A simulator for real hardware. This project provides a server able to spawn multiple simulated devices and serve requests concurrently.
This project provides only the required infrastructure to launch a server from a configuration file (YAML, TOML or json) and a means to register third-party device plugins through the python entry point mechanism.
So far, the project provides transports for TCP, UDP and serial line. Support for new transports (ex: USB, GPIB or SPI) is being implemented on a need basis.
PRs are welcome!
(TL;DR: pip install sinstruments[all]
)
From within your favorite python environment:
$ pip install sinstruments
Additionally, if you want to write YAML configuration files in YAML:
$ pip install sinstruments[yaml]
...or, for TOML based configuration:
$ pip install sinstruments[toml]
Once installed the server can be run with:
$ sinstruments-server -c <config file name>
The configuration file describes which devices the server should instantiate along with a series of options like the transport(s) each device listens for requests on.
Imagine you need to simulate 2 GE Pace 5000 reachable through a TCP port each and a CryoCon 24C accessible through the serial line.
First, make sure the dependencies are installed with:
$ pip install gepace[simulator] cryoncon[simulator]
Now we can prepare a YAML configuration called simulator.yml
:
devices:
- class: Pace
name: pace-1
transports:
- type: tcp
url: :5000
- class: Pace
name: pace-2
transports:
- type: tcp
url: :5001
- class: CryoCon
name: cryocon-1
transports:
- type: serial
url: /tmp/cryocon-1
We are now ready to launch the server:
$ sinstruments-server -c simulator.yml
That's it! You should now be able to connect to any of the Pace devices through TCP or the CryoCon using the local emulated serial line.
Let's try connecting to the first Pace with the nc (aka netcat) linux command
line tool and ask for the well known *IDN?
SCPI command:
$ nc localhost 5000
*IDN?
GE,Pace5000,204683,1.01A
This is a summary of the known third-party instrumentation libraries which provide their own simulators.
- cryocon
- fast-spinner
- gepace
- icepap
- julabo
- vacuubrand
- xia-pfcu
- Mythen detector (from Dectris) - not publicly available yet
If you wrote a publicly available device feel free complete the above list by creating a PR.
Hint: sinstruments-server ls
shows a list of available plugins.
The configuration file can be a YAML, TOML or JSON file as long as it translates to a dictionary with the description given below.
In this chapter we will use YAML as a reference example.
The file should contain at least a top-level key called devices
.
The value needs to be a list of device descriptions:
devices:
- class: Pace
name: pace-1
transports:
- type: tcp
url: :5000
Each device description must contain:
- class: Each third-party plugin should describe which text identify itself
- name: a unique name. Each device must be given a unique name at your choice
- transports: a list of transports from where the device is accessible.
Most devices provide only one transport.
- type: Each transport must define its type (supported are
tcp
,udp
,serial
) - url: the url where the device is listening on
- type: Each transport must define its type (supported are
Any other options given to each device are passed directly to the specific plugin object at runtime. Each plugin should describe which additional options it supports and how to use them.
For TCP and UDP transports, the url has the <host>:<port>
format.
An empty host (like in the above example) is interpreted as 0.0.0.0
(which
means listen on all network interfaces). If host is 127.0.0.1
or localhost
the device will only be accessible from the machine where the simulator is
running.
A port value of 0 means ask the OS to assign a free port (useful for running a test suite). Otherwise must be a valid TCP or UDP port.
The url represents a special file which is created by the simulator to
simulate a serial line accessible like a /dev/ttyS0
linux serial line file.
This feature is only available in linux and systems for which the pseudo
terminal pty
is implemented in python.
The url is optional. The simulator will always create a non deterministic
name like /dev/pts/4
and it will log this information in case you need to
access. This feature is most useful when running a test suite.
You are free to choose any url path you like (ex: /dev/ttyRP10
) as long
as you are sure the simulator has permissions to create the symbolic file.
For any of the transports (TCP, UDP and serial line) is is possible to do basic
simulation of the communication channel speed by providing an additional
baudrate
parameter to the configuration. Example:
- class: CryoCon
name: cryocon-1
transports:
- type: serial
url: /tmp/cryocon-1
baudrate: 9600
The simulator provides a gevent back door python console which you can activate if you want to access a running simulator process remotely. To activate this feature simply add to the top-level of the configuration the following:
backdoor: ["localhost": 10001]
devices:
- ...
You are free to choose any other TCP port and bind address. Be aware that this backdoor provides no authentication and makes no attempt to limit what remote users can do. Anyone that can access the server can take any action that the running python process can. Thus, while you may bind to any interface, for security purposes it is recommended that you bind to one only accessible to the local machine, e.g., 127.0.0.1/localhost.
Usage
Once the backdoor is configured and the server is running, in a another terminal, connect with:
$ nc 127.0.0.1 10001
Welcome to Simulator server console.
You can access me through the 'server()' function. Have fun!
>>> print(server())
...
Writting a new device is simple. Let's imagine you want to simulate a SCPI
oscilloscope. The only thing you need to do is write a class inheriting
from BaseDevice and implement the handle_message(self, message)
where you
should handle the different commands supported by your device:
# myproject/simulator.py
from sinstruments.simulator import BaseDevice
class Oscilloscope(BaseDevice):
def handle_message(self, message):
self._log.info("received request %r", message)
message = message.strip().decode()
if message == "*IDN?":
return b"ACME Inc,O-3000,23l032,3.5A"
elif message == "*RST":
self._log.info("Resetting myself!")
