How to

This is a small list of how-tos specific to PyTango. A more general Tango how-to list can be found here.

How to contribute

Everyone is welcome to contribute to PyTango project. If you don’t feel comfortable with writing core PyTango we are looking for contributors to documentation or/and tests.

It refers to the next section, see How to Contribute.

Check the default TANGO host

The default TANGO host can be defined using the environment variable TANGO_HOST or in a tangorc file (see Tango environment variables for complete information)

To check what is the current value that TANGO uses for the default configuration simple do:

1>>> import tango
2>>> tango.ApiUtil.get_env_var("TANGO_HOST")
3'homer.simpson.com:10000'

Check TANGO version

There are two library versions you might be interested in checking: The PyTango version:

1>>> import tango
2>>> tango.__version__
3'9.3.4'
4>>> tango.__version_info__
5(9, 3, 4)

and the Tango C++ library version that PyTango was compiled with:

1>>> import tango
2>>> tango.constants.TgLibVers
3'9.3.4'

Start server from command line

To start server from the command line execute the following command:

$ python <server_file>.py <instance_name>
Ready to accept request

To run server without database use option -nodb.

$ python <server_file>.py <instance_name> -nodb -port 10000
Ready to accept request

Note, that to start server in this mode you should provide a port with either --post, or --ORBendPoint option

Additionally, you can use the following options:

-h, -?, --help : show usage help

-v, --verbose: set the trace level. Can be user in count way: -vvvv set level to 4 or –verbose –verbose set to 2

-vN: directly set the trace level to N, e.g. -v3 - set level to 3

--file <file_name>: start a device server using an ASCII file instead of the Tango database

--host <host_name>: force the host from which server accept requests

--port <port>: force the port on which the device server listens

--nodb: run server without DB

--dlist <dev1,dev2,etc>: the device name list. This option is supported only with the -nodb option

--ORBendPoint giop:tcp:<host>:<port>: Specifying the host from which server accept requests and port on which the device server listens.

Note: any ORB option can be provided if it starts with -ORB<option>

Additionally in Windows the following option can be used:

-i: install the service

-s: install the service and choose the automatic startup mode

-u: uninstall the service

--dbg: run in console mode to debug service. The service must have been installed prior to use it.

Note: all long-options can be provided in non-POSIX format: -port or --port etc…

Report a bug

Bugs can be reported as issues in PyTango GitLab.

It is also helpful if you can put in the issue description the PyTango information. It can be a dump of:

$ python -c "from tango.utils import info; print(info())"

Test the connection to the Device and get it’s current state

One of the most basic examples is to get a reference to a device and determine if it is running or not:

 1from tango import DeviceProxy
 2
 3# Get proxy on the tango_test1 device
 4print("Creating proxy to TangoTest device...")
 5tango_test = DeviceProxy("sys/tg_test/1")
 6
 7# ping it
 8print(tango_test.ping())
 9
10# get the state
11print(tango_test.state())

Read and write attributes

Basic read/write attribute operations:

 1from tango import DeviceProxy
 2
 3# Get proxy on the tango_test1 device
 4print("Creating proxy to TangoTest device...")
 5tango_test = DeviceProxy("sys/tg_test/1")
 6
 7# Read a scalar attribute. This will return a tango.DeviceAttribute
 8# Member 'value' contains the attribute value
 9scalar = tango_test.read_attribute("long_scalar")
10print("Long_scalar value = {0}".format(scalar.value))
11
12# PyTango provides a shorter way:
13scalar = tango_test.long_scalar
14print("Long_scalar value = {0}".format(scalar))
15
16# Read a spectrum attribute
17spectrum = tango_test.read_attribute("double_spectrum")
18# ... or, the shorter version:
19spectrum = tango_test.double_spectrum
20
21# Write a scalar attribute
22scalar_value = 18
23tango_test.write_attribute("long_scalar", scalar_value)
24
25#  PyTango provides a shorter way:
26tango_test.long_scalar = scalar_value
27
28# Write a spectrum attribute
29spectrum_value = [1.2, 3.2, 12.3]
30tango_test.write_attribute("double_spectrum", spectrum_value)
31# ... or, the shorter version:
32tango_test.double_spectrum = spectrum_value
33
34# Write an image attribute
35image_value = [ [1, 2], [3, 4] ]
36tango_test.write_attribute("long_image", image_value)
37# ... or, the shorter version:
38tango_test.long_image = image_value

Note that if PyTango is compiled with numpy support the values got when reading a spectrum or an image will be numpy arrays. This results in a faster and more memory efficient PyTango. You can also use numpy to specify the values when writing attributes, especially if you know the exact attribute type:

 1import numpy
 2from tango import DeviceProxy
 3
 4# Get proxy on the tango_test1 device
 5print("Creating proxy to TangoTest device...")
 6tango_test = DeviceProxy("sys/tg_test/1")
 7
 8data_1d_long = numpy.arange(0, 100, dtype=numpy.int32)
 9
10tango_test.long_spectrum = data_1d_long
11
12data_2d_float = numpy.zeros((10,20), dtype=numpy.float64)
13
14tango_test.double_image = data_2d_float

