Other changes

Attribute write methods

To make code consistent we have renamed the parameters for all write attribute methods of DeviceProxy, AttributeProxy and Group to be either attr (for single attribute write methods) or attr_values (for multple attributes write methods):

Method	Old parameter name	New parameter name
DeviceProxy.write_attribute	attr_name	attr
DeviceProxy.write_attributes	name_val	attr_values
DeviceProxy.write_read_attribute	attr_name	attr
DeviceProxy.write_read_attributes	name_val	attr_values
DeviceProxy.write_attribute_asynch	attr_name	attr
DeviceProxy.write_attributes_asynch	attr_values	attr_values
AttributeProxy.write	attr_name	attr
AttributeProxy.write_read	attr_name	attr
AttributeProxy.write_asynch	attr_name	attr
Group.write_attribute_asynch	attr_name	attr

And for the all methods now there is a “fast” overload, where instead of attribute name you can give an AttributeInfoEx object (that overload existed before for DeviceProxy.write_attribute and AttributeProxy.write methods). Due to actual attribute write being done by cppTango, PyTango must know to which c++ data type to cast each python value. In case if only attribute name is provided it must fetch attribute info for this attribute from server by additional synchronous IO. But in case if AttributeInfoEx is given by user - this IO can be skiped. Note, that for multiple attribute write AttributeInfoEx must be provided for each pair of values. In case if at least for one pair just name is given - PyTango must do additional IO.

Tip

In case you write identical attributes, you can reuse one AttributeInfoEx object for all. You also can create an AttributeInfoEx object on your own and fill only name, data_type, data_format fields instead of fetching the full structure from the server.

Defining attributes that push events

If a device pushes events for attributes programmatically, the standard way to indicate this was to call a method like set_change_event, set_archive_event, etc. from the device’s init_device method. This is still possible, however, there is a new alternative that makes the attribute definition self-contained.

Instead of this:

from tango.server import Device, attribute

class MyDevice(Device):
    def init_device(self):
        self.set_change_event("voltage", implemented=True, detect=False)

    @attribute
    def voltage(self) -> float:
        return 1.23

    ...

You can now do this:

from tango.server import Device, attribute

class MyDevice(Device):

    @attribute(change_event_implemented=True, change_event_detect=False)
    def voltage(self) -> float:
        return 1.23

    ...

These are the new keyword arguments available for attribute() (they all default to False):

• alarm_event_implemented

• alarm_event_detect

• archive_event_implemented

• archive_event_detect

• change_event_implemented

• change_event_detect

• data_ready_event_implemented

Support setting command doc_in and doc_out from docstrings

Instead of providing documentation via the doc_in and doc_out kwargs for your commands, this can now be taken from the doctring.

Instead of:

from tango.server import Device, command

class MyDevice(Device):
    @command(
        doc_in="val_in (int): The input integer value.",
        doc_out="returns (int): An integer result after processing.",
    )
    def example(self, val_in: int) -> int:
        return val_in

You can do this:

from tango.server import Device, command

class MyDevice(Device):
        @command
        def example(self, val_in: int) -> int:
            """
            Example command, using Google docstring syntax

            Args:
                val_in (int): The input integer value.

            Returns:
                int: An integer result after processing.
            """
            return val_in

You can also use ReST, Numpydoc-style, or Epydoc docstrings. See the docstring_parser library.

Note

Text outside the input and return types is ignored. E.g., the summary of the command won’t be included. The keyword arguments take priority over the docstring.