pydop 1.0

Python package for the creation of delta-oriented Software Product Lines.

The pydop Library

pydop is a small python library for Delta-Oriented Software Product Lines development, that provides classes for

• feature models
• modular and transformation-based SPL (a generalization of Delta-Oriented Programming)
• modular Multi-SPL
• standard transformation operations, like the addition or the removal of a field or a method in a class

Installation

This library is available using the pip command:

$ pip install pydop

Alternatively, it is possible to simply clone this repository: this library is implemented in pure python and only depends on networkx: installing networkx using pip and cloning this repository is enough to install it:

$ pip install networkx
$ git clone https://github.com/onera/pydop.git

An Hello World Example

This example is taken from [1] and consists of a Multi-Lingual Hello World Product Line, i.e., an SPL that generates a class named Greater that can say Hello World in multiple languages, and possibly multiple times.

In pydop, a product line is split in 5 parts:

• the feature model, that states the variability of the SPL
• the base artifact factory, that generates the initial artifact for every variant generation
• the SPL object itself
• a list of deltas, that are python functions that modify the base artifact
• the configuration knowledge, that links the deltas to the SPL by stating which delta execute for which combination of features, and in which order

Part 0: the imports

In this Example, we have 5 imports:

from pydop.fm_constraint import *

This line loads all the classes to declare cross-tree constraints, i.e., boolean constraints over feature names.

from pydop.fm_diagram import *

This line loads all the classes used to declare a feature model.

from pydop.spl import SPL, RegistryGraph

This line loads the SPL class and RegistryGraph. The later is a class that allows to specify the ordering between delta as a generic dependency graph.

from pydop.operations.modules import *

This line loads all the basic delta operation on standard python objects (including classes and modules)

import sys

This line loads the sys module, to access sys.exit in case of problems.

Part 1: the Feature Model

fm = FD("HelloWorld",
 FDMandatory(
  FD("Language",
   FDAlternative(
    FD("English"),
    FD("Dutch"),
    FD("German")
 ))),
 FDOptional(
  FD("Repeat", times=Int(0,1000))
))

Here, FD introduces a new feature. Hence the root feature of this Feature Model is HelloWorld, with one mandatory feature Language (with the languages English, Dutch and German available); and with one optional feature Repeat. That last feature has an attribute times, that states that the number of repetition must be in the interval [0, 1000[.

Part 2: the Base Artifact Factory

def gen_base_artifact():
 class Greeter(object):
   def sayHello(self): return "Hello World"
 return Greeter

The factory is a python function with no parameter, that returns the initial artifact of a variant generation process. In our case, the initial variant is the Greeter class with the method sayHello that returns "Hello World" said in english.

Part 3: the SPL Object

spl = SPL(fm, RegistryGraph(), gen_base_artifact)

The SPL constructor takes three parameters:

• the feature model of the SPL
• an object stating how the ordering between delta can be specified (in our case, that object is an instance of the RegistryGraph class)
• and the base artifact factory

Note: it is possible in pydop to use a Pure Delta-Oriented Programming approach [2] by putting None in place of the base artifact factory. In that case, the parameter of the first delta being executed will be None, and this delta would be in charge of providing the base artifact.

Part 4: the Deltas

def Nl(Greeter):
 @modify(Greeter)
 def sayHello(self): return "Hallo wereld"

def De(Greeter):
 @modify(Greeter)
 def sayHello(self): return "Hallo Welt"

def Rpt(Greeter, product):
 @modify(Greeter)
 def sayHello(self):
  tmp_str = self.original()
  return " ".join(tmp_str for _ in range(product["times"]))

Deltas are python functions that modify the base artifact to construct the expected variant. The first delta implements the Dutch feature. It is a function that takes one argument: the artifact to be modified. This delta modifies that artifact (in our case, the class Greeter) by replacing its sayHello method and making it return "hello World in dutch. The second delta implements the German feature and is implemented like the Dutch one. The third delta implements the Repeat feature and takes two parameters: the first one is the artifact to be modified (like for the two previous deltas), and the second one is the product that triggered the variant generation. This delta uses this additional argument to retrieve the number of repetition requested by the user, with product["times"] in the last line.

Part 5: the Configuration Knowledge

spl.delta("Dutch")(Nl)
spl.delta("German")(De)
spl.delta("Repeat", after=["Nl", "De"])(Rpt)

This Configuration Knowledge (CK) simply states that the Nl delta is activated by the Dutch feature, the De delta is activated by the German feature, and the Rpt delta is activated by the Repeat feature. Moreover, the statement after=["Nl", "De"] means that the Nl delta cannot be executed before the Nl or the De delta, to ensure that the repeated sentence is the correct one.

Note: it is possible in pydop to declare the CK together with the delta. In this case, the delta declaration would look like

@spl.delta("Dutch")
def Nl(Greeter): ...

@spl.delta("German")
def De(Greeter): ...

@spl.delta("Repeat", after=["Nl", "De"])
def Rpt(Greeter, product): ...

Using the SPL

Now that the SPL is constructed, we can generate variant right away. For instance, the following generates a variant corresponding to selecting the Dutch language, with 4 repetition, and calls the sayHello method on an instance of that variant:

# 1. create a full configuration from a partial specification of features
conf, err = spl.close_configuration({"Dutch": True, "Repeat": True, "times": 4})
if(err): # possibly the partial specification was erroneous
  print(err); sys.exit(-1)

# 2. create the variant by simply calling the SPL with the configuration in parameter
Greater = spl(conf)

# 3. use the variant
print(Greater().sayHello())

Moreover, other variants can also be created in the same python program, like this second one, with the German language selected and no repetition:

conf, err = spl.close_configuration({"German": True, "Repeat": False})
if(err):
  print(err); sys.exit(-1)

Greater = spl(conf)
print(Greater().sayHello())

Other Examples

Other examples are available in the examples folder.

References

[1] Dave Clarke, Radu Muschevici, José Proença, Ina Schaefer, and Rudolf Schlatte. 2010. Variability Modelling in the ABS Language. In FMCO (LNCS, Vol. 6957). Springer, 204–224. doi: 10.1007/978-3-642-25271-6_11

[2] Ina Schaefer, and Ferruccio Damiani. 2010. Pure delta-oriented programming. In FOSD'10. ACM, 49--56. doi: 10.1145/1868688.1868696