Python Remote Server for Robot Framework

Robot Framework remote servers allow hosting test libraries on different processes or machines than Robot Framework itself is running on. This project implements a generic remote server using the Python programming language. See the remote library interface documentation for more information about the remote interface in general as well as for a list of remote server implementations in other programming languages.

This project is hosted on GitHub and downloads are available on PyPI.

• Supported Python versions
• Supported library APIs
• Installation
• Remote server configuration
• Starting remote server
• Starting server on background
• Getting active server port
• Stopping remote server
• Testing is server running
• Example

Supported Python versions

This remote server is implemented with Python and supports also Jython (JVM), IronPython (.NET) and PyPy. Remote server version 1.1 and newer support Python 2.6, 2.7, 3.3, and newer. Remote server 1.0 series supports Python versions 2.2-2.7.

Supported library APIs

Starting from the remote server version 1.1, Robot Framework's static, hybrid and dynamic library APIs are all supported. This includes setting custom name and tags for keywords using the robot.api.deco.keyword decorator, although the support for tags requires using Robot Framework 3.0.2 or newer. Earlier remote server versions support only the static and hybrid APIs and do not support the keyword decorator at all.

For most parts these APIs work exactly like when using with Robot Framework normally. There main limitation is that logging using robot.api.logger or Python's logging module is currently not supported.

Installation

The easiest installation approach is using pip:

pip install robotremoteserver

Alternatively you can download the source distribution from PyPI, extract it and install the remote server using:

python setup.py install

Remote server configuration

The remote server is implemented as a class RobotRemoteServer and it accepts the following configuration parameters when it is initialized:

Argument	Default	Explanation
library		Test library instance or module to host. Mandatory argument.
host	'127.0.0.1'	Address to listen. Use '0.0.0.0' to listen to all available interfaces.
port	8270	Port to listen. Use 0 to select a free port automatically. Can be given as an integer or as a string. The default port 8270 is registered by IANA for remote server usage.
port_file	None	File to write the port that is used. None (default) means no such file is written.
allow_stop	'DEPRECATED'	Deprecated since version 1.1. Use allow_remote_stop instead.
serve	True	If True, start the server automatically and wait for it to be stopped. If False, server can be started using the serve method. New in version 1.1.
allow_remote_stop	True	Allow/disallow stopping the server remotely using Stop Remote Server keyword and stop_remote_server XML-RPC method. New in version 1.1.

Starting remote server

The remote server can be started simply by creating an instance of the server and passing a test library instance or module to it:

from robotremoteserver import RobotRemoteServer
from mylibrary import MyLibrary

RobotRemoteServer(MyLibrary())

By default the server listens to address 127.0.0.1 and port 8270. As discussed above, the remote server accepts various configuration parameters. Some of them are used by this example:

from robotremoteserver import RobotRemoteServer
from examplelibrary import ExampleLibrary

RobotRemoteServer(ExampleLibrary(), host='10.0.0.42', port=0,
                  port_file='/tmp/remote-port.txt')

Starting from version 1.1, the server can be initialized without starting it by using the argument serve=False. The server can then started afterwards by calling its serve method explicitly. This example is functionally equivalent to the example above:

from robotremoteserver import RobotRemoteServer
from examplelibrary import ExampleLibrary

server = RobotRemoteServer(ExampleLibrary(), host='10.0.0.42', port=0,
                           port_file='/tmp/remote-port.txt', serve=False)
server.serve()

Starting server on background

The main benefit of separately initializing and starting the server is that it makes it easier to start the server in a background thread. Servers started in a thread work exactly like servers running in the main tread except that stopping the server gracefully using Ctrl-C or signals is not supported automatically. Users must thus register signal handlers separately if needed.

Also this following example is functionally nearly equivalent to the earlier examples except. The main difference is that not all same signals are handled.

import signal
import threading
from examplelibrary import ExampleLibrary
from robotremoteserver import RobotRemoteServer

server = RobotRemoteServer(ExampleLibrary(), port=0, serve=False)
signal.signal(signal.SIGINT, lambda signum, frame: server.stop())
server_thread = threading.Thread(target=server.serve)
server_thread.start()
while server_thread.is_alive():
    server_thread.join(0.1)

Getting active server port

If the server uses the default port 8270 or some other port is given explicitly when configuring the server, you obviously know which port to use when connecting the server. When using the port 0, the server selects a free port automatically, but there are various ways how to find out the actual port:

• Address and port that are used are printed into the console where the server is started.
• If port_file argument is used, the server writes the port into the specified file where other tools can easily read it. Starting from the remote server version 1.1, the server removes the port file automatically when the server is stopped.
• Starting from the version 1.1, the server has activate method that can be called to activate the server without starting it. This method returns the port that the server binds and also sets it available via the attributes discussed below.
• A started or actived server instance has server_address attribute that contains the address and the port as a tuple. Starting from the version 1.1 there is also server_port attribute that contains just the port as an integer.

Stopping remote server

The remote server can be gracefully stopped using several different methods:

• Hitting Ctrl-C on the console where the server is running. Not supported automatically if the server is started on a background thread.
• Sending the process SIGINT, SIGTERM, or SIGHUP signal. Does not work on Windows and not supported if the server is started on a background thread.
• Using Stop Remote Server keyword. Can be disabled by using allow_remote_stop=False when initializing the server.
• Using stop_remote_server function in the XML-RPC interface. Can be disabled with the allow_remote_stop=False initialization parameter.
• Running python -m robotremoteserver stop [uri] which uses the aforementioned stop_remote_server XML-RPC function internally. Can be disabled with the allow_remote_stop=False initialization parameter.
• Using the stop_remote_server function provided by the robotremoteserver module similarly as when testing is server running. Uses the stop_remote_server XML-RPC function internally and can be disabled with the allow_remote_stop=False initialization parameter.
• Calling the stop method of the running server instance. Mainly useful when running the server on background.

Testing is server running

Starting from the version 1.0.1, the robotremoteserver module supports testing is a remote server running. This can be accomplished by running the module as a script with test argument and an optional URI:

$ python -m robotremoteserver test
Remote server running at http://127.0.0.1:8270.
$ python -m robotremoteserver test http://10.0.0.42:57347
No remote server running at http://10.0.0.42:57347.

Starting from the version 1.1, the robotremoteserver module contains function test_remote_server that can be used programmatically:

from robotremoteserver import test_remote_server

if test_remote_server('http://localhost:8270'):
    print('Remote server running!')

The robotremoteserver module can be also used to stop a remote server by using stop argument on the command line or by using the stop_remote_server function programmatically. Testing and stopping should work also with other Robot Framework remote server implementations.

Example

The remote server project contains an example that can be studied and also executed once the library is installed. You can get the example by cloning the project on GitHub, and it is also included in the source distribution available on PyPI.