Evaluation extension for taurus core model.

The evaluation extension is a special extension that provides evaluation objects. The official scheme name is ‘eval’.

The main usage for this extension is to provide a way of performing mathematical evaluations with data from other source, as well as allowing fast interfacing with sources of data not supported by specific schemes.

The Evaluation Factory (EvaluationFactory) uses the following object naming for referring to attributes (EvaluationAttribute):


or the following for referring to evaluation devices (EvaluationDevice):


or the following for referring to an evaluation authority (EvaluationAuthority):


  • The <authority> segment is optional (except when referring to an EvaluationAuthority). At this point, only //localhost is supported.

  • The @<evaluator> is optional (except when referring to an EvaluationDevice). If not given, it defaults to DefaultEvaluator. See below for further details

  • <expr> is a mathematical expression (using python syntax) that may have references to other taurus attributes by enclosing them between { and }. Expressions will be evaluated by the evaluator device to which the attribute is assigned.

  • The optional <subst> segment is used to provide substitution symbols. <subst> is a semicolon-separated string of <key>=<value> strings.

The evaluator device inherits from SafeEvaluator which by default includes a large subset of mathematical functions from the numpy module. If access to other symbols are required, a custom evaluator can be used. <evaluator> is a unique identification name for the evaluator device object and may define a source of additional symbols to be present for the evaluation. The supported syntax for @<evaluator> is:

  • @<ID> (cannot contain dots or any of / ? # : =). This indicates just an alternative name for the EvaluationDevice, It does not add any extra symbol to the evaluation context.

  • @<modulename>.* (<modulename> may include dots for submodules). It will make all symbols found in the given module available during the evaluation (i.e., it emulates doing from <modulename> import * in the evaluation context).

  • @<modulename>.<customdeviceclass>. Use your own custom EvaluationDevice based class. This allows to define custom symbols see <taurus>/core/evaluation/test/res/dev_example.py, but note that this syntax is is now superseded by the “instance-based” one (see below), which is easier to use and provides write attribute support.

  • @<inst>=<modulename>.<class>() (e.g. @c=mymod.MyClass() ). This will import a class from a module, then instantiate it and then make the instance available for evaluation with the given name. Note that the <inst>= part may be omitted, in which case the instance will be available for evaluation as self. IMPORTANT: If the given class declares writable properties, EvaluationAttributes that access one such property will automatically be considered writable. See examples of usage in <taurus>/core/evaluation/test/res/mymod.py and in <taurus>/core/evaluation/test/res/ipap_example.py

Some examples of valid evaluation models are:

  • An attribute that multiplies a tango attribute by 2:

  • Same as above, but using substitutions:

  • An attribute that adds two tango attributes together (assuming that tango is set as the default scheme):

  • An attribute that generates an array of random values:

  • Same as above, but with units:

  • An attribute that adds noise to a tango image attribute:

  • An attribute that accesses a method from a given module (in this case to use os.path.exists):

  • Same as before, for getting today’s date as an attribute:

  • A default evaluator device named foo:

  • A custom evaluator device (implemented as class MyEvalDev in the mymod module):

  • A custom evaluator device (implemented as class MyEvalDev in the mymod module):

  • A writable attribute foo (implemented as a writable property of the MyClass class from the mymod module):


    assuming that the mymod module defines MyClass as:

    class MyClass(object):
        set_foo(self, value):
        foo = property(get_foo, set_foo)


Previous to SEP3, a RFC3986 non-compliant syntax was used for the evaluation scheme (e.g., allowing names such as eval://db=foo;dev=bar;a*b?k=2;a={tango:a/b/c/d}). This syntax is now deprecated and should not be used. Taurus will issue warnings if detected.



class EvaluationAttribute(name='', parent=None, **kwargs)[source]

A TaurusAttribute that can be used to perform mathematical operations involving other arbitrary Taurus attributes. The mathematical operation is described in the attribute name itself. An Evaluation Attribute will keep references to any other attributes being referenced and it will update its own value whenever any of the referenced attributes change.


In most cases this class should not be instantiated directly. Instead it should be done via the EvaluationFactory.getAttribute()

(more info)

class EvaluationAuthority(complete_name='', parent=None)[source]

Dummy authority class for Evaluation (the authority concept is not yet used in the Evaluation scheme)


In most cases this class should not be instantiated directly. Instead it should be done via the EvaluationFactory.getAuthority()

(more info)

class EvaluationDevice(name='', **kw)[source]

The evaluator object. It is a TaurusDevice and is used as the parent of EvaluationAttribute objects for which it performs the mathematical evaluation.


In most cases this class should not be instantiated directly. Instead it should be done via the EvaluationFactory.getDevice()

(more info)

class EvaluationFactory(*p, **k)[source]

A Singleton class that provides Evaluation related objects.

(more info)