grepros

grep for ROS bag files and live topics: read, filter, export.

Searches through ROS messages and matches any message field value by regular expression patterns or plain text, regardless of field type. Can also look for specific values in specific message fields only.

By default, matches are printed to console. Additionally, matches can be written to a bagfile or HTML/CSV/MCAP/Parquet/Postgres/SQL/SQLite, or published to live topics.

Supports both ROS1 and ROS2. ROS environment variables need to be set, at least ROS_VERSION.

Supported bag formats: .bag (ROS1), .db3 (ROS2), .mcap (ROS1, ROS2).

In ROS1, messages can be grepped even if Python packages for message types are not installed. Using ROS1 live topics requires ROS master to be running.

Using ROS2 requires Python packages for message types to be available in path.

Supports loading custom plugins, mainly for additional output formats.

Usable as a Python library, see LIBRARY.md. Full API documentation available at https://suurjaak.github.io/grepros.

• Example usage
• Installation
• Inputs
• Outputs
• Matching and filtering
• Plugins
• Notes on ROS1 vs ROS2
• All command-line arguments
• Dependencies
• Attribution
• License

Example usage

Search for "my text" in all bags under current directory and subdirectories:

grepros -r "my text"

Print 30 lines of the first message from each live ROS topic:

grepros --max-per-topic 1 --lines-per-message 30 --live

Find first message containing "future" (case-sensitive) in my.bag:

grepros future -I --max-count 1 --name my.bag

Find 10 messages, from geometry_msgs package, in "map" frame, from bags in current directory, reindexing any unindexed bags:

grepros frame_id=map --type geometry_msgs/* --max-count 10 --reindex-if-unindexed

Pipe all diagnostics messages with "CPU usage" from live ROS topics to my.bag:

grepros "CPU usage" --type *DiagnosticArray --no-console-output --write my.bag

Find messages with field "key" containing "0xA002", in topics ending with "diagnostics", in bags under "/tmp":

grepros key=0xA002 --topic *diagnostics --path /tmp

Find diagnostics_msgs messages in bags in current directory, containing "navigation" in fields "name" or "message", print only header stamp and values:

grepros --type diagnostic_msgs/* --select-field name message \
        --emit-field header.stamp status.values -- navigation

Print first message from each lidar topic on ROS1 host 1.2.3.4, without highlight:

ROS_MASTER_URI=http://1.2.3.4::11311 \
grepros --live --topic *lidar* --max-per-topic 1 --no-highlight

Export all bag messages to SQLite and Postgres, print only export progress:

grepros -n my.bag --write my.bag.sqlite --no-console-output --no-verbose --progress

grepros -n my.bag --write postgresql://user@host/dbname \
        --no-console-output --no-verbose --progress

Patterns use Python regular expression syntax, message matches if all match. '*' wildcards use simple globbing as zero or more characters, target matches if any value matches.

Note that some expressions may need to be quoted to avoid shell auto-unescaping or auto-expanding them, e.g. linear.x=2.?5 should be given as "linear.x=2\.?5".

Care must also be taken with unquoted wildcards, as they will auto-expanded by shell if they happen to match paths on disk.

Installation

grepros is written in Python, supporting both Python 2 and Python 3.

Developed and tested under ROS1 Noetic and ROS2 Foxy, works in later ROS2 versions (Galactic, Humble, Iron, and likewise Rolling); also works in the earlier ROS1 version Melodic (lacking MCAP and Parquet support).

Using pip

pip install grepros

This will add the grepros command to path.

Requires ROS Python packages (ROS1: rospy, roslib, rosbag, genpy; ROS2: rclpy, rosidl_parser, rosidl_runtime_py, builtin_interfaces).

For ROS1, if only using bag files and no live topics, minimal ROS1 Python packages can also be installed separately with:

pip install rospy rosbag roslib roslz4 \
--extra-index-url https://rospypi.github.io/simple/

Using apt

If ROS apt repository has been added to system:

sudo apt install ros-noetic-grepros  # ROS1 Noetic

sudo apt install ros-foxy-grepros    # ROS2 Foxy

This will add the grepros command to the global ROS1 / ROS2 environment.

Using catkin

In a ROS1 workspace, under the source directory:

git clone https://github.com/suurjaak/grepros.git
cd grepros
catkin build --this

This will add the grepros command to the local ROS1 workspace path.

Using colcon

In a ROS2 workspace, at the workspace root:

git clone https://github.com/suurjaak/grepros.git src/grepros
colcon build --packages-select grepros

This will add the grepros command to the local ROS2 workspace path.

