enpassant 0.3.1

En passant assignment for clearer if and while control structures

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

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

Python does not. This leads to more code lines and (in my opinion) 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 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.

Usage

from enpassant import *
result = Passer()

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

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. While item ([]) and attribute (.) access work directly on result, this is 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 Operations

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 ← (LEFTARDS ARROW), but alas! Until Python gets more Unicode-savvy, we have to choose some existing ASCII operator to repurpose.

Grabber and Similar

I’ve begun experimenting 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

• En passant is a chess term.

• En passant assignment / naming is discussed in Issue1714448 and PEP 379, 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 the wonderful pytest, and tox. Successfully packaged for, and tested against, all late-model versions of Python: 2.6, 2.7, 3.2, 3.3, 3.4, 3.5a4, as well as PyPy 2.5.1 (based on 2.7.9) and PyPy3 2.4.0 (based on 3.2.5). Test line coverage ~100%.

• On Python 2.6, uses Raymond Hettinger’s ordereddict module (which is included in the source tree for ease of installation) to provide OrderedDict. Thank you, Raymond!

• The simplere package similarly provides en passant handling (and other helpers) for the important, common case of regular expression searches.

• The author, Jonathan Eunice or @jeunice on Twitter welcomes your comments and suggestions.

Installation

To install the latest version:

pip install -U enpassant

To easy_install under a specific Python version (3.3 in this example):

python3.3 -m easy_install --upgrade 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.)