Version 0.3.0

Package documentation

Organize symbols and load them lazily

python 2:

pip install symboldict

python 3:

pip3 install symboldict

The symboldict module implements a dictionary type containing symbolic references to python objects in the hierarchy of importable modules or builtins objects. Examples of symbolic references are 'os.path.isdir', 'telnetlib.Telnet', 'complex.conjugate' etc. References to user-defined objects are equally easy.

The module implements a lazy loading mechanism for these symbols, which means that a dictionary can contain symbols from many different modules without importing any of these modules until this import becomes necessary in user code.

Symboldicts are meant to favor loose coupling between different parts of an application. For example three modules A, B, C may share a common symboldict sy instead of importing each others symbols directly. A and B can use method sy.func() instead of function C.func(), thus decreasing the knowledge that each module possesses about other modules.

A SymbolDict, or dictionary of symbols, is typically created by a statement similar to

>>> sy = SymbolDict(
...     isfile=symbol.os.path.isfile,
...     Telnet=symbol.telnetlib.Telnet,
...     conju=symbol.complex.conjugate,
...     eggs=symbol.spam.eggs,
... )

It is a regular dictionary, which keys are arbitrary but which values are instances of a new type Symbol:

>>> sy
SymbolDict({'eggs': Symbol('spam.eggs'), 'isfile': Symbol('os.path.isfile'), ...})
>>>
>>> isinstance(sy, dict)
True

The usual dictionary operations apply:

>>> sy['isfile']
Symbol('os.path.isfile')
>>> sy['eggs']
Symbol('spam.eggs')
>>> sy['Parser'] = Symbol('argparse.ArgumentParser')
>>> 'Telnet' in sy
True

Attribute access is overloaded for symboldicts. It supplies a lazy access to the value symbolically represented by the Symbol instance:

>>> sy.isfile
<function isfile at 0x7f7fd0f40050>
>>> sy.Telnet
<class 'telnetlib.Telnet at 0x7fbb3a518db8'>
>>> sy.conju
<method 'conjugate' of 'complex' objects>
>>> sy.eggs
Traceback (most recent call last)
    ...
ImportError: No module named spam

This syntax enables user code to manipulate the SymbolDict instance as if it was a module containing python variables available through qualified names. Actual modules are imported only when it is necessary to do so.

The values of these variables can also be obtained by the following alternative expression, which handles the case where the dictionary keys are not character strings or where they are existing attribute names of the dict class (such as 'update' or 'clear'):

>>> sy.getvalue('isfile')
<function isfile at 0x7f7fd0f40050>
>>> sy.getvalue('Parser')
<class 'argparse.ArgumentParser'>

Previous versions of this module used a call syntax here, requiring expressions such as sy().getvalue('isfile'). Calling a SymbolDict instance is deprecated and currently returns the instance itself.

Special attributes

As of version 0.3.0, the valued loaded lazily, such as sy.Telnet above are stored in the instance's __dict__ under the corresponding attribute. It means that subsequent accesses to these values have the efficiency of a simple attribute access.

To avoid collisions with ordinary dict method names, SymbolDict will normally reject the use of a certain number of keys such as

  _strict     has_key     itervalues   keys      iterkeys     items
  iteritems   viewkeys    hasvalue     update    fromkeys     clear
  pop         viewitems   symboldict   popitem   setdefault   values
  getvalue    strict      get          copy      viewvalues

and magic attribute names such as __doc__ or __setattr__. Using one of these keys in a symboldict will raise a TypeError exception.

This restriction on the keys can be removed by unsetting the strict property

sy.strict = False

It can also be removed at instantiation time by calling the LaxSymbolDict constructor instead of SymbolDict. For those lax symboldicts, forbidden keys such as 'popitem' can be used, but the lazy access can only be obtained through the getvalue() method. Loaded values won't be added to the instance's __dict__.

Symbol objects (used as values in SymbolDict instances) wrap a dot-separated path to a python object, which may exist or not, for example

>>> a = Symbol('wave.Error')
>>> b = Symbol('complex.conjugate')
>>> c = Symbol('spam.ham.eggs')

Attribute access is overloaded to allow building other instances with the dot syntax

>>> Symbol('spam').ham.eggs
Symbol('spam.ham.eggs')

A special instance named symbol is defined, which value is Symbol(''). Its path is empty, and it permits expressions such as

>>> a = symbol.wave.Error
>>> b = symbol.complex.conjugate
>>> c = symbol.spam.ham.eggs

which produce the same result as above.

Defining an instance does not trigger an attempt to retrieve the indicated python object by importing modules or accessing attributes. However, standalone Symbol instances have the ability to fetch this object by calling the getvalue() method:

>>> a().getvalue()
<class 'wave.Error'>

The call syntax a() enables to bypass the overloading of the attribute operator. It returns a special adapter having the type SymbolControl. Method getvalue() cannot be called directly on the Symbol instance.

A method hasvalue() indicates if a value can be obtained for the symbol's path. Unlike getvalue(), it does not raise an exception when there is no value:

>>> symbol.spam.ham().hasvalue()
False

This method also exists for symboldicts. It may raise KeyError if the key is missing in the dictionary:

>>> sy.hasvalue('conju')
True

This software is licensed under the MIT License

© 2014-2015 Eric Ringeisen