Inputs

Input is either from one or more ROS bag files (default), or from live ROS topics.

bag

Read messages from ROS bag files, by default all in current directory.

For reading bags in MCAP format, see the MCAP plugin.

Recurse into subdirectories when looking for bagfiles:

-r
--recursive

Read specific filenames (supports * wildcards):

--n        /tmp/*.bag
--filename my.bag 2021-11-*.bag

Scan specific paths instead of current directory (supports * wildcards):

-p     /home/bags/2021-11-*
--path my/dir

Reindex unindexed ROS1 bags before processing (note: creates backup copies of files, into same directory as file):

--reindex-if-unindexed
--reindex-if-unindexed --progress

Decompress archived ROS bags before processing (.zst .zstd extensions, requires zstandard Python package) (note: unpacks archived file to disk, into same directory as file):

--decompress
--decompress --progress

Order bag messages first by topic or type, and only then by time:

--order-bag-by topic
--order-bag-by type

live

--live

Read messages from live ROS topics instead of bagfiles.

Requires ROS_MASTER_URI and ROS_ROOT to be set in environment if ROS1.

Set custom queue size for subscribing (default 10):

--queue-size-in 100

Use ROS time instead of system time for incoming message timestamps:

--ros-time-in

Outputs

There can be any number of outputs: printing to console (default), publishing to live ROS topics, or writing to file or database.

console

Default output is to console, in ANSI colors, mimicking grep output.

Disable printing messages to console:

--no-console-output

Manage color output:

--color always  (default)
--color auto    (auto-detect terminal support)
--color never   (disable colors)

Note that when paging color output with more or less, the pager needs to accept raw control characters (more -f or less -R).

bag

--write path/to/my.bag [format=bag] [overwrite=true|false]
        [rollover-size=NUM] [rollover-count=NUM] [rollover-duration=NUM]
        [rollover-template=STR]

Write messages to a ROS bag file, the custom .bag format in ROS1, or the .db3 SQLite database format in ROS2. If the bagfile already exists, it is appended to, unless specified to overwrite.

Specifying format=bag is not required if the filename ends with .bag in ROS1 or .db3 in ROS2.

For writing bags in MCAP format, see the MCAP plugin.

More on rollover.

live

--publish

Publish messages to live ROS topics. Topic prefix and suffix can be changed, or topic name set to one specific name:

--publish-prefix  /myroot
--publish-suffix  /myend
--publish-fixname /my/singular/name

One of the above arguments needs to be specified if publishing to live ROS topics while grepping from live ROS topics, to avoid endless loops.

Set custom queue size for publishers (default 10):

--queue-size-out 100

csv

--write path/to/my.csv [format=csv] [overwrite=true|false]

Write messages to CSV files, each topic to a separate file, named path/to/my.full__topic__name.csv for /full/topic/name.

html

--write path/to/my.html [format=html] [overwrite=true|false]
        [template=/path/to/html.template]
        [rollover-size=NUM] [rollover-count=NUM] [rollover-duration=NUM]
        [rollover-template=STR]

Write messages to an HTML file, with a linked table of contents, message timeline, message type definitions, and a topically traversable message list.

More on rollover.

postgres

--write postgresql://username@host/dbname [format=postgres]
        [commit-interval=NUM] [nesting=array|all]
        [dialect-file=path/to/dialects.yaml]

Write messages to a Postgres database, with tables pkg/MsgType for each ROS message type, and views /full/topic/name for each topic.

sqlite

--write path/to/my.sqlite [format=sqlite] [overwrite=true|false]
        [commit-interval=NUM] [message-yaml=true|false] [nesting=array|all]
        [dialect-file=path/to/dialects.yaml]
        [rollover-size=NUM] [rollover-count=NUM] [rollover-duration=NUM]
        [rollover-template=STR]

Write an SQLite database with tables pkg/MsgType for each ROS message type and nested type, and views /full/topic/name for each topic.

More on outputs in doc/DETAIL.md.

console / html message formatting

Set maximum number of lines to output per message:

--lines-per-message 5

Set maximum number of lines to output per message field:

--lines-per-field 2

Start message output from, or stop output at, message line number:

--start-line  2   # (1-based if positive
--end-line   -2   # (count back from total if negative)

Output only the fields where patterns find a match:

--matched-fields-only

Output only matched fields and specified number of lines around match:

--lines-around-match 5

Output only specific message fields (supports nested.paths and * wildcards):

--emit-field *data

Skip outputting specific message fields (supports nested.paths and * wildcards):

--no-emit-field header.stamp

Wrap matches in custom texts:

--match-wrapper @@@
--match-wrapper "<<<<" ">>>>"

Set custom width for wrapping message YAML printed to console (auto-detected from terminal by default):

--wrap-width 120

Matching and filtering

Any number of patterns can be specified, message matches if all patterns find a match. If no patterns are given, any message matches.

