:8383", api_token="my-token") as client:
        bucket = await client.create_bucket(
            "my-bucket",
            exist_ok=True,
        )
        # Write an MCAP file with timestamps
        now = time_ns() // 1000
        with open(f"{HERE}/../data/multi_topic_5min.mcap", "rb") as f:
            data = f.read()

        await bucket.write(
            "mcap",
            data,
            content_length=len(data),
            timestamp=now,
            content_type="application/mcap",
        )

        # Prepare the query with the 'ros' extension (transform)
        condition = {
            "#ext": {
                "ros": {  # name of the extension to use
                    "transform": {
                        "include": ["/topic-.*"],
                        "exclude": ["/topic-b", "/ext-.*"],
                        "duration": "1m",
                        "size": "100MB"
                    },
                }
            }
        }

        # Query the data with the 'ros' extension
        async for record in bucket.query("mcap", start=now, when=condition):
            print(f"Record entry: {record.entry}")
            print(f"Record timestamp: {record.timestamp}")

            # Each record corresponds to a new MCAP episode
            data = await record.read_all()
            print(f"Episode file size: {len(data)} bytes")


if __name__ == "__main__":
    import asyncio
    asyncio.run(main())

Expected output

The expected output of the above code is as follows:

Record timestamp: 1755592486439270
Episode file size: 1602 bytes
Record timestamp: 1755592546439270
Episode file size: 1586 bytes
Record timestamp: 1755592606439270
Episode file size: 1582 bytes
Record timestamp: 1755592666439270
Episode file size: 1585 bytes
Record timestamp: 1755592726439270
Episode file size: 1558 bytes

Explanation

The extension reads an .mcap file stored in ReductStore.
Topics matching /topic-.* (aside from /topic-b) are included.
Topics matching /ext-.* are excluded.
The extension writes a sequence of new MCAP episodes capped at 1 minute or 100 MB, whichever comes first.

Query format​

Data extraction​

Encoding binary data​

Data transformation​

Transforming MCAP​

Examples​

Extracting messages as JSON​

Expected output​

Explanation​

Extracting messages as JSON with JPEG encoding​

Expected output​

Explanation​

Transforming MCAP with splitting and topic filtering​

Expected output​

Explanation​

Query format

Data extraction

Encoding binary data

Data transformation

Transforming MCAP

Examples

Extracting messages as JSON

Expected output

Explanation

Extracting messages as JSON with JPEG encoding

Expected output

Explanation

Transforming MCAP with splitting and topic filtering

Expected output

Explanation