Topic Store

Topic Store is a ROS package for storing ROS messages to a database or filesystem.

Unlike ROS bags Topic Store adds flexibility by serialising all messages into a data hierarchy that's easily searchable with database queries and allows for remote storage.

You can also use Topic Store as a standalone python package to read and write data without a ROS installation!

Installation

ROS

# From source
cd catkin_ws/src
git clone https://github.com/RaymondKirk/topic_store 
catkin build topic_store

# From apt
sudo apt install ros-melodic-topic-store  # you need to add the L-CAS ros source

Installing without ROS

You can install topic_store as an independent python2/3 package without a working ROS installation.

# From source 
git clone https://github.com/RaymondKirk/topic_store
cd topic_store/src
pip install --extra-index-url https://rospypi.github.io/simple/ -e .

# From PyPi
pip install --extra-index-url https://rospypi.github.io/simple/ topic-store

To install other dependencies i.e. ros_numpy you can run the following

pip install --extra-index-url https://rospypi.github.io/simple/ ros_numpy

Usage

Create a scenario file by following the documentation in scenarios or by following the example file default_config.yaml. The example below will save a history of all messages sent to the ros log topic.

# Save to /path/to/your/scenario/file.yaml. This collection behaviour will save your log history.
context: "save_ros_logs"

storage: 
  "method": "filesystem"
  "location": "default" 

data: { 
  "ros_msg": "/rosout", 
}

collection: 
  method: "event" 
  watch_topic: "/rosout"

If storage.method is database ensure that your database is accessible at the host/port in the config file found at storage.config. To launch the default database roslaunch topic_store start_database.launch, the default database creates/uses a Mongo 4.2 server instance in the default ${HOME}/.ros/topic_store/database folder exposed on localhost:65530 (defined in storage.config of the scenario file). Docker is a requirement to use a database backend to avoid conflicts with system mongo. A utility script is provided in scripts/install_docker.sh if you do not have it installed.

Launch your data collection scenario!

roslaunch topic_store run_scenario.launch scenario_file:="/path/to/your/scenario/file.yaml"

Examples

Database data

The below example shows how to load and use files stored in a database.

import topic_store as ts

# Read data
storage = ts.load("/path/to/scenario/file/containing/db/connection/info.yaml")
for item in storage:
    print("As Python Dict", item.dict)  # or item["key"]
    print("As ROS Msgs", item.msgs)  # or item("key")

# Write data
storage.insert_one({"important_data": "topic store is great!"})

Filesystem data

The below example shows how to load and use .topic_store files, saved from when scenarios are ran with the storage_method="filesystem" option.

import topic_store as ts

# Read data
storage = ts.load("/path/to/file.topic_store")
for item in storage:
    print("As Python Dict", item.dict)  # or item["key"]
    print("As ROS Msgs", item.msgs)  # or item("key")

# Write data
storage.insert_one({"important_data": "topic store is great!"})

Launch a database

When launching a data collection scenario where storage.method==database you must also launch the database or ensure it's already running at the uri defined in the host/port parameters of the config file located at storage.config.

To launch a database prior to running a data collection scenario.

roslaunch topic_store start_database.launch scenario_file:="/path/to/your/scenario/file.yaml"

Convert to ROS bags

Filesystem .topic_store files and database collections can be converted to ROS bags.

# Convert filesystem files to rosbags
rosrun topic_store convert.py -i input.topic_store -o output.bag

# Convert database collections to rosbags 
# Pass scenario file containing connection and collection info as the input
rosrun topic_store convert.py -i scenario_config.yaml -o output.bag

Convert between Filesystem and Database

Filesystem .topic_store files can be migrated to a MongoDB database. Pass the scenario file that contains the database connection information as the output file.

rosrun topic_store convert.py -i input.topic_store -o scenario_config.yaml

Database collections can be converted to a filesystem .topic_store file. Pass the scenario file that contains the database connection information as the input file.

rosrun topic_store convert.py -i scenario_config.yaml -o output.topic_store

Export from Mongodb query into rosbag/filesystem

Example call:

With typical mongodb URI for SSL and authentication
On database ff_rasberry and collection 2020_riseholme_framos_cameras
Specify database connection URI (-i) with db name authSource=database
Including a query (-q) for a specific document
Add projection to return only sub-documents (-p)

convert.py -i "mongodb://USER:PASS@HOST:PORT/?authSource=ff_rasberry&tls=true&tlsAllowInvalidCertificates=true" -c 2020_riseholme_framos_cameras -q '{"_id":"ObjectId(5f115ee6af915351df739757)"}' -p '{"cameras.top.color":1, "robot": 1}' -o out.bag

Storing high bandwidth image data

To store a lot of image data you will be limited to the IO of your machine. Topic Compression offers a compression solution to increase capture performance.

# Install the ROS package
cd catkin_ws/src
git clone https://github.com/RaymondKirk/topic_compression
catkin build topic_compression

# Compression/Decompression is chosen automatically so just pass the input topic name and optionally the out topic name
rosrun topic_compression run in:=/camera/colour/image_raw  # out=/camera/colour/image_raw/compressed 
rosrun topic_compression run in:=/camera/depth/image_raw  # out=/camera/depth/image_raw/compressed

Mark-Niemeyer / topic_store