Match messages containing all of the words:

cpu memory speed

Match messages where frame_id contains "world":

frame_id=world

Match messages where header.frame_id is present:

header.frame_id=.*

Match as plaintext, not Python regular expression patterns:

-F
--fixed-strings

Select non-matching messages instead:

-v
--invert-match

Use case-sensitive matching in patterns (default is insensitive):

-I
--no-ignore-case

Limits

Stop after matching a specified number of messages (per each file if bag input):

-m          100
--max-count 100

Read only a specified number of topics (per each file if bag input):

--max-topics 10

Emit a specified number of matches per topic (per each file if bag input):

--max-per-topic 20

Emit every Nth match in topic:

--every-nth-match 10  # (skips 9 matches in topic after each match emitted)

Filtering

Read specific topics only (supports * wildcards):

-t      *lidar* *ins*
--topic /robot/sensors/*

Skip specific topics (supports * wildcards):

-nt        *lidar* *ins*
--no-topic /robot/sensors/*

Read specific message types only (supports * wildcards):

-d     *Twist*
--type sensor_msgs/*

Skip specific message types from reading (supports * wildcards):

-nd       *Twist*
--no-type sensor_msgs/*

Set specific message fields to scan (supports nested.paths and * wildcards):

-sf            twist.linear
--select-field *data

Skip specific message fields in scan (supports nested.paths and * wildcards):

-ns               twist.linear
--no-select-field *data

Only emit matches that are unique in topic, taking --select-field and --no-select-field into account (per each file if bag input):

--unique-only

Start reading from a specific timestamp:

-t0          2021-11     # (using partial ISO datetime)
--start-time 1636900000  # (using UNIX timestamp)
--start-time +100        # (seconds from bag start time, or from script startup time if live input)
--start-time -100        # (seconds from bag end time, or script startup time if live input)

Stop reading at a specific timestamp:

-t1        2021-11     # (using partial ISO datetime)
--end-time 1636900000  # (using UNIX timestamp)
--end-time +100        # (seconds from bag start time, or from script startup time if live input)
--end-time -100        # (seconds from bag end time, or from script startup time if live input)

Start reading from a specific message index in topic:

-n0           -100  # (counts back from topic total message count in bag)
--start-index   10  # (1-based index)

Stop reading at a specific message index in topic:

-n1         -100  # (counts back from topic total message count in bag)
--end-index   10  # (1-based index)

Read every Nth message in topic:

--every-nth-message 10  # (skips 9 messages in topic with each step)

Read messages in topic with timestamps at least N seconds apart:

--every-nth-interval 5  # (samples topic messages no more often than every 5 seconds)

Conditions

--condition "PYTHON EXPRESSION"

Specify one or more Python expressions that must evaluate as true to search encountered messages. Expressions can access topics, by name or * wildcard, and refer to message fields directly.

# (Match while last message in '/robot/enabled' has data=true)
--condition "<topic /robot/enabled>.data"

# (Match if at least 10 messages have been encountered in /robot/alerts)
--condition "len(<topic /robot/alerts>) > 10"

# (Match if last two messages in /robot/mode have equal .value)
--condition "<topic /robot/mode>[-2].value == <topic /robot/mode>[-1].value"

# (Match while control is enabled and robot is moving straight and level)
--condition "<topic */control_enable>.data and <topic */cmd_vel>.linear.x > 0 " \
            "and <topic */cmd_vel>.angular.z < 0.02"

Condition namespace:

Condition is automatically false if trying to access attributes of a message not yet received.

Plugins

--plugin some.python.module some.other.module.Class

Load one or more Python modules or classes as plugins.

Specifying --plugin someplugin and --help will include plugin options in printed help.

Built-in plugins:

embag

--plugin grepros.plugins.embag

Use the embag library for reading ROS1 bags.

Significantly faster, but library tends to be unstable.

mcap

--plugin grepros.plugins.mcap

Read or write messages in MCAP format.

parquet

--plugin grepros.plugins.parquet \
--write path/to/my.parquet [format=parquet] [overwrite=true|false] \
        [column-name=rostype:value] [type-rostype=arrowtype] \
        [idgenerator=callable] [nesting=array|all] [writer-argname=argvalue]

Write messages to Apache Parquet files (columnar storage format, version 2.6), each message type to a separate file.

sql

--plugin grepros.plugins.sql \
--write path/to/my.sql [format=sql] [overwrite=true|false] \
        [nesting=array|all] [dialect=clickhouse|postgres|sqlite] \
        [dialect-file=path/to/dialects.yaml]

Write SQL schema to output file, CREATE TABLE for each message type and CREATE VIEW for each topic.