Execute commands

As you can see in the following example, when scalar types are used, the Tango binding automagically manages the data types, and writing scripts is quite easy:

 1from tango import DeviceProxy
 2
 3# Get proxy on the tango_test1 device
 4print("Creating proxy to TangoTest device...")
 5tango_test = DeviceProxy("sys/tg_test/1")
 6
 7# First use the classical command_inout way to execute the DevString command
 8# (DevString in this case is a command of the Tango_Test device)
 9
10result = tango_test.command_inout("DevString", "First hello to device")
11print("Result of execution of DevString command = {0}".format(result))
12
13# the same can be achieved with a helper method
14result = tango_test.DevString("Second Hello to device")
15print("Result of execution of DevString command = {0}".format(result))
16
17# Please note that argin argument type is automatically managed by python
18result = tango_test.DevULong(12456)
19print("Result of execution of DevULong command = {0}".format(result))

Execute commands with more complex types

In this case you have to use put your arguments data in the correct python structures:

 1from tango import DeviceProxy
 2
 3# Get proxy on the tango_test1 device
 4print("Creating proxy to TangoTest device...")
 5tango_test = DeviceProxy("sys/tg_test/1")
 6
 7# The input argument is a DevVarLongStringArray so create the argin
 8# variable containing an array of longs and an array of strings
 9argin = ([1,2,3], ["Hello", "TangoTest device"])
10
11result = tango_test.DevVarLongStringArray(argin)
12print("Result of execution of DevVarLongArray command = {0}".format(result))

Work with Groups

Todo

write this how to

Handle errors

Todo

write this how to

For now check Exception API.

Registering devices

Here is how to define devices in the Tango DataBase:

 1from tango import Database, DbDevInfo
 2
 3#  A reference on the DataBase
 4db = Database()
 5
 6# The 3 devices name we want to create
 7# Note: these 3 devices will be served by the same DServer
 8new_device_name1 = "px1/tdl/mouse1"
 9new_device_name2 = "px1/tdl/mouse2"
10new_device_name3 = "px1/tdl/mouse3"
11
12# Define the Tango Class served by this  DServer
13new_device_info_mouse = DbDevInfo()
14new_device_info_mouse._class = "Mouse"
15new_device_info_mouse.server = "ds_Mouse/server_mouse"
16
17# add the first device
18print("Creating device: %s" % new_device_name1)
19new_device_info_mouse.name = new_device_name1
20db.add_device(new_device_info_mouse)
21
22# add the next device
23print("Creating device: %s" % new_device_name2)
24new_device_info_mouse.name = new_device_name2
25db.add_device(new_device_info_mouse)
26
27# add the third device
28print("Creating device: %s" % new_device_name3)
29new_device_info_mouse.name = new_device_name3
30db.add_device(new_device_info_mouse)

Setting up device properties

A more complex example using python subtilities. The following python script example (containing some functions and instructions manipulating a Galil motor axis device server) gives an idea of how the Tango API should be accessed from Python:

 1from tango import DeviceProxy
 2
 3# connecting to the motor axis device
 4axis1 = DeviceProxy("microxas/motorisation/galilbox")
 5
 6# Getting Device Properties
 7property_names = ["AxisBoxAttachement",
 8                  "AxisEncoderType",
 9                  "AxisNumber",
10                  "CurrentAcceleration",
11                  "CurrentAccuracy",
12                  "CurrentBacklash",
13                  "CurrentDeceleration",
14                  "CurrentDirection",
15                  "CurrentMotionAccuracy",
16                  "CurrentOvershoot",
17                  "CurrentRetry",
18                  "CurrentScale",
19                  "CurrentSpeed",
20                  "CurrentVelocity",
21                  "EncoderMotorRatio",
22                  "logging_level",
23                  "logging_target",
24                  "UserEncoderRatio",
25                  "UserOffset"]
26
27axis_properties = axis1.get_property(property_names)
28for prop in axis_properties.keys():
29    print("%s: %s" % (prop, axis_properties[prop][0]))
30
31# Changing Properties
32axis_properties["AxisBoxAttachement"] = ["microxas/motorisation/galilbox"]
33axis_properties["AxisEncoderType"] = ["1"]
34axis_properties["AxisNumber"] = ["6"]
35axis1.put_property(axis_properties)

Using clients with multiprocessing

Since version 9.3.0 PyTango provides cleanup() which resets CORBA connection. This static function is needed when you want to use tango with multiprocessing in your client code.

