short-con 2.0.0

short-con: Constants collections without hassle

Motivation

When your Python code needs constants, the process often starts simply enough with the worthy goal of getting the magic strings and numbers out of your code.

BLACK = 'black'
WHITE = 'white'

KING = 0
QUEEN = 9
ROOK = 5
BISHOP = 3
KNIGHT = 3
PAWN = 1

At some point, you might need to operate on those constants in groups, so you add some derived constants. We've hardly gotten out of the gate and the journey already seems tedious.

COLORS = (BLACK, WHITE)
PIECES = (KING, QUEEN, ROOK, BISHOP, KNIGHT, PAWN)

Starting in Python 3.4, the enum library became available:

from enum import Enum

Colors = Enum('Colors', 'BLACK WHITE')
Pieces = Enum('Pieces', dict(KING = 0, QUEEN = 9, ROOK = 5, BISHOP = 3, KNIGHT = 3, PAWN = 1))

Although that library helps a lot, there is one annoyance. We started with the simple goal of wrangling magic strings and values, but we end up forced to interact with special enum instances:

Pieces.QUEEN        # Will this give us the value we want? No.
Pieces.QUEEN.value  # Dig a level deeper, friend.

Although there are use cases where such formalism might be desirable, in the vast majority of practical programming situations the intermediate object is just a hassle — a form of robustness theater rather than an actual best practice with concrete benefits.

An easier way

The short-con project simplifies the creation of constants collections: just supply names and values via keyword arguments.

from short_con import cons

PIECES = cons(king = 0, queen = 9, rook = 5, bishop = 3, knight = 3, pawn = 1)

Behind the scenes cons() defines a frozen dataclass and then returns an instance of that class.

Pieces.queen = 99   # Fails with FrozenInstanceError.

The underlying values are directly accessible — no need to interact with a bureaucratic object standing guard in the middle:

PIECES.queen == 9  # True

The object is directly iterable and convertible to other collections, in the manner of dict.items():

for name, value in PIECES:
    print(name, value)

d = dict(PIECES)
tups = list(PIECES)

The object also supports relevant read-only dict behaviors:

# Always supported.
PIECES['queen']      # 9
len(PIECES)          # 6
'queen' in PIECES    # True

# Supported if the attribute names do not conflict with the method names.
PIECES.keys()        # ('king', 'queen', 'rook', 'bishop', 'knight', 'pawn')
PIECES.values()      # (0, 9, 5, 3, 3, 1)
PIECES.get('rook')   # 5
PIECES.get('blort')  # None

For situations when the values are the same as the attribute names, usage is even more compact: just supply names as positional arguments or via one or more space-delimited strings.

COLORS = cons('black white')
COLORS = cons('black', 'white')

print(COLORS)  # ShortCon(black='black', white='white')

The library also supports the creation of enum-like collections: supply the names and, optionally, start and step parameters to control the generation of the numeric values.

PETS1 = enumcons('dog cat parrot')
PETS2 = enumcons('dog cat parrot', start = 100, step = -10)

print(PETS1)  # ShortCon(dog=1, cat=2, parrot=3)
print(PETS2)  # ShortCon(dog=100, cat=90, parrot=80)

Finally, the library provides a constants() function that supports (1) the ability to control the class name of the underlying dataclass, and (2) use cases where the constant values can be computed from the names. The first argument to constant() should be the names and values (via a dict) or just the names (via a list, tuple, or space-delimited str).

COLORS = constants(
    'black white',         # dict, list, tuple, or str
    cls_name = 'Colors',
    val_func = str.upper,  # Callable: f(NAME) => VALUE
)

print(COLORS)  # Colors(black='BLACK', white='WHITE')