More on plugins in doc/DETAIL.md.

Notes on ROS1 vs ROS2

In ROS1, message type packages do not need to be installed locally to be able to read messages from bags or live topics, as bags and topic publishers provide message type definition texts, and message classes can be generated at run-time from the type definition text. This is what rosbag does automatically, and so does grepros.

Additionally, each ROS1 message type has a hash code computed from its type definition text, available both in live topic metadata, and bag metadata. The message type definition hash code allows to recognize changes in message type packages and use the correct version of the message type.

ROS2 does not provide message type definitions, neither in the .db3 bagfiles nor in live topics. Due to this, the message type packages always need to be installed. Also, ROS2 does not provide options for generating type classes at run-time, and it does not have the concept of a message type hash.

These are serious limitations in ROS2 compared to ROS1, at least with versions up to ROS2 Humble and counting, and require extra work to smooth over. Without knowing which version of a message type package a bag was recorded with, reading bag messages with changed definitions can result in undefined behaviour.

If the serialized message structure happens to match (e.g. a change swapped the order of two int32 fields), messages will probably be deserialized seemingly successfully but with invalid content. If the serialized structure does not match, the result is a run-time error.

Because of this, it is prudent to always include a snapshot archive of used message type packages, when recording ROS2 bags.

grepros does provide the message type hash itself in ROS2 exports, by calculating the ROS2 message type hash on its own from the locally installed type definition.

The situation in ROS2 with the newer MCAP format is a bit better: at least parsed message data can be read from MCAP bags without needing the specific message packages installed. However, reading from MCAP bags yields only data structs, not usable as ROS messages e.g. for publishing to live topics. grepros tries to smooth over this difference by defaulting to locally installed message classes if available, with definitions matching message types in bag.

All command-line arguments

positional arguments:
  PATTERN               pattern(s) to find in message field values,
                        all messages match if not given,
                        can specify message field as NAME=PATTERN
                        (supports nested.paths and * wildcards)

optional arguments:
  -h, --help            show this help message and exit
  -F, --fixed-strings   PATTERNs are ordinary strings, not regular expressions
  -I, --no-ignore-case  use case-sensitive matching in PATTERNs
  -v, --invert-match    select messages not matching PATTERN
  --version             display version information and exit
  --live                read messages from live ROS topics instead of bagfiles
  --publish             publish matched messages to live ROS topics
  --write TARGET [format=bag|csv|html|mcap|parquet|postgres|sql|sqlite] [KEY=VALUE ...]
                        write matched messages to specified output,
                        format is autodetected from TARGET if not specified.
                        Bag or database will be appended to if it already exists.
                        Keyword arguments are given to output writer.
                          column-NAME=ROSTYPE:VALUE
                                                   additional column to add in Parquet output,
                                                   like column-bag_hash=string:26dfba2c
                          commit-interval=NUM      transaction size for Postgres/SQLite output
                                                   (default 1000, 0 is autocommit)
                          dialect-file=path/to/dialects.yaml
                                                   load additional SQL dialect options
                                                   for Postgres/SQL/SQLite output
                                                   from a YAML or JSON file
                          dialect=clickhouse|postgres|sqlite
                                                   use specified SQL dialect in SQL output
                                                   (default "sqlite")
                          idgenerator=CALLABLE     callable or iterable for producing message IDs 
                                                   in Parquet output, like 'uuid.uuid4' or 'itertools.count()';
                                                   nesting uses UUID values by default
                          message-yaml=true|false  whether to populate table field messages.yaml
                                                   in SQLite output (default true)
                          nesting=array|all        create tables for nested message types
                                                   in Parquet/Postgres/SQL/SQLite output,
                                                   only for arrays if "array" 
                                                   else for any nested types
                                                   (array fields in parent will be populated 
                                                    with foreign keys instead of messages as JSON)
                          overwrite=true|false     overwrite existing file 
                                                   in bag/CSV/HTML/MCAP/Parquet/SQL/SQLite output
                                                   instead of appending to if bag or database
                                                   or appending unique counter to file name
                                                   (default false)
                          rollover-size=NUM        size limit for individual files
                                                   in bag/HTML/MCAP/SQLite output
                                                   as bytes (supports abbreviations like 1K or 2M or 3G)
                          rollover-count=NUM       message limit for individual files
                                                   in bag/HTML/MCAP/SQLite output
                                                   (supports abbreviations like 1K or 2M or 3G)
                          rollover-duration=INTERVAL
                                                   message time span limit for individual files
                                                   in bag/HTML/MCAP/SQLite output
                                                   as seconds (supports abbreviations like 60m or 2h or 1d)
                          rollover-template=STR    output filename template for individual files
                                                   in bag/HTML/MCAP/SQLite output,
                                                   supporting strftime format codes like "%H-%M-%S"
                                                   and "%(index)s" as output file index
                          template=/my/path.tpl    custom template to use for HTML output
                          type-ROSTYPE=ARROWTYPE   custom mapping between ROS and pyarrow type
                                                   for Parquet output, like type-time="timestamp('ns')"
                                                   or type-uint8[]="list(uint8())"
                          writer-ARGNAME=ARGVALUE  additional arguments for Parquet output
                                                   given to pyarrow.parquet.ParquetWriter
  --plugin PLUGIN [PLUGIN ...]
                        load a Python module or class as plugin
                        (built-in plugins: grepros.plugins.embag, 
                         grepros.plugins.mcap, grepros.plugins.parquet, 
                         grepros.plugins.sql)
  --stop-on-error       stop further execution on any error like unknown message type

