Lookup Decorator¶

LookupDecorator is a feature extension to Lookup Class. Whereas the Lookup class allows users to write different classes, functions and variables in tokenized modules and dynamically reference them, the lookup decorator operates at the class method level, allowing users to write a single class with different method implementations per each token variance combination.

                                                .--> TokenX.Y class method
                                               /
UserScript -> import cls -> call cls method --+----> TokenX class method
                                               \
                                                `--> Default (no token)
                                                       class method

# Example
# -------
#
#  a simple lookup decorator example

# my_library/config.py
# --------------------

# import the decorator
# (note the lowercase 'lookup')
from abstract import lookup

# define a class using the decorator on its methods
class ConfigureRouting(object)
    def __init__(self, os):
        self.os = os

    # apply the decorator on methods to be abstracted
    @lookup('os')
    def apply_config(self):
        # ... insert generic/non-os specific code here


# my_library/nxos/config.py
# -------------------------
from ..config import ConfigureRouting as BaseConfigRouting

# inherit the parent class
class ConfigureRouting(BaseConfigRouting):

    # define the same method specific to this token
    def apply_config(self):
        # ... insert nxos specific code here

The main benefit of using LookupDecorator is that it allows the user to perform standard python import and deal with only one class instance. During runtime, the engine looks up the class’s attributes and forms a list of tokens based on these values, and replaces the decorated methods during with a “more” appropriate one from a tokenized search (see Search Algorithm).

# Example
# -------
#
#   using the above code

# import the main entry class directly
from my_library.config import ConfigureRouting

# use it as you would naturally
obj = ConfigureRouting(os = 'nxos')

# when a decorated method is called, the lookup occurs and the
# most appriorate method from one of its subclasses is called instead.
result = obj.apply_config()
# lookup information
# ------------------
#   attributes to read: os
#   attribute value: os = 'nxos'
#
# thus, the search result equivalence is:
#   from my_library.nxos.config import ConfigureRouting
#   result = ConfigureRouting.apply_config(obj)

Usages¶

To use LookupDecorator, start with writing your abstraction-enabled library as you normally would. When arriving at defining classes that requires methods level abstraction, simply apply the decorator onto each method that needs to be abstracted. Behaviors:

Lookup decorator can be imported as lookup (note the lowercase), or as decorator.LookupDecorator. They are exactly the same, but some may prefer one name over the other.
```
from abstract import lookup
from abstract.decorator import LookupDecorator
```
The usage of lookup decorator does not mandate a top-level Abstraction-Enabled Package declaration. It only requires Abstraction Tokens definitions under the module where the lookup decorator is used.
```
Example:
  if LookupDecorator is used in on class X under module A.B,
  tokens should be defined as child modules under A.B.
```
Lookup decorator takes in a list of attributes names as arguments. During runtime, the engine will lookup the given class instance for these attributes to be used as tokens. This mechanism is called an attribute getter. The default attribute getter looks up both the class instance and instance.device (if exists) for the named attribute.
```
class MyClass(object):

    @lookup('attr_1', 'attr_2')
    def some_func(self):
        # ...

# equivalent to
#     obj = MyClass()
#     token_1 = getattr(obj, 'attr_1', getattr(obj.device, 'attr_1'))
#     token_2 = getattr(obj, 'attr_2', getattr(obj.device, 'attr_2'))
```
The search for matching token combinations always begins at this class’s module declaration level onwards. It will match for the same relative path as the current module, and the same class name (or names in nested class defs) and target method.
```
Example:
  a search originating from: moduleX.moduleY.classA.classB.some_func()
  may match: moduleX.moduleY.tokenJ.tokenK.classA.classB.some_func()
```
the default attribute getter can be replaced by providing a new function through attr_getter argument. The provided function must take in two arguments: obj and attr for both the object under scrutiny and the attribute to lookup

# Examples
# --------
#
#   lookup decorator usage

# assuming we had a lookup-decorator enabled library
# my_library.my_module.ConfigureOspf

# import it regularly
from my_library.my_module import ConfigureOspf

# instaciate it naturally
# (in this case our class requires argument 'os' and mgmt_context)
routing = ConfigureOspf(os = 'iosxr', mgmt_context = 'yang')

# if we call a decorated method, say, apply_configuration
# eg, code snippet:
#       @lookup('os', 'mgmt_context')
#       def apply_configuration(self):
#           # ... code

routing.apply_configuration()
# the engine translates this to:
#    token_os = routing.os = 'iosxr'
#    token_mgmt_context = routing.mgmt_context = 'yang'
# and the resulting lookup equivalent could be:
#    from my_library.my_module.iosxr.yang import ConfigureOspf
#    result = ConfigureOspf.apply_configuration(routing)

# note
# ----
#   after lookup is performed, notice that the found target class's method
#   is called directly with the original class instance as first argument.
#   This is a python property: class methods can be treated as "functions"
#   if you pass in a "similar" class instance as the first argument.
#   See: https://docs.python.org/3.4/tutorial/classes.html#method-objects

LookupDecorator Class Argument List¶
Argument	Description
`*attrs`	list of attributes to be used read as input tokens for lookup
`attr_getter`	class instance attribute getter (optional)
`builder`	token permutation builder (optional)
`**builder_kwargs`	any keyword arguments/values to be passed to the builder (optional)

Lookup From Device Decorator¶

LookupDecorator.from_device is a feature extension to LookupDecorator. The lookup.from_device decorator operates at the runtime, allowing users to write a single class with different method implementations and dynamically based on the token variance combination from device’s custom abstraction or pre-defined at class method level.

# Example
# -------
#
#  a simple lookup.from_device decorator example

# my_library/config.py
# --------------------

# import the decorator
# (note the lowercase 'lookup')
from abstract import lookup

# define a class using the decorator on its methods
class ConfigureRouting(object)
    def __init__(self, os):
        self.os = os

    # apply the decorator on methods to be abstracted dynamically based on
    # custom abstraction data
    @lookup.from_device
    def apply_config(self):
        # ... insert generic/non-os specific code here

    # apply the decorator on methods to be abstracted dynamically based on
    # custom abstraction data or fallback to token 'os'
    @lookup.from_device('os')
    def check_config(self):
        # ... insert generic/non-os specific code here