In the case when both your parent process and your child process create DeviceProxy, Database or/and AttributeProxy your child process inherits the context from your parent process, i.e. open file descriptors, the TANGO and the CORBA state. Sharing the above objects between the processes may cause unpredictable errors, e.g. TRANSIENT_CallTimedout, unidentifiable C++ exception. Therefore, when you start a new process you must reset CORBA connection:

 1import time
 2import tango
 3
 4from multiprocessing import Process
 5
 6
 7class Worker(Process):
 8
 9    def __init__(self):
10        Process.__init__(self)
11
12    def run(self):
13        # reset CORBA connection
14        tango.ApiUtil.cleanup()
15
16        proxy = tango.DeviceProxy('test/tserver/1')
17
18        stime = time.time()
19        etime = stime
20        while etime - stime < 1.:
21            try:
22                proxy.read_attribute("Value")
23            except Exception as e:
24                print(str(e))
25            etime = time.time()
26
27
28def runworkers():
29    workers = [Worker() for _ in range(6)]
30    for wk in workers:
31        wk.start()
32    for wk in workers:
33        wk.join()
34
35
36db = tango.Database()
37dp = tango.DeviceProxy('test/tserver/1')
38
39for i in range(4):
40    runworkers()

After cleanup() all references to DeviceProxy, AttributeProxy or Database objects in the current process become invalid and these objects need to be reconstructed.

Using clients with multithreading

When performing Tango I/O from user-created threads, there can be problems. This is often more noticeable with event subscription and unsubscription, but it could affect any Tango I/O. As PyTango wraps the cppTango library, we needs to consider how cppTango’s threads work.

cppTango was originally developed at a time where C++ didn’t have standard threads. All the threads currently created in cppTango are omni threads, since this is what the omniORB library is using to create threads and since this implementation is available for free with omniORB.

In C++, users used to create omni threads in the past so there was no issue. Since C++11, C++ comes with an implementation of standard threads. cppTango is currently (version 9.3.3) not directly thread safe when a user is using C++11 standard threads or threads different than omni threads. This lack of thread safety includes threads created from Python’s threading module.

In an ideal future cppTango should should protect itself, regardless of what type of threads are used. In the meantime, we need a work-around.

The work-around when using threads which are not omni threads is to create an object of the C++ class omni_thread::ensure_self in the user thread, just after the thread creation, and to delete this object only when the thread has finished its job. This omni_thread::ensure_self object provides a dummy omniORB ID for the thread. This ID is used when accessing thread locks within cppTango, so the ID must remain the same for the lifetime of the thread. Also note that this object MUST be released before the thread has exited, otherwise omniORB will throw an exception.

A Pythonic way to implement this work-around for multithreaded applications is available via the EnsureOmniThread class. It was added in PyTango version 9.3.2. This class is best used as a context handler to wrap the target method of the user thread. An example is shown below:

 1import tango
 2from threading import Thread
 3from time import sleep
 4
 5
 6def thread_task():
 7    with tango.EnsureOmniThread():
 8        eid = dp.subscribe_event(
 9            "double_scalar", tango.EventType.PERIODIC_EVENT, cb)
10        while running:
11            print("num events stored {}".format(len(cb.get_events())))
12            sleep(1)
13        dp.unsubscribe_event(eid)
14
15
16cb = tango.utils.EventCallback()  # print events to stdout
17dp = tango.DeviceProxy("sys/tg_test/1")
18dp.poll_attribute("double_scalar", 1000)
19thread = Thread(target=thread_task)
20running = True
21thread.start()
22sleep(5)
23running = False
24thread.join()

Another way to create threads in Python is the concurrent.futures.ThreadPoolExecutor. The problem with this is that the API does not provide an easy way for the context handler to cover the lifetime of the threads, which are created as daemons. One option is to at least use the context handler for the functions that are submitted to the executor. I.e., executor.submit(thread_task). This is not guaranteed to work. A second option to investigate (if using at least Python 3.7) is the initializer argument which could be used to ensure a call to the __enter__() method for a thread-specific instance of EnsureOmniThread. However, calling the __exit__() method on the corresponding object at shutdown is a problem. Maybe it could be submitted as work.

Write a server

Before reading this chapter you should be aware of the TANGO basic concepts. This chapter does not explain what a Tango device or a device server is. This is explained in detail in the Tango control system manual

Since version 8.1, PyTango provides a helper module which simplifies the development of a Tango device server. This helper is provided through the tango.server module.

Here is a simple example on how to write a Clock device server using the high level API

 1 import time
 2 from tango.server import Device, attribute, command, pipe
 3
 4
 5 class Clock(Device):
 6
 7     @attribute
 8     def time(self):
 9         return time.time()
10
11     @command(dtype_in=str, dtype_out=str)
12     def strftime(self, format):
13         return time.strftime(format)
14
15     @pipe
16     def info(self):
17         return ('Information',
18                 dict(manufacturer='Tango',
19                      model='PS2000',
20                      version_number=123))
21
22
23 if __name__ == "__main__":
24     Clock.run_server()
line 2

import the necessary symbols

line 5

tango device class definition. A Tango device must inherit from tango.server.Device

line 7-9

definition of the time attribute. By default, attributes are double, scalar, read-only. Check the attribute for the complete list of attribute options.

line 11-13

the method strftime is exported as a Tango command. In receives a string as argument and it returns a string. If a method is to be exported as a Tango command, it must be decorated as such with the command() decorator

line 15-20

definition of the info pipe. Check the pipe for the complete list of pipe options.

line 24

start the Tango run loop. This method automatically determines the Python class name and exports it as a Tango class. For more complicated cases, check run() for the complete list of options

There is a more detailed clock device server in the examples/Clock folder.

