Notes on Python

Ah, I love python. Oh no, it's Python. Here's a memo just for myself, it's quite conceptional, not just a copy from document. It will be updated at any time.

• What does it mean by "built-in" modules and names?

  It's with respect to the interpreter. Those modules built into the interpreter are built-in modules. When you start an interpreter, using the interactive environment in the command line or an IDE like Pycharm or whatever ways else, those modules are already there. If you check all names your current module already defines by using dir(), you probably get a result like

  ['__builtins__', '__doc__', '__name__', '__package__']
  

  That means some actions are done when the interpreter starts. Here the value of the name __builtins__ is usually the standard module __builtin__. So if you further check all names defined in this module using dir(__builtins__), you'll get a bunch of names like follows:

  ['ArithmeticError', 'AssertionError', 'AttributeError', 'BaseException', 'BufferError', 'BytesWarning', 'DeprecationWarning', 'EOFError', 'Ellipsis', 'EnvironmentError', 'Exception', 'False', 'FloatingPointError', 'FutureWarning', 'GeneratorExit', 'IOError', 'ImportError', 'ImportWarning', 'IndentationError', 'IndexError', 'KeyError', 'KeyboardInterrupt', 'LookupError', 'MemoryError', 'NameError', 'None', 'NotImplemented', 'NotImplementedError', 'OSError', 'OverflowError', 'PendingDeprecationWarning', 'ReferenceError', 'RuntimeError', 'RuntimeWarning', 'StandardError', 'StopIteration', 'SyntaxError', 'SyntaxWarning', 'SystemError', 'SystemExit', 'TabError', 'True', 'TypeError', 'UnboundLocalError', 'UnicodeDecodeError', 'UnicodeEncodeError', 'UnicodeError', 'UnicodeTranslateError', 'UnicodeWarning', 'UserWarning', 'ValueError', 'Warning', 'WindowsError', 'ZeroDivisionError', '_', '__debug__', '__doc__', '__import__', '__name__', '__package__', 'abs', 'all', 'any', 'apply', 'basestring', 'bin', 'bool', 'buffer', 'bytearray', 'bytes', 'callable', 'chr', 'classmethod', 'cmp', 'coerce', 'compile', 'complex', 'copyright', 'credits', 'delattr', 'dict', 'dir', 'divmod', 'enumerate', 'eval', 'execfile', 'exit', 'file', 'filter', 'float', 'format', 'frozenset', 'getattr', 'globals', 'hasattr', 'hash', 'help', 'hex', 'id', 'input', 'int', 'intern', 'isinstance', 'issubclass', 'iter', 'len', 'license', 'list', 'locals', 'long', 'map', 'max', 'memoryview', 'min', 'next', 'object', 'oct', 'open', 'ord', 'pow', 'print', 'property', 'quit', 'range', 'raw_input', 'reduce', 'reload', 'repr', 'reversed', 'round', 'set', 'setattr', 'slice', 'sorted', 'staticmethod', 'str', 'sum', 'super', 'tuple', 'type', 'unichr', 'unicode', 'vars', 'xrange', 'zip']
  

  Notice that this whole series of modules are already built in your interpreter, you can use or check them directly specifying their names without importing.

  But what's a bit funny is that the module __builtin__ itself is not a built-in module. The value of the built-in variable __builtins__ is this module.

• What is the list of paths used for the interpreter to search for modules?

  The variable sys.path. It's initialized with the environment variable PYTHONPATH or a built-in default. Its type is python list.

• How to check linear algebra library linked in numpy?

  There's a question on stackoverflow, see the most voted answer, which is not the accepted one.

  >>> import numpy as np
  >>> np.__config__.show()
  

• What is metaclass?

  There's a perfect answer on stackoverflow, see the most voted but not accepted one. But there's an even better article.

  In short, class (here we assume old-style class is gone from python, we will never use it) is a class, type is also a class, and classes are all objects in python. When representing a noun, the word class is usually used for user-defined classes, and type for built-in classes.

  type(yourobject) is not a function but a class creator. However it often acts as a function. It creates an "type" object (a class), which is the type of this object. type(name, bases, dict) returns a new "type" object (a class), it's actually a dynamic form of the class statement. So in the second way, type creates a class object, just as class statement does. What creates class objects is called metaclass. type is the default metaclass used by python. You can set your metaclass to replace type to create a class, by putting __metaclass__=yourmetaclass under class Yourclass(baseclass): statement.

  Roughly speaking, thisobject.__class__ attribute is just "who creates this object". If thisobject is a class, then the value of thisobject.__class__ is the metaclass who created it, usually type.

• What is proxy object?

  Refer to this excellent article, and the document. In short, a proxy object can be used just like a strong reference of an object, say if a = Foo(); b = weakref.proxy(a), then you can just use b like a, b is an alternative of a. When a is deleted, for example by del a, then that object is deleted, and further using of b will raise exceptions.

  If b = weakref.ref(a), then b is a weak reference, which is a callable. We retrieve the pointed object by calling it: b(). When a is deleted, then calling b() will get None without any exception.

• What is closure?

  Refer to this article. I'd like to just put the summary in that article here, because it's brief and clear enough:

  • Closure is just a fancy name for a function that remembers the values from the enclosing lexical scope even when the program flow is no longer in the enclosing scope.
  • If you've ever written a function that returned another function, you may have used closures.

  So there's nothing deep or mysterious. func.__closure__ is a tuple including details of the variables defined in the enclosing scope, one can access the content of those variables using func.__closure__[i].cell_contents.

• What is "method"?

  Refer to this question, the answer by AndiDog, not the most voted one. In short, a method is a function defined in a class definition.

• What is the use of -m in the command python -m themodule?

  Two uses. One is in the case that you're in some directory other than the path in sys.path, but you want to run a built-in module, like SimpleHTTPServer. Another is when your objective module is in the path dira/, but it requires external modules in dirb/, then you can walk into dirb/ and run python -m dira/yourmodule. The underlying reason is that -m will search along sys.path, and current working diretory will be the first path in sys.path. By the way, __name__ is always __main__ with or without -m.