Answers to general Tango questions can be found at http://www.tango-controls.org/tutorials
Please also check http://www.tango-controls.org/howtos for a list of Tango howtos
Starting from PyTango 7.0.0 the prefered way to build PyTango is using the standard python distutils package. This means that:
Please check the compilation chapter for details on how to build PyTango.
>>> import PyTango
ImportError: libboost_python-gcc43-mt-1_38.so.1.38.0: cannot open shared object file: No such file or directory
You must check that you have the correct boost python installed on your computer. To see which boost python file PyTango needs type:
$ ldd /usr/lib/python2.5/site-packages/PyTango/_PyTango.so
linux-vdso.so.1 => (0x00007fff48bfe000)
libtango.so.7 => /home/homer/local/lib/libtango.so.7 (0x00007f393fabb000)
liblog4tango.so.4 => /home/homer/local/lib/liblog4tango.so.4 (0x00007f393f8a0000)
**libboost_python-gcc43-mt-1_38.so.1.38.0 => not found**
libpthread.so.0 => /lib/libpthread.so.0 (0x00007f393f65e000)
librt.so.1 => /lib/librt.so.1 (0x00007f393f455000)
libdl.so.2 => /lib/libdl.so.2 (0x00007f393f251000)
libomniORB4.so.1 => /usr/local/lib/libomniORB4.so.1 (0x00007f393ee99000)
libomniDynamic4.so.1 => /usr/local/lib/libomniDynamic4.so.1 (0x00007f393e997000)
libomnithread.so.3 => /usr/local/lib/libomnithread.so.3 (0x00007f393e790000)
libCOS4.so.1 => /usr/local/lib/libCOS4.so.1 (0x00007f393e359000)
libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x00007f393e140000)
libc.so.6 => /lib/libc.so.6 (0x00007f393ddce000)
libstdc++.so.6 => /usr/lib/libstdc++.so.6 (0x00007f393dac1000)
libm.so.6 => /lib/libm.so.6 (0x00007f393d83b000)
/lib64/ld-linux-x86-64.so.2 (0x00007f3940a4c000)
To ease migration effort, PyTango 7 provides an alternative module called PyTango3.
Changing your python import from:
import PyTango
to:
import PyTango3 as PyTango
should allow you to execute your old PyTango code using the new PyTango 7 library.
Please note that you should as soon as possible migrate the code to Tango 7 since the PyTango team cannot assure the maintainability of the PyTango3 module.
Please find below a basic set of rules to migrate from PyTango 3.0.x to 7:
The first important thing to be aware of when migrating from PyTango <= 3.0.4 to PyTango >= 7 is that the data type mapping from tango to python and vice versa is not always the same. The following table summarizes the differences:
Tango data type | PyTango 7 type | PyTango <= 3.0.4 type |
---|---|---|
DEV_VOID | No data | No data |
DEV_BOOLEAN | bool | bool |
DEV_SHORT | int | int |
DEV_LONG | int | int |
DEV_LONG64 | long (on a 32 bits computer) or int (on a 64 bits computer) | long (on a 32 bits computer) or int (on a 64 bits computer) |
DEV_FLOAT | float | float |
DEV_DOUBLE | float | float |
DEV_USHORT | int | int |
DEV_ULONG | int | int |
DEV_ULONG64 | long (on a 32 bits computer) or int (on a 64 bits computer) | long (on a 32 bits computer) or int (on a 64 bits computer) |
DEV_STRING | str | str |
DEVVAR_CHARARRAY | sequence<int> | list<int> |
DEVVAR_SHORTARRAY | sequence<int> | list<int> |
DEVVAR_LONGARRAY | sequence<int> | list<int> |
DEVVAR_LONG64ARRAY | sequence<long> (on a 32 bits computer) or sequence<int> (on a 64 bits computer) | list<long> (on a 32 bits computer) or list<int> (on a 64 bits computer) |
DEVVAR_FLOATARRAY | sequence<float> | list<float> |
DEVVAR_DOUBLEARRAY | sequence<float> | list<float> |
DEVVAR_USHORTARRAY | sequence<int> | list<int> |
DEVVAR_ULONGARRAY | sequence<int> | list<int> |
DEVVAR_ULONG64ARRAY | sequence<long> (on a 32 bits computer) or sequence<int> (on a 64 bits computer) | list<long> (on a 32 bits computer) or list<int> (on a 64 bits computer) |
DEVVAR_STRINGARRAY | sequence<str> | list<str> |
DEVVAR_LONGSTRINGARRAY | A sequence with two elements: 1. sequence<int> 2. sequence<str> |
|
DEVVAR_DOUBLESTRINGARRAY | A sequence with two elements: 1. sequence<float> 2. sequence<str> |
|
Note that starting from PyTango 7 you cannot assume anything about the concrete sequence implementation for the tango array types in PyTango. This means that the following code (valid in PyTango <= 3.0.4):
import PyTango
dp = PyTango.DeviceProxy("my/device/experiment")
da = dp.read_attribute("array_attr")
if isinstance(da.value, list):
print "array_attr is NOT a scalar attribute"
must be replaced with:
import operator, types
import PyTango
dp = PyTango.DeviceProxy("my/device/experiment")
da = dp.read_attribute("array_attr")
if operator.isSequence(da.value) and not type(da.value) in types.StringTypes:
print "array_attr is NOT a scalar attribute"
Note that the above example is intended for demonstration purposes only. For reference, the proper code would be:
import PyTango
dp = PyTango.DeviceProxy("my/device/experiment")
da = dp.read_attribute("array_attr")
if not da.data_format is PyTango.AttrDataFormat.SCALAR:
print "array_attr is NOT a scalar attribute"
replace PyTango.PyUtil with Util
replace PyTango.PyDeviceClass with DeviceClass
in PyTango <= 3.0.4, in order to overwrite the default state and status in a device server, you had to reimplement State() and Status() methods respectively.