Here is a more complete example on how to write a PowerSupply device server using the high level API. The example contains:

  1. a read-only double scalar attribute called voltage

  2. a read/write double scalar expert attribute current

  3. a read-only double image attribute called noise

  4. a ramp command

  5. a host device property

  6. a port class property

 1from time import time
 2from numpy.random import random_sample
 3
 4from tango import AttrQuality, AttrWriteType, DispLevel
 5from tango.server import Device, attribute, command
 6from tango.server import class_property, device_property
 7
 8
 9class PowerSupply(Device):
10
11    current = attribute(label="Current", dtype=float,
12                        display_level=DispLevel.EXPERT,
13                        access=AttrWriteType.READ_WRITE,
14                        unit="A", format="8.4f",
15                        min_value=0.0, max_value=8.5,
16                        min_alarm=0.1, max_alarm=8.4,
17                        min_warning=0.5, max_warning=8.0,
18                        fget="get_current", fset="set_current",
19                        doc="the power supply current")
20
21    noise = attribute(label="Noise", dtype=((float,),),
22                      max_dim_x=1024, max_dim_y=1024,
23                      fget="get_noise")
24
25    host = device_property(dtype=str)
26    port = class_property(dtype=int, default_value=9788)
27
28    @attribute
29    def voltage(self):
30        self.info_stream("get voltage(%s, %d)" % (self.host, self.port))
31        return 10.0
32
33    def get_current(self):
34        return 2.3456, time(), AttrQuality.ATTR_WARNING
35
36    def set_current(self, current):
37        print("Current set to %f" % current)
38
39    def get_noise(self):
40        return random_sample((1024, 1024))
41
42    @command(dtype_in=float)
43    def ramp(self, value):
44        print("Ramping up...")
45
46
47if __name__ == "__main__":
48    PowerSupply.run_server()

Server logging

This chapter instructs you on how to use the tango logging API (log4tango) to create tango log messages on your device server.

The logging system explained here is the Tango Logging Service (TLS). For detailed information on how this logging system works please check:

The easiest way to start seeing log messages on your device server console is by starting it with the verbose option. Example:

python PyDsExp.py PyDs1 -v4

This activates the console tango logging target and filters messages with importance level DEBUG or more. The links above provided detailed information on how to configure log levels and log targets. In this document we will focus on how to write log messages on your device server.

Basic logging

The most basic way to write a log message on your device is to use the Device logging related methods:

Example:

1def read_voltage(self):
2    self.info_stream("read voltage attribute")
3    # ...
4    return voltage_value

This will print a message like:

1282206864 [-1215867200] INFO test/power_supply/1 read voltage attribute

every time a client asks to read the voltage attribute value.

The logging methods support argument list feature (since PyTango 8.1). Example:

1def read_voltage(self):
2    self.info_stream("read_voltage(%s, %d)", self.host, self.port)
3    # ...
4    return voltage_value

Logging with print statement

This feature is only possible since PyTango 7.1.3

It is possible to use the print statement to log messages into the tango logging system. This is achieved by using the python’s print extend form sometimes refered to as print chevron.

Same example as above, but now using print chevron:

1def read_voltage(self, the_att):
2    print >>self.log_info, "read voltage attribute"
3    # ...
4    return voltage_value

Or using the python 3k print function:

1def read_Long_attr(self, the_att):
2    print("read voltage attribute", file=self.log_info)
3    # ...
4    return voltage_value

Logging with decorators

This feature is only possible since PyTango 7.1.3

PyTango provides a set of decorators that place automatic log messages when you enter and when you leave a python method. For example:

1@tango.DebugIt()
2def read_Long_attr(self, the_att):
3    the_att.set_value(self.attr_long)

will generate a pair of log messages each time a client asks for the ‘Long_attr’ value. Your output would look something like:

1282208997 [-1215965504] DEBUG test/pydsexp/1 -> read_Long_attr()
1282208997 [-1215965504] DEBUG test/pydsexp/1 <- read_Long_attr()
Decorators exist for all tango log levels:
The decorators receive three optional arguments:
  • show_args - shows method arguments in log message (defaults to False)

  • show_kwargs shows keyword method arguments in log message (defaults to False)

  • show_ret - shows return value in log message (defaults to False)

Example:

1@tango.DebugIt(show_args=True, show_ret=True)
2def IOLong(self, in_data):
3    return in_data * 2

will output something like:

1282221947 [-1261438096] DEBUG test/pydsexp/1 -> IOLong(23)
1282221947 [-1261438096] DEBUG test/pydsexp/1 46 <- IOLong()

Multiple device classes (Python and C++) in a server

Within the same python interpreter, it is possible to mix several Tango classes. Let’s say two of your colleagues programmed two separate Tango classes in two separated python files: A PLC class in a PLC.py:

 1# PLC.py
 2
 3from tango.server import Device
 4
 5class PLC(Device):
 6
 7    # bla, bla my PLC code
 8
 9if __name__ == "__main__":
10    PLC.run_server()

… and a IRMirror in a IRMirror.py:

 1# IRMirror.py
 2
 3from tango.server import Device
 4
 5class IRMirror(Device):
 6
 7    # bla, bla my IRMirror code
 8
 9if __name__ == "__main__":
10    IRMirror.run_server()