...
Don't forget to always return bytes
! The simulator doesn't make any guesses
on how to encode str
Assuming this file simulator.py
is part of a python package called myproject
,
the second thing to do is register your simulator plugin in your setup.py:
setup(
...
entry_points={
"sinstruments.device": [
"Oscilloscope=myproject.simulator:Oscilloscope"
]
}
)
You should now be able to launch your simulator by writing a configuration file:
# oscilo.yml
devices:
- class: Oscilloscope
name: oscilo-1
transports:
- type: tcp
url: :5000
Now launch the server with
$ sinstruments-server -c oscillo.yml
and you should be able to connect with:
$ nc localhost 5000
*IDN?
ACME Inc,O-3000,23l032,3.5A
By default the eol
is set to \n
. You can change it to any character with:
class Oscilloscope(BaseDevice):
newline = b"\r"
If your device implements a protocol which answers with multiple (potentially
delayed) answers to a single request, you can support this by
converting the handle_message()
into a generator:
class Oscilloscope(BaseDevice):
def handle_message(self, message):
self._log.info("received request %r", message)
message = message.strip().decode()
if message == "*IDN?":
yield b"ACME Inc,O-3000,23l032,3.5A"
elif message == "*RST":
self._log.info("Resetting myself!")
elif message == "GIVE:ME 10":
for i in range(1, 11):
yield f"Here's {i}\n".encode()
...
Don't forget to always yield bytes
! The simulator doesn't make any guesses
on how to encode str
If your simulated device requires additional configuration, it can be supplied through the same YAML file.
Let's say you want to be able to configure if your device is in CONTROL
mode
at startup. Additionally, if no initial value is configured, it should default
to 'OFF'.
First lets add this to our configuration example:
# oscilo.yml
devices:
- class: Oscilloscope
name: oscilo-1
control: ON
transports:
- type: tcp
url: :5000
Then, we re-implement our Oscilloscope __init__()
to intercept this new
parameter and we handle it in handle_message()
:
class Oscilloscope(BaseDevice):
def __init__(self, name, **opts):
self._control = opts.pop("control", "OFF").upper()
super().__init__(name, **opts)
def handle_message(self, message):
...
elif message == "CONTROL":
return f"CONTROL {self._control}\n".encode()
...
You are free to add as many options as you want as long as they don't conflict
with the reserved keys name
, class
and transports
.
Some instruments implement protocols that are not suitably managed by a EOL based message protocol.
The simulator allows you to write your own message protocol. Here is an example:
from sinstruments.simulator import MessageProtocol
class FixSizeProtocol(MessageProtocol):
Size = 32
def read_messages(self):
transport = self.transport
buff = b''
while True:
buff += transport.read(self.channel, size=4096)
if not buff:
return
for i in range(0, len(buff), self.Size):
message = buff[i:i+self.Size]
if len(message) < self.Size:
buff = message
break
yield message
class Oscilloscope(BaseDevice):
protocol = FixSizeProtocol
...
If you are developing a python library that provides access to an instrument accessible through socket or serial line and you wrote a simulator for it, you might be interested in testing your library against the simulator.
sinstruments provides a pair of pytest helpers that spawn a simulator in a separate thread.
The first usage is simply using the server_context
helper.
There is actually nothing pytest speficic about this helper so you could
imagine using it in other scenarios as well.
Here is an example:
import pytest
from sinstruments.pytest import server_context
cfg = {
"devices": [{
"name": "oscillo-1",
"class": "Oscilloscope",
"transports": [
{"type": "tcp", "url": "localhost:0"}
]
}]
}
def test_oscilloscope_id():
with server_context(cfg) as server:
# put here code to perform your tests that need to communicate with
# the simulator. In this example an oscilloscope client
addr = server.devices["oscillo-1"].transports[0].address
oscillo = Oscilloscope(addr)
assert oscillo.idn().startswith("ACME Inc,O-3000")
You might notice that in the configuration we use port 0
. This is telling
the simulator to listen on any free port provided by the OS.
The actual test retrieves the current address assigned by the OS and uses it in the test.
As you can see, the tests are not dependent of the availability of one specific port which makes them portable.
Here is a suggestion on how you could write your own fixture using the
server_context
helper. The aim was to reduce the amount of boilerplate
code you need to write your test:
@pytest.fixture
def oscillo_server():
with server_context(config) as server:
server.oscillo1 = server.devices["oscillo-1"]
server.oscillo1.addr = server.oscillo1.transports[0].address
yield server
def test_oscilloscope_current(oscillo_server):
oscillo = Oscilloscope(oscillo_server.oscillo1.addr)
assert .05 < oscillo.current() < 0.01
A second helper is the server
fixture. This fixture depends on an existing
config
feature that must be present in your module. Here is an example
following the previous code:
from sinstruments.pytest import server
@pytest.fixture
def config()
yield cfg
def test_oscilloscope_voltage(server):
addr = server.devices["oscillo-1"].transports[0].address
oscillo = Oscilloscope(addr)
assert 5 < oscillo.voltage() < 10