Underscode, a Python identifier-like encoding

Ivan Vilata i Balaguer

About Underscode

Underscode is an encoding which is capable of representing any Unicode string as a valid (and quite similar) Python identifier. The way Unicode strings are encoded minimises the chances of clashing with other existing names, while not obscuring the resulting string too much.

Some method decorators are provided which allow arbitrary objects to be accessed as normal instance attributes, with optional tab-completion support for interactive usage. The standard Python codec API is also supported.

Underscode-encoded (or underscoded) strings can be quickly spotted because they end with an odd number of underscores, and they contain escape sequences beginning with an underscore where characters not allowed in identifiers would be found. Some examples of underscoded strings are:

As you see, underscoded strings are quite similar to their decoded counterparts when these are more or less identifier-like, but complex strings can still be handled.

Underscode is a very basic tool which may have several uses:

The Underscode package is released under the GNU Lesser General Public License (LGPL) version 3 or later (see http://www.gnu.org/licenses/).

Underscoded strings as attributes

Underscode provides a module with decorators that allow you to use plain attribute access as a flexible way of accessing all kinds of "child objects" without polluting the normal attribute namespace, and with optional interactive completion if you wish so. For instance, you can make the (string) keys of a dictionary accessible as attributes:

from underscode.decorators import proxy_method

class AttributedDict(dict):
    def __getattr__(self, name):
        return super(AttributedDict, self).__getattr__(name)

    def __setattr__(self, name, value):
        super(AttributedDict, self).__setattr__(name, value)

    def __delattr__(self, name):
        super(AttributedDict, self).__delattr__(name)

Then, access to an attribute which looks like an underscoded string gets the name decoded and used as an argument to __getitem__():

>>> d = AttributedDict()
>>> d
>>> d.foo = 1
>>> d.foo_ = 42
>>> d.foo_, d['foo'], d.foo
(42, 42, 1)
>>> d
{u'foo': 42}
>>> del d.foo_
>>> d

Adding tab-completion on underscoded attributes to this simple example is as easy as applying some ready-to-use decorators on the methods used as arguments to proxy_method. See the documentation of the underscode.decorators module for more information and examples.

Python codec API support

Since the Underscode package is compliant with the standard Python codec API, you can use Underscode to encode and decode strings with the usual unicode.encode() and str.decode() calls at any time just by importing the underscode.codec subpackage (it is not automatically imported by the main underscode package):

>>> import underscode.codec
>>> print u'this is \u201ca test\u201d'
this is “a test”
>>> u'this is \u201ca test\u201d'.encode('underscode')
>>> 'this__is___u201ca__test_u201d_'.decode('underscode')
u'this is \u201ca test\u201d'

Getting Underscode

You can download the source code distribution of Underscode from the Python Package Index at http://pypi.python.org/. It uses the standard setup.py method for installation, runs on any platform and has no additional dependencies but Python version 2.4 or greater.

You may also be interested in following the development of Underscode; you can make a clone of its Fossil repository and checkout it with::

$ fossil clone https://elvil.net/fossil/underscode/ underscode.fossil
$ mkdir underscode && cd underscode
$ fossil open ../underscode.fossil

Helping Underscode

I would like to hear your opinions and feelings on Underscode, to know how you used it in your projects, and to help you solve your problems with it. If you come across a bug or you have some enhancement proposal, please send me an e-mail or create a ticket at https://elvil.net/fossil/underscode/