Python's dynamic nature (rough slides, November 2004)

Python's dynamic nature (rough slides, November 2004)

• 1.
  Python’s Dynamic Nature An investigationof how Python’s dynamic nature allows it to scale up from one-line scripts to massive applications. Kiran Jonnalagadda <jace@seacrow.com> These are rough slides. The finals will make heavy use of diagrams in place of text.
• 2.
  What Does Interpreted Mean? •Python is not line interpreted like shell script. • Python is byte-code compiled like Java, but compilation is on-the-fly. • However, Python in some ways behaves like a line-interpreted language, making the learning curve easy.
• 3.
  Loading a Module •When a module (file) is loaded, it is first compiled into byte-code. • Python then executes all the code in the module that is not in a function. • This code can modify the module as it loads, making for the line-interpreted feel. • Code in functions is only executed on call.
• 4.
  Example of Loading #Example code print “Top level executes on load.” def somefunc(): print “Executes only when called.” class foo: print “Class level also executes on load.” def bar(self): print “Executes only when called.”
• 5.
  Bags of Attributes •Python is a fully object oriented language. • All symbols are objects with attributes. • Objects contain other objects. Therefore, Python’s namespace is nested. • A module is an object, a string is an object, an integer is an object, everything is an object.
• 6.
  __attributes__ • Python insertsseveral special attributes into objects. These attributes are all named in the __attribute__ style. • Example 1: __name__ indicates the name given to the object in code. • Example 2: __class__ refers to the class this object is an instance of.
• 7.
  Operator Overloading • Operatoroverloading is also achieved via functions with special attribute names. • Examples: • __eq__ and __ne__ for == and != • __add__ to add two objects • __cmp__ to compare, for sorting
• 8.
  Controlling Attribute Access • Anobject can take control of the way its attributes are accessed. • The __getitem__ function for list or dictionary access. Example: object[‘foo’] • The __getattr__ function for named access. Example: using object.foo will call object.__getattr__(‘foo’) • __setitem__ and __setattr__ for editing.
• 9.
  Documentation • Python codeis self-documenting. • Documentation can be placed in a string, which must be the first non-remark declaration in a module, class or function. • This documentation is accessible as the __doc__ attribute of the object. • Example: class foo: “This is the documentation.” pass
• 10.
  Namespaces • In Python,everything is contained within something else. • The top-level, invisible container is called __builtins__. Contains internal functions. • Top-level symbols not from __builtins__ are local symbols. • Local symbols are not private. They can be accessed via the namespace.
• 11.
  Namespaces #2 • Thetop-level namespace is not crowded; an app can have several modules with conflicting names and not have a conflict. • “module1.cname” is different from “module2.cname” • Using “name = module1.cname” makes “name” a local alias for “module1.cname”
• 12.
  Symbols and Objects •Symbols are not the same as objects. • Symbols are only references to objects. • An object can have multiple references. • Objects are garbage collected when they have zero references. GC algorithms used are both reference counting and markand-sweep.
• 13.
  Multiple References • Anmodule can consist of nothing but references to symbols in other modules. • Hence the term, “bags of attributes.” • This allows a large package to have a single API module that links to the rest of the package internally.
• 14.
  References Example module1.py: def foo(bar,baz): return bar + baz module2.py: import module1 new_name = module1.foo module3.py: from module2 import new_name print new_name(1, 2)
• 15.
  Runtime Editing • Pythoncode can edit itself at run-time. • The only way to define variables in a class is by editing the instance at runtime. • Editing involves simply changing the object a symbol refers to. • This only works when all code refers to the object via the symbol, which is the case almost all the time.
• 16.
  How Classes Work •{To insert code examples and diagram here, showing a complex class definition and how it works when different parts are accessed. This will be across several slides.}
• 17.
  Scalability • Because ofthe clear namespace separation, a Python application can contain thousands of modules without conflicts. • Advanced class definition ability helps avoid repetitive code. • This makes Python code easy to maintain and scale, while still remaining simple.
• 18.
  The Type/Class Dichotomy • Priorto version 2.2, Python had two types of objects: Types and Classes. • Types could only be defined in C code. • Classes could only be defined in Python. • Classes could not be derived from Types. • This meant a base-class could not be defined externally in C/C++ code.
• 19.
  Zope’s Solution • JimFulton, the creator of Zope, created a new Type called ExtensionClass. • ExtensionClass can be derived from, like a regular Python Class, but is a Type. • Achieved by patching Python internals. • ExtensionClass made a lot of extensions possible that could not be done in Python.
• 20.
  Type/Class Merger • Python2.2 merges Types and Classes, so they are all one type of object. • But changes are incompatible with ExtensionClass, which Zope depends on. • So Zope still uses the old Type/Class divide, even with Python 2.2 and 2.3. • Zope will upgrade with Zope 3.