Filtering:
  -t TOPIC [TOPIC ...], --topic TOPIC [TOPIC ...]
                        ROS topics to read if not all (supports * wildcards)
  -nt TOPIC [TOPIC ...], --no-topic TOPIC [TOPIC ...]
                        ROS topics to skip (supports * wildcards)
  -d TYPE [TYPE ...], --type TYPE [TYPE ...]
                        ROS message types to read if not all (supports * wildcards)
  -nd TYPE [TYPE ...], --no-type TYPE [TYPE ...]
                        ROS message types to skip (supports * wildcards)
  --condition CONDITION [CONDITION ...]
                        extra conditions to require for matching messages,
                        as ordinary Python expressions, can refer to last messages
                        in topics as <topic /my/topic>; topic name can contain wildcards.
                        E.g. --condition "<topic /robot/enabled>.data" matches
                        messages only while last message in '/robot/enabled' has data=true.
  -t0 TIME, --start-time TIME
                        earliest timestamp of messages to read
                        as relative seconds if signed,
                        or epoch timestamp or ISO datetime
                        (for bag input, relative to bag start time
                        if positive or end time if negative,
                        for live input relative to system time,
                        datetime may be partial like 2021-10-14T12)
  -t1 TIME, --end-time TIME
                        latest timestamp of messages to read
                        as relative seconds if signed,
                        or epoch timestamp or ISO datetime
                        (for bag input, relative to bag start time
                        if positive or end time if negative,
                        for live input relative to system time,
                        datetime may be partial like 2021-10-14T12)
  -n0 INDEX, --start-index INDEX
                        message index within topic to start from
                        (1-based if positive, counts back from bag total if negative)
  -n1 INDEX, --end-index INDEX
                        message index within topic to stop at
                        (1-based if positive, counts back from bag total if negative)
  --every-nth-message NUM
                        read every Nth message within topic
  --every-nth-interval SECONDS
                        read messages at least N seconds apart within topic
  --every-nth-match NUM
                        emit every Nth match in topic
  -sf FIELD [FIELD ...], --select-field FIELD [FIELD ...]
                        message fields to use in matching if not all
                        (supports nested.paths and * wildcards)
  -ns FIELD [FIELD ...], --no-select-field FIELD [FIELD ...]
                        message fields to skip in matching
                        (supports nested.paths and * wildcards)
  -m NUM, --max-count NUM
                        number of matched messages to emit (per each file if bag input)
  --max-per-topic NUM   number of matched messages to emit from each topic
                        (per each file if bag input)
  --max-topics NUM      number of topics to emit matches from (per each file if bag input)
  --unique-only         only emit matches that are unique in topic,
                        taking --select-field and --no-select-field into account
                        (per each file if bag input)