You want to create a Tango server called PLCMirror that is able to contain devices from both PLC and IRMirror classes. All you have to do is write a PLCMirror.py containing the code:

1# PLCMirror.py
2
3from tango.server import run
4from PLC import PLC
5from IRMirror import IRMirror
6
7run([PLC, IRMirror])
It is also possible to add C++ Tango class in a Python device server as soon as:
  1. The Tango class is in a shared library

  2. It exist a C function to create the Tango class

For a Tango class called MyTgClass, the shared library has to be called MyTgClass.so and has to be in a directory listed in the LD_LIBRARY_PATH environment variable. The C function creating the Tango class has to be called _create_MyTgClass_class() and has to take one parameter of type “char *” which is the Tango class name. Here is an example of the main function of the same device server than before but with one C++ Tango class called SerialLine:

 1import tango
 2import sys
 3
 4if __name__ == '__main__':
 5    py = tango.Util(sys.argv)
 6    util.add_class('SerialLine', 'SerialLine', language="c++")
 7    util.add_class(PLCClass, PLC, 'PLC')
 8    util.add_class(IRMirrorClass, IRMirror, 'IRMirror')
 9
10    U = tango.Util.instance()
11    U.server_init()
12    U.server_run()
Line 6

The C++ class is registered in the device server

Line 7 and 8

The two Python classes are registered in the device server

Create attributes dynamically

It is also possible to create dynamic attributes within a Python device server. There are several ways to create dynamic attributes. One of the way, is to create all the devices within a loop, then to create the dynamic attributes and finally to make all the devices available for the external world. In C++ device server, this is typically done within the <Device>Class::device_factory() method. In Python device server, this method is generic and the user does not have one. Nevertheless, this generic device_factory method calls a method named dyn_attr() allowing the user to create his dynamic attributes. It is simply necessary to re-define this method within your <Device>Class and to create the dynamic attribute within this method:

dyn_attr(self, dev_list)

where dev_list is a list containing all the devices created by the generic device_factory() method.

There is another point to be noted regarding dynamic attribute within Python device server. The Tango Python device server core checks that for each attribute it exists methods named <attribute_name>_read and/or <attribute_name>_write and/or is_<attribute_name>_allowed. Using dynamic attribute, it is not possible to define these methods because attributes name and number are known only at run-time. To address this issue, the Device_3Impl::add_attribute() method has a diferent signature for Python device server which is:

add_attribute(self, attr, r_meth = None, w_meth = None, is_allo_meth = None)

attr is an instance of the Attr class, r_meth is the method which has to be executed with the attribute is read, w_meth is the method to be executed when the attribute is written and is_allo_meth is the method to be executed to implement the attribute state machine. The method passed here as argument as to be class method and not object method. Which argument you have to use depends on the type of the attribute (A WRITE attribute does not need a read method). Note, that depending on the number of argument you pass to this method, you may have to use Python keyword argument. The necessary methods required by the Tango Python device server core will be created automatically as a forward to the methods given as arguments.

Here is an example of a device which has a TANGO command called createFloatAttribute. When called, this command creates a new scalar floating point attribute with the specified name:

 1from tango import Util, Attr, AttrWriteType
 2from tango.server import Device, command
 3
 4class MyDevice(Device):
 5
 6    @command(dtype_in=str)
 7    def CreateFloatAttribute(self, attr_name):
 8        attr = Attr(attr_name, tango.DevDouble, AttrWriteType.READ_WRITE)
 9        self.add_attribute(attr, self.read_General, self.write_General)
10
11    def read_General(self, attr):
12        self.info_stream("Reading attribute %s", attr.get_name())
13        attr.set_value(99.99)
14
15    def write_General(self, attr):
16        self.info_stream("Writting attribute %s", attr.get_name())

Create/Delete devices dynamically

This feature is only possible since PyTango 7.1.2

Starting from PyTango 7.1.2 it is possible to create devices in a device server “en caliente”. This means that you can create a command in your “management device” of a device server that creates devices of (possibly) several other tango classes. There are two ways to create a new device which are described below.

Tango imposes a limitation: the tango class(es) of the device(s) that is(are) to be created must have been registered before the server starts. If you use the high level API, the tango class(es) must be listed in the call to run(). If you use the lower level server API, it must be done using individual calls to add_class().

Dynamic device from a known tango class name

If you know the tango class name but you don’t have access to the tango.DeviceClass (or you are too lazy to search how to get it ;-) the way to do it is call create_device() / delete_device(). Here is an example of implementing a tango command on one of your devices that creates a device of some arbitrary class (the example assumes the tango commands ‘CreateDevice’ and ‘DeleteDevice’ receive a parameter of type DevVarStringArray with two strings. No error processing was done on the code for simplicity sake):

 1from tango import Util
 2from tango.server import Device, command
 3
 4class MyDevice(Device):
 5
 6    @command(dtype_in=[str])
 7    def CreateDevice(self, pars):
 8        klass_name, dev_name = pars
 9        util = Util.instance()
10        util.create_device(klass_name, dev_name, alias=None, cb=None)
11
12    @command(dtype_in=[str])
13    def DeleteDevice(self, pars):
14        klass_name, dev_name = pars
15        util = Util.instance()
16        util.delete_device(klass_name, dev_name)

An optional callback can be registered that will be executed after the device is registed in the tango database but before the actual device object is created and its init_device method is called. It can be used, for example, to initialize some device properties.

