Note: This tutorial assumes that you have completed the previous tutorials: Recording and playing back data.
(!) Please ask about problems and questions regarding this tutorial on answers.ros.org. Don't forget to include in your question the link to this page, the versions of your OS & ROS, and also add appropriate tags.

Reading messages from a bag file

Description: Learn two ways to read messages from desired topics in a bag file, including using the really handy ros_readbagfile script.

Keywords: data, rosbag, extract, play, info, bag, messages, readbagfile, ros_readbagfile

Tutorial Level: BEGINNER

Next Tutorial: Producing filtered bag files

Contents

  1. Download or record a bag file
  2. Option 1: play back the messages immediately and look at the output in multiple terminals
  3. Option 2: use the ros_readbagfile script to easily extract the topics of interest
    1. More .yaml file analysis
      1. Example 1:
      2. Example 2:
  4. Why use `ros_readbagfile` for this purpose instead of `rostopic echo -b`?

Download or record a bag file

First, you need a bag file. Either produce your own by following this tutorial (ROS/Tutorials/Recording and playing back data), or download the https://webviz.io demo one from here (https://open-source-webviz-ui.s3.amazonaws.com/demo.bag), using the wget command like this: 

wget https://open-source-webviz-ui.s3.amazonaws.com/demo.bag

Alternatively, assuming you are on a system with ROS already running, here is a quick command to record a 30 second snippet of data into a bag file for just topics you are interested in, ex: /topic1, /topic2, and /topic3. Since we are setting a duration of 30 seconds, the recording will automatically stop after this time: 

rosbag record --duration=30 --output-name=/tmp/mybagfile.bag \
    /topic1 /topic2 /topic3

---

The rest of this tutorial will be done assuming you've downloaded the webviz.io demo bag file using the wget command as shown above. You will go through two options to read/extract messages from the bag file.

Note that in any of the commands below, the time command is prepended simply because it will print out how long each command takes, and since sometimes these commands can take a long time, it is useful to use the time command to gain an idea of how long a given command should take. If you don't want to use it, you may remove the time part of any of the commands below.

Option 1: play back the messages immediately and look at the output in multiple terminals

Source: this material was adapted from instructions first published in this document here.

  1. You need to know the exact topic names you'd like to read from the bag file. So, let's see what's in the bag file. In any terminal, manually inspect all published topics and how many messages were published to each topic with this command: 

    time rosbag info demo.bag  
# OR (if you know part of the topic names of interest before-hand):
time rosbag info mybag.bag | grep -E "(topic1|topic2|topic3)"
    Sample output: 
    $ time rosbag info demo.bag  
path:        demo.bag
version:     2.0
duration:    20.0s
start:       Mar 21 2017 19:37:58.00 (1490150278.00)
end:         Mar 21 2017 19:38:17.00 (1490150298.00)
size:        696.2 MB
messages:    5390
compression: none [600/600 chunks]
types:       bond/Status                      [eacc84bf5d65b6777d4c50f463dfb9c8]
             diagnostic_msgs/DiagnosticArray  [60810da900de1dd6ddd437c3503511da]
             diagnostic_msgs/DiagnosticStatus [d0ce08bc6e5ba34c7754f563a9cabaf1]
             nav_msgs/Odometry                [cd5e73d190d741a2f92e81eda573aca7]
             radar_driver/RadarTracks         [6a2de2f790cb8bb0e149d45d297462f8]
             sensor_msgs/Image                [060021388200f6f0f447d0fcd9c64743]
             sensor_msgs/NavSatFix            [2d3a8cd499b9b4a0249fb98fd05cfa48]
             sensor_msgs/PointCloud2          [1158d486dd51d683ce2f1be655c3c181]
             sensor_msgs/Range                [c005c34273dc426c67a020a87bc24148]
             sensor_msgs/TimeReference        [fded64a0265108ba86c3d38fb11c0c16]
             tf2_msgs/TFMessage               [94810edda583a504dfda3829e70d7eec]
             velodyne_msgs/VelodyneScan       [50804fc9533a0e579e6322c04ae70566]
topics:      /diagnostics                      140 msgs    : diagnostic_msgs/DiagnosticArray 
             /diagnostics_agg                   40 msgs    : diagnostic_msgs/DiagnosticArray 
             /diagnostics_toplevel_state        40 msgs    : diagnostic_msgs/DiagnosticStatus
             /gps/fix                          146 msgs    : sensor_msgs/NavSatFix           
             /gps/rtkfix                       200 msgs    : nav_msgs/Odometry               
             /gps/time                         192 msgs    : sensor_msgs/TimeReference       
             /image_raw                        600 msgs    : sensor_msgs/Image               
             /obs1/gps/fix                      30 msgs    : sensor_msgs/NavSatFix           
             /obs1/gps/rtkfix                  200 msgs    : nav_msgs/Odometry               
             /obs1/gps/time                    136 msgs    : sensor_msgs/TimeReference       
             /radar/points                     400 msgs    : sensor_msgs/PointCloud2         
             /radar/range                      400 msgs    : sensor_msgs/Range               
             /radar/tracks                     400 msgs    : radar_driver/RadarTracks        
             /tf                              1986 msgs    : tf2_msgs/TFMessage              
             /velodyne_nodelet_manager/bond     80 msgs    : bond/Status                     
             /velodyne_packets                 200 msgs    : velodyne_msgs/VelodyneScan      
             /velodyne_points                  200 msgs    : sensor_msgs/PointCloud2