Output control:
  -B NUM, --before-context NUM
                        emit NUM messages of leading context before match
  -A NUM, --after-context NUM
                        emit NUM messages of trailing context after match
  -C NUM, --context NUM
                        emit NUM messages of leading and trailing context
                        around match
  -ef FIELD [FIELD ...], --emit-field FIELD [FIELD ...]
                        message fields to emit in console/CSV/HTML/Parquet output if not all
                        (supports nested.paths and * wildcards)
  -nf FIELD [FIELD ...], --no-emit-field FIELD [FIELD ...]
                        message fields to skip in console/CSV/HTML/Parquet output
                        (supports nested.paths and * wildcards)
  -mo, --matched-fields-only
                        emit only the fields where PATTERNs find a match in console/HTML output
  -la NUM, --lines-around-match NUM
                        emit only matched fields and NUM message lines
                        around match in console/HTML output
  -lf NUM, --lines-per-field NUM
                        maximum number of lines to emit per field in console/HTML output
  -l0 NUM, --start-line NUM
                        message line number to start emitting from in console/HTML output
                        (1-based if positive, counts back from total if negative)
  -l1 NUM, --end-line NUM
                        message line number to stop emitting at in console/HTML output
                        (1-based if positive, counts back from total if negative)
  -lm NUM, --lines-per-message NUM
                        maximum number of lines to emit per message in console/HTML output
  --match-wrapper [STR [STR ...]]
                        string to wrap around matched values in console/HTML output,
                        both sides if one value, start and end if more than one,
                        or no wrapping if zero values
                        (default "**" in colorless output)
  --wrap-width NUM      character width to wrap message YAML console output at,
                        0 disables (defaults to detected terminal width)
  --color {auto,always,never}
                        use color output in console (default "always")
  --no-meta             do not print source and message metainfo to console
  --no-filename         do not print bag filename prefix on each console message line
  --no-highlight        do not highlight matched values
  --no-console-output   do not print matches to console
  --progress            show progress bar when not printing matches to console
  --verbose             print status messages during console output
                        for publishing and writing
  --no-verbose          do not print status messages during console output
                        for publishing and writing

Bag input control:
  -n FILE [FILE ...], --filename FILE [FILE ...]
                        names of ROS bagfiles to read if not all in directory
                        (supports * wildcards)
  -p PATH [PATH ...], --path PATH [PATH ...]
                        paths to scan if not current directory
                        (supports * wildcards)
  -r, --recursive       recurse into subdirectories when looking for bagfiles
  --order-bag-by {topic,type}
                        order bag messages by topic or type first and then by time
  --decompress          decompress archived bagfiles with recognized extensions (.zst .zstd)
  --reindex-if-unindexed
                        reindex unindexed bagfiles (ROS1 only), makes backup copies

Live topic control:
  --publish-prefix PREFIX
                        prefix to prepend to input topic name on publishing match
  --publish-suffix SUFFIX
                        suffix to append to input topic name on publishing match
  --publish-fixname TOPIC
                        single output topic name to publish all matches to,
                        overrides prefix and suffix
  --queue-size-in SIZE  live ROS topic subscriber queue size (default 10)
  --queue-size-out SIZE
                        output publisher queue size (default 10)
  --ros-time-in         use ROS time instead of system time for incoming message
                        timestamps from subsribed live ROS topics

Dependencies

grepros requires Python 3.8+ or Python 2.7, and the following 3rd-party Python packages:

• pyyaml (https://pypi.org/project/PyYAML)
• six (https://pypi.org/project/six)
• ROS1: rospy, roslib, rosbag, genpy
• ROS2: rclpy, rosidl_parser, rosidl_runtime_py, builtin_interfaces

Optional, for decompressing archived bags:

• zstandard (https://pypi.org/project/zstandard)

Optional, for faster reading of ROS1 bags:

• embag (https://github.com/embarktrucks/embag)

Optional, for Postgres output:

• psycopg2 (https://pypi.org/project/psycopg2)

Optional, for Parquet output:

• pandas (https://pypi.org/project/pandas)
• pyarrow (https://pypi.org/project/pyarrow)

Optional, for MCAP input-output:

• mcap (https://pypi.org/project/mcap)
• mcap_ros1_support (https://pypi.org/project/mcap_ros1_support)
• mcap_ros2_support (https://pypi.org/project/mcap_ros2_support)

Optional, for generating API documentation:

• doxypypy (https://pypi.org/project/doxypypy; needs latest master: pip install git+https://github.com/Feneric/doxypypy)

All dependencies other than rospy/rclpy can be installed with:

pip install pyyaml six zstandard embag psycopg2 pandas pyarrow \
            mcap mcap_ros1_support mcap_ros2_support \
            git+https://github.com/Feneric/doxypypy

Attribution

Includes a modified version of step, Simple Template Engine for Python, (c) 2012, Daniele Mazzocchio, https://github.com/dotpy/step, released under the MIT license.

License

Copyright (c) 2021 by Erki Suurjaak. Released as free open source software under the BSD License, see LICENSE.md for full details.