Dynamic device from a known tango class

If you already have access to the DeviceClass object that corresponds to the tango class of the device to be created you can call directly the create_device() / delete_device(). For example, if you wish to create a clone of your device, you can create a tango command called Clone:

 1class MyDevice(tango.Device):
 2
 3    def fill_new_device_properties(self, dev_name):
 4        prop_names = db.get_device_property_list(self.get_name(), "*")
 5        prop_values = db.get_device_property(self.get_name(), prop_names.value_string)
 6        db.put_device_property(dev_name, prop_values)
 7
 8        # do the same for attributes...
 9        ...
10
11    def Clone(self, dev_name):
12        klass = self.get_device_class()
13        klass.create_device(dev_name, alias=None, cb=self.fill_new_device_properties)
14
15    def DeleteSibling(self, dev_name):
16        klass = self.get_device_class()
17        klass.delete_device(dev_name)

Note that the cb parameter is optional. In the example it is given for demonstration purposes only.

Write a server (original API)

This chapter describes how to develop a PyTango device server using the original PyTango server API. This API mimics the C++ API and is considered low level. You should write a server using this API if you are using code generated by Pogo tool or if for some reason the high level API helper doesn’t provide a feature you need (in that case think of writing a mail to tango mailing list explaining what you cannot do).

The main part of a Python device server

The rule of this part of a Tango device server is to:

  • Create the Util object passing it the Python interpreter command line arguments

  • Add to this object the list of Tango class(es) which have to be hosted by this interpreter

  • Initialize the device server

  • Run the device server loop

The following is a typical code for this main function:

1if __name__ == '__main__':
2    util = tango.Util(sys.argv)
3    util.add_class(PyDsExpClass, PyDsExp)
4
5    U = tango.Util.instance()
6    U.server_init()
7    U.server_run()
Line 2

Create the Util object passing it the interpreter command line arguments

Line 3

Add the Tango class PyDsExp to the device server. The Util.add_class() method of the Util class has two arguments which are the Tango class PyDsExpClass instance and the Tango PyDsExp instance. This Util.add_class() method is only available since version 7.1.2. If you are using an older version please use Util.add_TgClass() instead.

Line 7

Initialize the Tango device server

Line 8

Run the device server loop

The PyDsExpClass class in Python

The rule of this class is to :

  • Host and manage data you have only once for the Tango class whatever devices of this class will be created

  • Define Tango class command(s)

  • Define Tango class attribute(s)

In our example, the code of this Python class looks like:

 1class PyDsExpClass(tango.DeviceClass):
 2
 3    cmd_list = { 'IOLong' : [ [ tango.ArgType.DevLong, "Number" ],
 4                              [ tango.ArgType.DevLong, "Number * 2" ] ],
 5                 'IOStringArray' : [ [ tango.ArgType.DevVarStringArray, "Array of string" ],
 6                                     [ tango.ArgType.DevVarStringArray, "This reversed array"] ],
 7    }
 8
 9    attr_list = { 'Long_attr' : [ [ tango.ArgType.DevLong ,
10                                    tango.AttrDataFormat.SCALAR ,
11                                    tango.AttrWriteType.READ],
12                                  { 'min alarm' : 1000, 'max alarm' : 1500 } ],
13
14                 'Short_attr_rw' : [ [ tango.ArgType.DevShort,
15                                       tango.AttrDataFormat.SCALAR,
16                                       tango.AttrWriteType.READ_WRITE ] ]
17    }
Line 1

The PyDsExpClass class has to inherit from the DeviceClass class

Line 3 to 7

Definition of the cmd_list dict defining commands. The IOLong command is defined at lines 3 and 4. The IOStringArray command is defined in lines 5 and 6

Line 9 to 17

Definition of the attr_list dict defining attributes. The Long_attr attribute is defined at lines 9 to 12 and the Short_attr_rw attribute is defined at lines 14 to 16

If you have something specific to do in the class constructor like initializing some specific data member, you will have to code a class constructor. An example of such a contructor is

1def __init__(self, name):
2    tango.DeviceClass.__init__(self, name)
3    self.set_type("TestDevice")

The device type is set at line 3.

