rosbag  

rosbag is a node.js & browser compatible module for reading rosbag binary data files.

Installation

npm install rosbag

or

yarn add rosbag

Then, depending on your environment, you can import {open} from 'rosbag' or require('rosbag').

If you're not running your code in node.js or building for the browser using a package manager like webpack, you can import the script directly into the page:

<script src="node_modules/rosbag/dist/web/index.js"></script>
<script>
  // use rosbag.open() here...
</script>

Quick start

The most common way to interact with a rosbag is to read data records for a specific set of topics. The rosbag format encodes type information for topics, and rosbag reads this type information and parses the data records into JavaScript objects and arrays.

Here is an example of reading messages from a rosbag in node.js:

const { open } = require('rosbag');

async function logMessagesFromFooBar() {
  // open a new bag at a given file location:
  const bag = await open('../path/to/ros.bag');

  // read all messages from both the '/foo' and '/bar' topics:
  await bag.readMessages({ topics: ['/foo', '/bar'] }, (result) => {
    // topic is the topic the data record was in
    // in this case it will be either '/foo' or '/bar'
    console.log(result.topic);

    // message is the parsed payload
    // this payload will likely differ based on the topic
    console.log(result.message);
  });
}

logMessagesFromFooBar();

API

Opening a new rosbag reader

function open(fileOrPath: File | string) => Promise<Bag>

Opening a new rosbag reader is done with the open function. In the browser the function takes a File instance which you will generally get from a file input element. In node.js the function takes a string which should be the full path to a rosbag file. Node.js will read the file off of the disk. The promise will reject if there is an issue opening the file or if the file format is invalid, otherwise it will resolve with an instance of a Bag.

Bag instance

class Bag {
  // the time of the earliest message in the bag
  startTime: Time,

  // the time of the last message in the bag
  endTime: Time,

  // a hash of connection records by their id
  connections: { [number]: Connection },

  // an array of ChunkInfos describing the chunks within the bag
  chunkInfos: Array<ChunkInfo>,

  // call to consume from the bag - see 'Consuming messages from the bag instance' below
  readMessages(options: BagOptions, cb: (result: ReadResult) => void) => Promise<void>
}

Consuming messages from the bag instance

bag.readMessages method returns a Promise<void> which resolves when the read operation is completed or rejects in the event of a read error. During the read operation individual ReadResult objects are passed to the callback supplied to the open function. The callback may be called multiple times on the same tick as multiple data records can be encoded within a single binary chunk read within the bag reader.

BagOptions

const bagOptions = {

  // an optional array of topics used to filter down
  // which data records will be read
  // the default is all records on all topics
  topics?: Array<string>,

  // an optional Time instance used to filter data records
  // to only those which start on or after the given start time
  // the default is undefined which will apply no filter
  startTime?: Time,

  // an optional Time instance used to filter data records
  // to only those which end on or before the given end time
  // the default is undefined which will apply no filter
  endTime? Time,

  // decompression callbacks:
  // if your bag is compressed you can supply a callback to decompress it
  // based on the compression type. The callback should accept a buffer of compressed bytes
  // and return a buffer of uncompressed bytes.  For examples on how to decompress lz4 and bz2 compressed bags
  // please see the tests here: https://github.com/cruise-automation/rosbag.js/blob/545529344c8c2a0b3a3126646d065043c2d67d84/src/bag.test.js#L167-L192
  // The decompression callback is also passed the uncompressedByteLength which is stored in the bag.
  // This byte length can be used with some decompression libraries to increase decompression efficiency.
  decompress?: {|
    bz2?: (buffer: Buffer, uncompressedByteLength: number) => Buffer,
    lz4?: (buffer: Buffer, uncompressedByteLength: number) => Buffer,
  |},

  // by default the individual parsed binary messages will be parsed based on their [ROS message definition](http://wiki.ros.org/msg)
  // if you set noParse to true the read operation will skip the message parsing step
  noParse?: boolean,

  // Whether the resulting messages should be deeply frozen using Object.freeze(). (default: false)
  // Useful to make sure your code or libraries doesn't accidentally mutate bag messages.
  freeze?: boolean,
}

All options are optional and used to filter down from the sometimes enormous and varied data records in a rosbag. One could omit all options & filter the messages in memory within the readMessages callback; however, due to the rosbag format optimizations can be made during reading & parsing which will yield significant performance and memory gains if you specify topics and/or date ranges ahead of time.

For ROS message definitions that contain a string field preceded by a #pragma rosbag_parse_json comment, rosbag will parse that string field into JSON. For example, the message definition below has a data field containing stringified JSON; rosbag will parse that string into JSON while reading messages from a bag instance.

#pragma rosbag_parse_json
string data

ReadResult

class ReadResult {

  // the topic from which the current record was read
  topic: string,

  // the parsed message contents as a JavaScript object
  // this can contain nested complex types
  // and arrays of complex & simple types
  // this will be undefined if you supply { noParse: true } to `bag.readMessages`
  message: { [string]: any },

  // a Time instance - the receive time of the message
  timestamp: Time

  // the raw buffer data from the data record
  // a node.js buffer in node & an array buffer in the browser
  data: Array<int8>,

  // the offset of the chunk being read
  // starts at 0 and eventually increments to totalChunks
  // useful for computing read progress as a percentage
  chunkOffset: number,

  // the total chunks to eventually be consumed
  // during the current read operation
  totalChunks: number,
}

Connection

class Connection {
  // the id of the connection
  conn: number,

  // the topic for the connection
  topic: string,

  // the md5 hash for the connection message definition
  md5sum: string,

  // the rosbag formatted message definition for records on this connection's topic
  messageDefinition: string,
}

Time

ROS represents time with nanosecond granularity. In JavaScript, a ROS Time value is stored as an object containing sec and nsec fields. The TimeUtil module has various utility methods for comparison, arithmetic, and conversion to/from JavaScript Date objects.

interface TimeUtil {
  // convert a Time object to a JavaScript Date object
  // note: this is a lossy conversion as JavaScript dates do not store nanoseconds
  toDate(Time): Date,

  // build a time instance from a JavaScript date object
  fromDate(Date): Time,

  // returns a positive number if left is greater than right
  // returns a negative number if right is greater than left
  // returns 0 if the times are the same
  // useful to sort an array of times:
  // const times = [{sec: 1, nsec: 1000}, {sec: 2, nsec: 2000}, {sec: 0, nsec: 100}]
  // const sortedTimes = times.sort(Time.compare)
  compare(left: Time, right: Time): number,

  // returns true if left is less than right, otherwise false
  isLessThan(left: Time, right: Time): boolean,

  // returns true if left is greater than right, otherwise false
  isGreaterThan(left: Time, right: Time): boolean,

  // returns true if the times are the same, otherwise false
  areSame(left: Time, right: Time): boolean,

  // computes the sum of two times and returns a new time
  add(left: Time, right: Time): Time,
}

Supported platforms

Currently rosbag is used & heavily tested in [email protected] as well as Google Chrome (via webpack). It should also work under all modern browsers which have the FileReader and typed array APIs available. If you run into issues with Firefox, Edge, or Safari please feel free to open an issue or submit a pull request with a fix.