Changelog for package grepros

1.1.0 (2024-01-03)

• add support for splitting files in bag/HTML/MCAP/SQLite output
• add api.to_duration()
• do not use partial match for topic and typename filters without wildcards
• support patterns yielding zero-length matches like \"(?!not_this)\"
• support search over message fields in conditions as [--condition \"value in msg\"]{.title-ref}
• support giving dashed names like \"rollover-size\" in format-specific write options as underscored \"rollover_size\" instead
• disallow unknown command-line flags and partial abbreviations
• fix matching nested message values for patterns using start or end flags
• fix ISO datetime support in earliest and latest timestamp arguments
• fix --end-line not being used
• fix api.dict_to_message() erroring on temporal types in dict
• fix api.get_ros_time_category() erroring on time/duration instances in ROS2
• fix including topics in match only used for conditions not matching
• fix using len() in conditions on topics not reached yet
• ensure api.get_message_value() returning collections as lists not tuples

1.0.4 (2023-09-28)

• fix Parquet sink validation resulting in silent failure if additional columns specified

1.0.3 (2023-08-31)

• fix Unicode characters in HTML output template

1.0.2 (2023-08-30)

• fix raising error in ROS1 live topics if message class not locally available
• fix generating ROS1 message classes dynamically in multi-threaded environment
• fix caching generated message classes in ROS1
• update step, the vendored template library
• ensure compatibility with Python 3.12+

1.0.1 (2023-07-14)

• ensure Python2 compatibility under ROS1 Melodic

1.0.0 (2023-07-13)

• make grepros conveniently usable as a library
• add --no-highlight option
• add --stop-on-error option
• rename options --print-field and --no-print-field to --emit-field and --no-emit-field
• support --emit-field --no-emit-field in CSV and Parquet exports
• handle all numpy types in ROS2 messages, not only ndarrays
• avoid raising errors for unknown message types in ROS2 bags if not reading those topics
• fix grepping and emitting specific messages fields only
• fix not skipping live topics published by grepros itself in ROS2
• fix embag reader
• verify output targets being writable on startup
• provide connection header in writing ROS1 bag if topic has multiple types
• smooth over rosbag bug of ignoring topic and time filters in format v1.2
• print ROS1 master URI in verbose mode on connecting to live topics
• raise error on loading Parquet plugin if libraries unavailable
• support \"postgres://\" as auto-detected Postgres target in addition to \"postgresql://\"
• use bagfile format as last when auto-detecting output format
• add MCAP bag interface
• auto-detect MCAP output by file extension
• add inputs.AppSource and outputs.AppSink
• add api.deserialize_message() dict_to_message() make_full_typename() time_message() to_time()
• rename api.get_message_data() to serialize_message()
• fix api.message_to_dict() giving invalid names for temporal types in ROS2

0.6.0 (2023-03-27)

• add nesting=array|all to --write Parquet options
• add idgenerator=callable to --write Parquet options
• add api.canonical()
• match bounded array fields to configured output types properly in Parquet/Postgres/SQL/SQLite output, like \"uint8[10]\" for \"BYTEA\" in Postgres
• workaround for ROS1 time/duration fields defined as int32 while actually being uint32
• fix date formatting in HTML output

0.5.0 (2022-10-18)

• add --plugin grepros.plugins.mcap (MCAP input and output)
• refactor internal bag API
• fix message type definition parsing yielding duplicate subtypes
• fix error in example usage text

0.4.7 (2022-06-20)

• fix space leak in caching message metadata

0.4.6 (2022-05-26)

• add forgotten implementation for --every-nth-match
• fix --every-nth-message
• fix error on grepping bags where no topic or type name matches given filter

0.4.5 (2022-04-19)

• fix forcing all numeric array fields to integer lists regardless of type
• fix error on subscribing to defunct topic

0.4.4 (2022-03-16)

• add support for reading zstd-compressed bagfiles
• use message type definition from ROS1 live topics instead of locally installed package
• optimize partial printing of very long array fields
• optimize CSV output of very long array fields
• strip leading \"./\" from printed filename prefix if grepping working directory
• fix not skipping ROS2 bag if all topics filtered out
• fix making compatible QoS for ROS2 topic subscriptions
• fix making unique filename on error in HTML output

0.4.3 (2022-03-01)

• continue subscribing to other live topics even if one causes error
• continue reading from ROS2 bag even if one message type causes error
• subscribe to live topics in ROS2 with QoS matching publisher
• fix not saving publisher QoS profiles in written ROS2 bags
• fix package build not including submodules (#1)

0.4.2 (2022-02-09)

• add overwrite=true|false to --write options
• add column-name=rostype:value to --write Parquet options
• add scripts/generate_msgs.py
• create message type definition from .idl if .msg file not available in ROS2
• improve DDS type parsing in ROS2
• handle ROS2 char and byte int8/uint8 reversal vs ROS1
• fix assembling message type full definition in ROS2
• fix processing byte values in ROS2
• fix processing bounded string types in ROS2
• fix inserting chars in Postgres
• fix inserting very large integers in SQLite