real    0m1.003s
user    0m0.620s
sys 0m0.283s

    Notice that there are 30 messages published on topic /obs1/gps/fix, and 40 on topic /diagnostics_agg. Let's just extract those.

  2. In terminal 1 (this terminal, for example), start up a ros core, which runs the required ROS master node: 
    roscore

  3. Open up another terminal. Note: to open up another terminal tab in the same terminal window you can use Ctrl + Shift + T on Ubuntu. Subscribe to the /obs1/gps/fix topic, echoing (printing) everything published on this topic, while also teeing it to a file for later review, all in yaml format: 

    rostopic echo /obs1/gps/fix | tee topic1.yaml
    You'll see: 
    $ rostopic echo /obs1/gps/fix | tee topic1.yaml
WARNING: topic [/obs1/gps/fix] does not appear to be published yet

  4. Open up another terminal. Subscribe to the other topic: /diagnostics_agg. 

    rostopic echo /diagnostics_agg | tee topic2.yaml
    You'll see: 
    $ rostopic echo /diagnostics_agg | tee topic2.yaml
WARNING: topic [/diagnostics_agg] does not appear to be published yet
  5. Repeat this process for as many topics as you like. Each topic must have its own terminal.

  6. Open up another terminal to play the bag file. We will now play back the bag file as quickly as possible (using the --immediate option), publishing ONLY the topics of interest. The format is: 

    time rosbag play --immediate demo.bag --topics /topic1 /topic2 /topic3 /topicN
    So in our case, the command would be: 
    time rosbag play --immediate demo.bag --topics /obs1/gps/fix /diagnostics_agg
    You'll see: 
    $ time rosbag play --immediate demo.bag --topics /obs1/gps/fix /diagnostics_agg
[ INFO] [1591916465.758724557]: Opening demo.bag

Waiting 0.2 seconds after advertising topics... done.

Hit space to toggle paused, or 's' to step.
 [RUNNING]  Bag Time: 1490150297.770734   Duration: 19.703405 / 19.703405               
Done.

real  0m1.570s
user  0m0.663s
sys 0m0.394s

  7. Done! Now go look at your two terminals which were each subsribed to a topic, and you'll see the output of all messages for each topic type, in YAML format, with a --- line between each message. Use a text editor of your choice, preferably with Syntax Highlighting for YAML file types (ex: Sublime Text 3) to view the messages in the files. The last two messages in topic1.yaml, for instance, look like this: 

    ---
header: 
  seq: 4027
  stamp: 
    secs: 1490150296
    nsecs:  66947432
  frame_id: "gps"
status: 
  status: 0
  service: 1
latitude: 37.4008017844
longitude: -122.108119889
altitude: -6.4380177824
position_covariance: [1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0]
position_covariance_type: 0
---
header: 
  seq: 4028
  stamp: 
    secs: 1490150297
    nsecs: 744347249
  frame_id: "gps"
status: 
  status: 0
  service: 1
latitude: 37.4007565466
longitude: -122.108159482
altitude: -6.35130467023
position_covariance: [1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0]
position_covariance_type: 0
---

    If for some reason one of your rostopic processes missed the messages, just kill its process with Ctrl + C, restart it, and call the rosbag play command again.