Defining commands

As shown in the previous example, commands have to be defined in a dict called cmd_list as a data member of the xxxClass class of the Tango class. This dict has one element per command. The element key is the command name. The element value is a python list which defines the command. The generic form of a command definition is:

'cmd_name' : [ [in_type, <"In desc">], [out_type, <"Out desc">], <{opt parameters}>]

The first element of the value list is itself a list with the command input data type (one of the tango.ArgType pseudo enumeration value) and optionally a string describing this input argument. The second element of the value list is also a list with the command output data type (one of the tango.ArgType pseudo enumeration value) and optionaly a string describing it. These two elements are mandatory. The third list element is optional and allows additional command definition. The authorized element for this dict are summarized in the following array:

key

Value

Definition

“display level”

DispLevel enum value

The command display level

“polling period”

Any number

The command polling period (mS)

“default command”

True or False

To define that it is the default command

Defining attributes

As shown in the previous example, attributes have to be defined in a dict called attr_list as a data member of the xxxClass class of the Tango class. This dict has one element per attribute. The element key is the attribute name. The element value is a python list which defines the attribute. The generic form of an attribute definition is:

'attr_name' : [ [mandatory parameters], <{opt parameters}>]

For any kind of attributes, the mandatory parameters are:

[attr data type, attr data format, attr data R/W type]

The attribute data type is one of the possible value for attributes of the tango.ArgType pseudo enunmeration. The attribute data format is one of the possible value of the tango.AttrDataFormat pseudo enumeration and the attribute R/W type is one of the possible value of the tango.AttrWriteType pseudo enumeration. For spectrum attribute, you have to add the maximum X size (a number). For image attribute, you have to add the maximun X and Y dimension (two numbers). The authorized elements for the dict defining optional parameters are summarized in the following array:

key

value

definition

“display level”

tango.DispLevel enum value

The attribute display level

“polling period”

Any number

The attribute polling period (mS)

“memorized”

“true” or “true_without_hard_applied”

Define if and how the att. is memorized

“label”

A string

The attribute label

“description”

A string

The attribute description

“unit”

A string

The attribute unit

“standard unit”

A number

The attribute standard unit

“display unit”

A string

The attribute display unit

“format”

A string

The attribute display format

“max value”

A number

The attribute max value

“min value”

A number

The attribute min value

“max alarm”

A number

The attribute max alarm

“min alarm”

A number

The attribute min alarm

“min warning”

A number

The attribute min warning

“max warning”

A number

The attribute max warning

“delta time”

A number

The attribute RDS alarm delta time

“delta val”

A number

The attribute RDS alarm delta val

The PyDsExp class in Python

The rule of this class is to implement methods executed by commands and attributes. In our example, the code of this class looks like:

 1class PyDsExp(tango.Device):
 2
 3    def __init__(self,cl,name):
 4        tango.Device.__init__(self, cl, name)
 5        self.info_stream('In PyDsExp.__init__')
 6        PyDsExp.init_device(self)
 7
 8    def init_device(self):
 9        self.info_stream('In Python init_device method')
10        self.set_state(tango.DevState.ON)
11        self.attr_short_rw = 66
12        self.attr_long = 1246
13
14    #------------------------------------------------------------------
15
16    def delete_device(self):
17        self.info_stream('PyDsExp.delete_device')
18
19    #------------------------------------------------------------------
20    # COMMANDS
21    #------------------------------------------------------------------
22
23    def is_IOLong_allowed(self):
24        return self.get_state() == tango.DevState.ON
25
26    def IOLong(self, in_data):
27        self.info_stream('IOLong', in_data)
28        in_data = in_data * 2
29        self.info_stream('IOLong returns', in_data)
30        return in_data
31
32    #------------------------------------------------------------------
33
34    def is_IOStringArray_allowed(self):
35        return self.get_state() == tango.DevState.ON
36
37    def IOStringArray(self, in_data):
38        l = range(len(in_data)-1, -1, -1)
39        out_index=0
40        out_data=[]
41        for i in l:
42            self.info_stream('IOStringArray <-', in_data[out_index])
43            out_data.append(in_data[i])
44            self.info_stream('IOStringArray ->',out_data[out_index])
45            out_index += 1
46        self.y = out_data
47        return out_data
48
49    #------------------------------------------------------------------
50    # ATTRIBUTES
51    #------------------------------------------------------------------
52
53    def read_attr_hardware(self, data):
54        self.info_stream('In read_attr_hardware')
55
56    def read_Long_attr(self, the_att):
57        self.info_stream("read_Long_attr")
58
59        the_att.set_value(self.attr_long)
60
61    def is_Long_attr_allowed(self, req_type):
62        return self.get_state() in (tango.DevState.ON,)
63
64    def read_Short_attr_rw(self, the_att):
65        self.info_stream("read_Short_attr_rw")
66
67        the_att.set_value(self.attr_short_rw)
68
69    def write_Short_attr_rw(self, the_att):
70        self.info_stream("write_Short_attr_rw")
71
72        self.attr_short_rw = the_att.get_write_value()
73
74    def is_Short_attr_rw_allowed(self, req_type):
75        return self.get_state() in (tango.DevState.ON,)
Line 1

The PyDsExp class has to inherit from the tango.Device (this will used the latest device implementation class available, e.g. Device_5Impl)

Line 3 to 6

PyDsExp class constructor. Note that at line 6, it calls the init_device() method

Line 8 to 12

The init_device() method. It sets the device state (line 9) and initialises some data members

Line 16 to 17

The delete_device() method. This method is not mandatory. You define it only if you have to do something specific before the device is destroyed

Line 23 to 30

The two methods for the IOLong command. The first method is called is_IOLong_allowed() and it is the command is_allowed method (line 23 to 24). The second method has the same name than the command name. It is the method which executes the command. The command input data type is a Tango long and therefore, this method receives a python integer.

