enpassant

| |travisci| |versions| |impls| |wheel| |coverage| |br-coverage|

.. |travisci| image:: https://api.travis-ci.org/jonathaneunice/enpassant.svg :target: http://travis-ci.org/jonathaneunice/enpassant

.. |versions| image:: https://img.shields.io/pypi/pyversions/enpassant.svg :alt: Supported versions :target: https://pypi.python.org/pypi/enpassant

.. |impls| image:: https://img.shields.io/pypi/implementation/enpassant.svg :alt: Supported implementations :target: https://pypi.python.org/pypi/enpassant

.. |wheel| image:: https://img.shields.io/pypi/wheel/enpassant.svg :alt: Wheel packaging support :target: https://pypi.python.org/pypi/enpassant

.. |coverage| image:: https://img.shields.io/badge/test_coverage-100%25-blue.svg :alt: Test line coverage :target: https://pypi.python.org/pypi/enpassant

.. |br-coverage| image:: https://img.shields.io/badge/branch_coverage-99%25-blue.svg :alt: Test branch coverage :target: https://pypi.python.org/pypi/enpassant

Simple en passant assignment, giving Python clearer conditional statements

Usage

::

from __future__ import print_function # Python 2/3 compat
from enpassant import *
result = Passer()

while result / expensive_request():
    print(result.report())

    # assuming report() is a method of the object
    # expensive_request() would naturally hand back

Discussion

Many languages support en passant (in passing) assignment, like so::

if result = expensive_request():
    print(result.report())

In Python, that's a syntax error. This leads to more code lines and, in some cases, less visual clarity::

result = expensive_request()
if result:
    print(result.report())

Or worse, in the case of looping structures::

result = expensive_request()
while result:
    print(result.report())
    result = expensive_request()

It doesn't look so bad here, in a highly distilled example. But in real programs, the called function often has parameters to be managed, and the surrounding code is invariably longer and more complicated. The more complicated the surrounding computations and requests, the simpler the comparison itself should be. As the Zen of Python <http://www.python.org/dev/peps/pep-0020/>_ intones: "Simple is better than complex." and "Readability counts."

I hope that Python will eventually provide a concise way of handling this, such as::

while expensive_request() as result:
    print(result.report())

But in the meanwhile, enpassant provides a workaround.

How it Works

::

from enpassant import *
result = Passer()

while result / expensive_request():
    print(result.report())

Here result / expensive_request() is read "the result of the expensive_request." result is merely a proxy object that, when it encounters the division operator, returns the denominator. That is, result / whatever == whatever. But it also remembers the denominator value. Then, whenever you want the result value provided (presumably, later in the body of your loop or conditional), simply access it through result. If you want the full object returned by expensive_request() you can get it via result.value. Or or the result has items or attributes, they are available by indexing or naming the attribute directly. Easy peasy!

NB: If you change the items or attributes of result, those settings are also forwarded to the underlying object. result is not a copy, but a true proxy, and as close to the actual object returned as I can make it given current Python strictures.

Some Details

enpassant "assignment" is transparent to conditional expressions, because the value of the expression is always the value of the denominator. But Passers are also guaranteed to have a Boolean value identical to that of the value they contain, should you wish to use them in subsequent tests.

The result in the example above isn't the pure result of the following function call (or expression), but rather a proxy to it. Item ([]) and attribute (.) access work directly on result because Passer objects pass on getitem and get-attribute requests to their enclosed value. Usually, this is a convenience, and avoids having to needlessly state that it's really result.value that's being indexed or dereferenced. But if you need the specific object returned (say for an object identity or isinstance test, use result.value directly.

Alternative Value Access

It is also possible to retrieve the value of a Passer by calling it::

if result / expensive_request():
    print(result().report())

This technique makes clear that the value is being rendered via some process, rather than just presented as a normal Python name / variable. And the resulting object from result() is the true and complete result of the earlier function call, with no need for implicit / auto-magical forwarding of items and attributes. Which style makes sense is a matter of judgment and taste.

Or, if you prefer something terser, the + (unary positive) operation will also yield the value::

if result / expensive_request():
    print(+result.report())

Alternative Invocations

.. |larrow| unicode:: 0x2190 .. leftwards arrow

If you prefer the less-than (<) or less-than-or-equal (<=) operators as indicators that result takes the value of the following value, they are supported as aliases of the division operation (/). Thus, the following are identical::

if result / expensive_request():
    print(result.report())

if result < expensive_request():
    print(result.report())

if result <= expensive_request():
    print(result.report())

It's a matter of preference which seems most logical, appropriate, and expressive. None of them are as good Note, however, that the operation usually known as division (/) has a much higher precedence (i.e. tighter binding to its operands) than the typical comparison operations (< and <=). If used with a more complex expressions, either know your precedence or use parenthesis to disambiguate!

It'd be swell if Python supported arbitrary symbols. Unicode has what would be reasonable alternative assignment markers, such as |larrow| (LEFTARDS ARROW <http://www.fileformat.info/info/unicode/char/2190/index.htm>_), but alas! Until Python gets more Unicode-savvy, we have to choose some existing ASCII operator to repurpose.

It is also possible to use a function call idiom if you prefer::

if result(expensive_request()):
    print(result.report())

This has the virtue of looking like a "wrapping" of the expensive request value, rather than reusing / overloading an existing operation.

Grabber and Similar

I've experimented with other forms of collecting and rendering values. This version of enpassant includes the results of one of those experiments. Objects of the Grabber class can have their attributes set on their first access. Subsequent uses of that attribute yield the set value.::

info = Grabber()
info.name('Joe')
assert info.name == 'Joe'

The challenge with this approach is that once set, attribute values cannot be reset.

Notes

• Se CHANGES.yml for the change log.

• En passant <http://en.wikipedia.org/wiki/En_passant>_ is a chess term.

• En passant assignment / naming is discussed in Issue1714448 <http://bugs.python.org/issue1714448>_ and PEP 379 <http://www.python.org/dev/peps/pep-0379/>_, which have been rejected and withdrawn, respectively. But that is years gone by. I hope the idea will be productively reconsidered in the future.

• Automated multi-version testing managed with pytest <http://pypi.python.org/pypi/pytest>, pytest-cov <http://pypi.python.org/pypi/pytest-cov>, coverage <https://pypi.python.org/pypi/coverage/4.0b1>, and tox <http://pypi.python.org/pypi/tox>. Packaging linting with pyroma <https://pypi.python.org/pypi/pyroma>_.

• Successfully packaged for, and tested against, all early 2017 versions of Python: 2.6, 2.7, 3.2, 3.3, 3.4, 3.5, and 3.6, as well as latest PyPy and PyPy3.

• On Python 2.6, uses Raymond Hettinger's ordereddict <https://pypi.python.org/pypi/ordereddict>_ module (which is included in the source tree for ease of installation) to provide OrderedDict. Thank you, Raymond!

• The simplere <http://pypi.python.org/pypi/simplere>_ package similarly provides en passant handling (and other helpers) for the important, common case of regular expression searches.

• An alterantive module: dataholder <https://pypi.python.org/pypi/dataholder/1.0.1>_.

• The author, Jonathan Eunice <mailto:[e-mail]>_ or @jeunice on Twitter <http://twitter.com/jeunice>_ welcomes your comments and suggestions.

Installation

To install or upgrade to the latest version::

pip install -U enpassant

You may need to prefix these with sudo to authorize installation. In environments without super-user privileges, you may want to use pip's --user option, to install only for a single user, rather than system-wide. You may also need version-specific versions of pip, such as pip2 and pip3, depending on your system configuration.