0.4.1 (2022-01-08)

• refactor database and SQL sinks onto a common base
• allow specifying dialect options in Postgres/SQLite output
• allow overriding table and view names in SQL dialects
• allow specifying field value adapters in SQL dialects
• allow specifying structured type mappings in Parquet output like type-uint8[]=\"list(uint8())\"
• drop meta-table from Postgres output
• drop default value from --publish-prefix option
• fix error in parsing subtypes from message definitions
• fix converting ROS2 temporal messages to seconds/nanoseconds

0.4.0 (2021-12-26)

• add --plugin grepros.plugins.parquet (Parquet output)
• add --plugin grepros.plugins.sql (SQL schema output)
• add --plugin grepros.plugins.embag (faster ROS1 bag reader)
• add --reindex-if-unindexed option
• add --every-nth-match option
• add --every-nth-message option
• add --every-nth-interval option
• allow multiple write sinks, combine --write-format and --write-option to --write
• refactor plugins interface
• populate topics.offered_qos_profiles in ROS2 bag output where possible
• fix progress bar afterword not updating when grepping multiple bags
• fix error on empty bag with no messages
• fix error in Postgres output for NaNs in nested JSON values
• fix skipping some messages in ROS1 bag for types with identical hashes
• fix not being able to specify list arguments several times
• ensure no conflicts from changed message types or identical type hashes
• add tests

0.3.5 (2021-12-14)

• fix Postgres output not having content

0.3.4 (2021-12-14)

• add --write-option message-yaml=true|false, for SQLite output
• speed up SQLite output (~4-8x)
• speed up YAML formatting (~2x)
• fix no engine name in console texts for Postgres output

0.3.3 (2021-12-13)

• fix errors in Postgres/SQLite output

0.3.2 (2021-12-12)

• rename --write-option commit_interval to commit-interval
• raise Postgres default commit-interval from 100 to 1000
• add --write-option commit-interval support to SQLite output
• add --write-option subtypes=array|all, for Postgres/SQLite output
• speed up SQLite output (~2x)
• refactor Postgres/SQLite sinks onto common base class

0.3.1 (2021-12-06)

• add --write-option template=/my/html.template, for HTML output
• add --write-option commit_interval=NUM, for Postgres output
• drop --write-format-template
• fix highlighting subtype arrays

0.3.0 (2021-12-05)

• add --write-format postgres
• add --no-verbose option
• add --condition option
• add --plugin option
• add wildcard support to fields in \"field=PATTERN\"
• use up to nanosecond precision in HTML output timeline
• highlight empty arrays on any-match regardless of type
• select meta-fields as last in SQLite topic views
• fix potential error on using --max-per-topic with live topics
• fix detecting ROS2 bags
• fix using --progress with --live

0.2.5 (2021-11-28)

• add --progress option
• match anything by default if no patterns given
• add timeline to HTML output
• auto-detect output format from given filename
• fix breaking too early on --max-per-topic
• fix adding topic views to existing SQLite output database
• fix error on adding message type tables for empty list values in SQLite output
• fix sorting table of contents in HTML output
• do not auto-expand table of contents in HTML output

0.2.4 (2021-11-18)

• skip retrieving full message counts from ROS2 bag before any match
• ensure message YAMLs in html output always in color and wrapped at 120 characters
• fix inserting duplicate types-rows when adding to an existing SQLite output file
• improve wrapping lists and nunbers

0.2.3 (2021-11-15)

• add --write-format csv
• add --write-format sqlite
• local Python packages no longer required for custom message types in ROS1
• add topic toggle checkboxes to HTML output
• add topic count to live source metainfo
• break early when max matches per topic reached
• improve HTML output

0.2.2 (2021-11-10)

• shut down ROS2 live node properly
• better support for ROS2 primitive types
• make HTML output table of contents sortable
• stop requiring unneeded environment variables

0.2.1 (2021-11-09)

• add --write-format option, with HTML support
• add --wrap-width option
• add --order-bag-by option
• handle topics with multiple message types
• improve console output wrapping
• fix detecting ROS2 primitive array types
• fix using ROS2 bag start-end timestamps

0.2.0 (2021-11-04)

• add ROS2 support
• flush stdout on every print, to avoid buffering in redirected output
• add --ros-time-in option
• add --unique-only option
• rename options --noselect-field and --noprint-field to --no-select-field and --no-print-field

0.1.0 (2021-10-31)

• grep for ROS1 bag files and live topics, able to print and publish and write bagfiles