Line 34 to 47

The two methods for the IOStringArray command. The first method is its is_allowed method (Line 34 to 35). The second one is the command execution method (Line 37 to 47). The command input data type is a string array. Therefore, the method receives the array in a python list of python strings.

Line 53 to 54

The read_attr_hardware() method. Its argument is a Python sequence of Python integer.

Line 56 to 59

The method executed when the Long_attr attribute is read. Note that before PyTango 7 it sets the attribute value with the tango.set_attribute_value function. Now the same can be done using the set_value of the attribute object

Line 61 to 62

The is_allowed method for the Long_attr attribute. This is an optional method that is called when the attribute is read or written. Not defining it has the same effect as always returning True. The parameter req_type is of type AttReqtype which tells if the method is called due to a read or write request. Since this is a read-only attribute, the method will only be called for read requests, obviously.

Line 64 to 67

The method executed when the Short_attr_rw attribute is read.

Line 69 to 72

The method executed when the Short_attr_rw attribute is written. Note that before PyTango 7 it gets the attribute value with a call to the Attribute method get_write_value with a list as argument. Now the write value can be obtained as the return value of the get_write_value call. And in case it is a scalar there is no more the need to extract it from the list.

Line 74 to 75

The is_allowed method for the Short_attr_rw attribute. This is an optional method that is called when the attribute is read or written. Not defining it has the same effect as always returning True. The parameter req_type is of type AttReqtype which tells if the method is called due to a read or write request.

General methods

The following array summarizes how the general methods we have in a Tango device server are implemented in Python.

Name

Input par (with “self”)

return value

mandatory

init_device

None

None

Yes

delete_device

None

None

No

always_executed_hook

None

None

No

signal_handler

int

None

No

read_attr_hardware

sequence<int>

None

No

Implementing a command

Commands are defined as described above. Nevertheless, some methods implementing them have to be written. These methods names are fixed and depend on command name. They have to be called:

  • is_<Cmd_name>_allowed(self)

  • <Cmd_name>(self, arg)

For instance, with a command called MyCmd, its is_allowed method has to be called is_MyCmd_allowed and its execution method has to be called simply MyCmd. The following array gives some more info on these methods.

Name

Input par (with “self”)

return value

mandatory

is_<Cmd_name>_allowed

None

Python boolean

No

Cmd_name

Depends on cmd type

Depends on cmd type

Yes

Please check Data types chapter to understand the data types that can be used in command parameters and return values.

The following code is an example of how you write code executed when a client calls a command named IOLong:

1def is_IOLong_allowed(self):
2    self.debug_stream("in is_IOLong_allowed")
3    return self.get_state() == tango.DevState.ON
4
5def IOLong(self, in_data):
6    self.info_stream('IOLong', in_data)
7    in_data = in_data * 2
8    self.info_stream('IOLong returns', in_data)
9    return in_data
Line 1-3

the is_IOLong_allowed method determines in which conditions the command ‘IOLong’ can be executed. In this case, the command can only be executed if the device is in ‘ON’ state.

Line 6

write a log message to the tango INFO stream (click here for more information about PyTango log system).

Line 7

does something with the input parameter

Line 8

write another log message to the tango INFO stream (click here for more information about PyTango log system).

Line 9

return the output of executing the tango command

Implementing an attribute

Attributes are defined as described in chapter 5.3.2. Nevertheless, some methods implementing them have to be written. These methods names are fixed and depend on attribute name. They have to be called:

  • is_<Attr_name>_allowed(self, req_type)

  • read_<Attr_name>(self, attr)

  • write_<Attr_name>(self, attr)

For instance, with an attribute called MyAttr, its is_allowed method has to be called is_MyAttr_allowed, its read method has to be called read_MyAttr and its write method has to be called write_MyAttr. The attr parameter is an instance of Attr. Unlike the commands, the is_allowed method for attributes receives a parameter of type AttReqtype.

Please check Data types chapter to understand the data types that can be used in attribute.

The following code is an example of how you write code executed when a client read an attribute which is called Long_attr:

1def read_Long_attr(self, the_att):
2    self.info_stream("read attribute name Long_attr")
3    the_att.set_value(self.attr_long)
Line 1

Method declaration with “the_att” being an instance of the Attribute class representing the Long_attr attribute

Line 2

write a log message to the tango INFO stream (click here for more information about PyTango log system).

Line 3

Set the attribute value using the method set_value() with the attribute value as parameter.

The following code is an example of how you write code executed when a client write the Short_attr_rw attribute:

1def write_Short_attr_rw(self,the_att):
2    self.info_stream("In write_Short_attr_rw for attribute ",the_att.get_name())
3    self.attr_short_rw = the_att.get_write_value(data)
Line 1

Method declaration with “the_att” being an instance of the Attribute class representing the Short_attr_rw attribute

Line 2

write a log message to the tango INFO stream (click here for more information about PyTango log system).

Line 3

Get the value sent by the client using the method get_write_value() and store the value written in the device object. Our attribute is a scalar short attribute so the return value is an int