Option 2: use the ros_readbagfile script to easily extract the topics of interest

Source: this material was adapted from instructions first published in this document here, and the Python script is from here: ros_readbag.py.

Note: you can kill any running processes. No roscore, for instance, is required.

  1. Download and install `ros_readbag.py` using these commands. Be sure to read the comments in the top of the file for the latest information on installation instructions and python dependencies. 

    # Download the file
wget https://raw.githubusercontent.com/ElectricRCAircraftGuy/eRCaGuy_dotfiles/master/useful_scripts/ros_readbagfile.py
# Make it executable
chmod +x ros_readbagfile.py
# Ensure you have the ~/bin directory for personal binaries
mkdir -p ~/bin
# Move this executable script into that directory as `ros_readbagfile`, so that it will
# be available as that command
mv ros_readbagfile.py ~/bin/ros_readbagfile
# Re-source your ~/.profile file to ensure ~/bin is in your PATH, so you can use this
# new `ros_readbagfile` command you just installed
# Note: this assumes you are on Linux Ubuntu with a default `~/.profile` file. If you're
# on some other Linux distribution, such as Arch Linux, you can manually create the
# `~/.profile` file. See here for details: 
# https://answers.ros.org/question/371583/ros_readbagfile-command-not-found/?answer=403354#post-id-403354
. ~/.profile
#
# Install python dependencies (see the comments at the top of the ros_readbagfile.py 
# file for the latest dependencies and instructions)
pip install bagpy
pip3 install bagpy

    Note: if your terminal still says it cannot find the command when trying to run it, you may need to ensure ~/bin is part of your PATH variable. See my answer here for details.

  2. Determine the exact topic names you'd like to read from the bag file, by using rosbag info, as shown in Step 1 of Option 1 above.

  3. Now use ros_readbagfile. The general format is: 

    # read these topics and print them to stdout
time ros_readbagfile <mybagfile.bag> [topic1] [topic2] [topic3] [...]

# Write to the topics.yaml file withOUT also printing to stdout
time ros_readbagfile <mybagfile.bag> [topic1] [topic2] [topic3] [...] > topics.yaml

# OR (preferred, so you can easily see it is still running): write to the
# topics.yaml file AND print to stdout
time ros_readbagfile <mybagfile.bag> [topic1] [topic2] [topic3] [...] | tee topics.yaml
    To read the same messages shown in Option 1 above, use: 
    # (preferred, so you can easily see it is still running): write to the
# topics.yaml file AND print to stdout
time ros_readbagfile demo.bag /obs1/gps/fix /diagnostics_agg | tee topics.yaml
    Optionally, to see "progress" by watching the yaml file grow in size, in a new terminal, run the following. Note: the yaml text file will be about 2x as large as the original binary .bag file when complete. 
    watch -n 1 'du -sk topics.yaml | awk '\''{printf "%.3f MiB %s\n", $1/1024, $2}'\'''
    That's it! You'll see it print out all 70 messages quickly. Here is what the last little bit of the terminal output looks like: 
            key: "Early diagnostic update count:"
        value: "0"
      - 
        key: "Zero seen diagnostic update count:"
        value: "0"

# =======================================
# topic:           /obs1/gps/fix
# msg_count:       30
# timestamp (sec): 1490150297.770734310
# - - -
header: 
  seq: 4028
  stamp: 
    secs: 1490150297
    nsecs: 744347249
  frame_id: "gps"
status: 
  status: 0
  service: 1
latitude: 37.40075654660259
longitude: -122.10815948235131
altitude: -6.351304670230949
position_covariance: [1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0]
position_covariance_type: 0

# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
# Total messages found:               70
#
#    /diagnostics_agg:                40
#    /obs1/gps/fix:                   30
#
# DONE.
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

real    0m0.812s
user    0m0.599s
sys 0m0.077s

    Now view topics.yaml with your preferred text editor or viewer (ex: Sublime Text 4, gedit, emacs, vim, less, etc) to see all of the messages it extracted from the bag file.