in PyTango 7 the methods have been renamed to dev_state() and dev_status() in order to match the C++ API.
(See DeviceProxy API documentation for more details)
For example, a tango command ‘DevVoid Go(DevVarDoubleArray)’ in tango 3.0.4 could be executed by calling:
dev_proxy.command_inout( 'Go', [[1.0, 2.0], [1, 2, 3]] )
and the second list would internally be converted to [‘1’, ‘2’, ‘3’]. Starting from PyTango 7 this is not allowed anymore. So the above code must be changed to:
dev_proxy.command_inout( 'Go', [[1.0, 2.0], ['1', '2', '3']] )
in 3.0.4 DevFailed was a tuple of dictionaries. Now DevFailed is a tuple of DevError. This means that code:
try:
tango_fail()
except PyTango.DevFailed as e:
print e.args[0]['reason']
needs to be replaced with:
try:
tango_fail()
except PyTango.DevFailed as e:
print e.args[0].reason
The following is a list of API improvements. Some where added for performance reasons, others to allow for a more pythonic interface, others still to reflect more adequately the C++ interface. They are not mandatory since the original interface will still be available.
If you want your server to support the V4 interface provided by Tango 7 instead of the V3 provided by Tango 6:
replace the inheritance of your device class from Device_3Impl to Device_4Impl
in the init_device method replace the call:
Device_3Impl.init_device(self)
with:
Device_4Impl.init_device(self)
or better yet, if your device class only inherits from Device_4Impl:
super(<your class>, self).init_device()
In PyTango <= 3.0.4, to set the value of an image attribute you needed it as a flat list. Consider you want to set as value the following image:
# Image:
# | 1 2 |
# | 3 4 |
In order to tell tango the dimensions of the image you had to specify them as:
image = [ 1, 2, 3, 4]
dim_x = 2
dim_y = 2
attr.set_value(image, dim_x, dim_y)
In PyTango 8 it is still supported, but the preferred way is to use a sequence of sequences (instead of a flat sequence), so the dimensions are inherent and not needed anymore:
image = [ [1, 2], [3, 4]]
attr.set_value(image)
If you use a numpy array as the sequence of sequences you can get better performance:
image = numpy.array([ [1, 2], [3, 4]], dtype=numpy.int32)
attr.set_value(image)
Likewise, calls to:
PyTango.set_attribute_value_date_quality(attr, value, date, quality, dim_x, dim_y)
can be replaced with:
attr.set_value_date_quality(value, date, quality)
Imagine the following value is written to our IMAGE attribute:
# Image:
# | 1 2 |
# | 3 4 |
This is what you would do with PyTango <= 3.0.4:
flatList = []
attr.get_write_value(flatList)
print "flatList =", flatList
# flatList = [ 1, 2, 3, 4 ]
You can still do it with PyTango 8. However I recommend:
image = attr.get_write_value()
print "image =", image
# image = numpy.array([[1, 2], [3, 4]])
If PyTango 8 is compiled without numpy support, you will get a sequence of sequences, which makes more sense than a flat list.
If PyTango 8 is compiled with numpy support it does not only makes more sense but it is also considerably faster and memory friendlier.
If PyTango is compiled with numpy support but you prefer a list of lists for some attribute, you can do:
image = attr.get_write_value(PyTango.ExtractAs.List)
print "image =", image
# image = [[1, 2], [3, 4]]
Also the SCALAR attribute case is much cleaner now. Instead of:
data = []
attr.get_write_value(data)
actualData = data[0]
You can just write:
actualData = attr.get_write_value()
The PyTango prefered build system (distutils) uses the same flags used to compile Python to compile PyTango. It happens that Python is compiled as a pure C library while PyTango is a C++ library. Unfortunately one of the flags used by Python is the “-Wstrict-prototypes” which makes sence in a C compilation but not in a C++ compilation. For reference here is the complete error message you may have:
cc1plus: warning: command line option “-Wstrict-prototypes” is valid for Ada/C/ObjC but not for C++
Do not worry about this warning since the compiler is ignoring the presence of this flag in the compilation.
PyTango uses boost python for the binding between C++ and Python and sphinx for document generation. When sphinx generates the PyTango API documentation it uses introspection to search for documentation in the python code. It happens that boost overrides some python introspection API for functions and methods which sphinx expects to have. Therefore you should see many warnings of type:
(WARNING/2) error while formatting signature for PyTango.Device_4Impl.always_executed_hook: **arg is not a Python function**
Do not worry since sphinx is able to generate the proper documentation.