    Note that topics.yaml is a true YAML-formatted file. All non-YAML content, such as message separators, is commented out with the # symbol.

More .yaml file analysis

Now that you have the .yaml file (produced from the .bag file), here are some demonstrations of how to scan it for a list of keys or text strings you are interested in to see if certain data is present.

Note that the following examples require the ripgrep (rg) command, which is like grep only waaay faster. See here: https://github.com/BurntSushi/ripgrep#installation. Install ripgrep (rg) with: 

sudo apt update && sudo apt install ripgrep

Example 1:

Let's say you want to find a list of all key entries which begin with "piksi_". You can do so by searching for the string "key: "piksi_", as follows. 

time rg 'key: "piksi_' topics.yaml | sort -V | awk '!seen[$0]++'

The rg 'key: "piksi_' topics.yaml part searches the "topics.yaml" text file for the 'key: "piksi_' string, the sort -V part sorts all of the output lines, and the awk '!seen[$0]++' part removes duplicate entries, so that you only see one line of each match.

Command and output: 

$ time rg 'key: "piksi_' topics.yaml | sort -V | awk '!seen[$0]++'
        key: "piksi_llh_diag: Frequency Status"
        key: "piksi_rtk_diag: Frequency Status"
        key: "piksi_rtk_diag: Piksi Status"
        key: "piksi_time_diag: Frequency Status"

real    0m0.009s
user    0m0.003s
sys     0m0.011s

Example 2:

Let's search for all key data entries which begin with "GPS", "Duration", or "Minimum". "OR" type regular expression searches take the general format: (str1|str2|str3|etc), with | being read as "or" in this case. So, searching for "GPS", "Duration", or "Minimum" can be done with the following regular expression search string: '(key: "GPS|key: "Duration|key: "Minimum)'.

Here is the command to run: 

time rg '(key: "GPS|key: "Duration|key: "Minimum)' topics.yaml | sort -V | awk '!seen[$0]++'

...and here is the full output: 

$ time rg '(key: "GPS|key: "Duration|key: "Minimum)' topics.yaml | sort -V | awk '!seen[$0]++'
        key: "Duration of window (s)"
        key: "GPS RTK height difference (m)"
        key: "GPS RTK horizontal accuracy (m)"
        key: "GPS RTK meters east"
        key: "GPS RTK meters north"
        key: "GPS RTK orientation east"
        key: "GPS RTK orientation north"
        key: "GPS RTK orientation up"
        key: "GPS RTK solution status (4 = good)"
        key: "GPS RTK velocity east"
        key: "GPS RTK velocity flags"
        key: "GPS RTK velocity north"
        key: "GPS RTK velocity up"
        key: "GPS altitude"
        key: "GPS latitude"
        key: "GPS lat/lon horizontal accuracy (m)"
        key: "GPS lat/lon solution status"
        key: "GPS longitude"
        key: "Minimum acceptable frequency (Hz)"

real    0m0.023s
user    0m0.023s
sys     0m0.008s

Why use `ros_readbagfile` for this purpose instead of `rostopic echo -b`?

Answer:

  1. Because rostopic is extremley slow! This command, running on a fast computer (4-core/8-thread Pentium i7 w/m.2 SSD), for instance, takes 11.5 minutes to read an 18 GB bag file! 

    time rostopic echo -b large_bag_file.bag /topic1

    The ros_readbagfile script, however, takes only 1 min 37 sec on the same computer to read the same topic from the same 18 GB bag file! Therefore, ros_readbagfile is 11.5/(1+37/60) = ~7x faster! 

    time ros_readbagfile large_bag_file.bag /topic1

  2. Because rostopic can only read 1 single topic at a time, whereas ros_readbagfile can read any number of topics at once! 

    ros_readbagfile <mybagfile.bag> [topic1] [topic2] [topic3] [...] [topic1000]

END.

Wiki: rosbag/Tutorials/reading msgs from a bag file (last edited 2022-07-05 00:54:41 by Gabriel Staples)

Except where otherwise noted, the ROS wiki is licensed under the
Creative Commons Attribution 3.0