on-the-lam 1.1.0

Easier Python lambdas

Onthelam

Tired of painstakingly writing the characters lambda: every time you want a simple lambda function? Onthelam is the package for you.

>>> from onthelam import _
>>> mapping = dict(foo=2, bar=1, baz=3)
>>> sorted_by_value = sorted(mapping.items(), key=_[1])
>>> sorted_by_value
[('bar', 1), ('foo', 2), ('baz', 3)]

Sure, you could type lambda pair: pair[1], but doesn't _[1] feel so much nicer? I think it does.

Onthelam supports painless definition of lambda functions that use any combination of comparators, arithmetic operations, bitwise operations, indexing, and attribute getting.

Installation and Requirements

Onthelam is installable from PyPI:

$ pip install on-the-lam

This initial release is only tested with and supports Python 3.9, but future releases will aim at supporting older Python versions as well.

Onthelam is written in pure Python and brings in no additional dependencies.

Readable repr

Onthelam lambdas provide a user-friendly repr string for easier debugging.

>>> fn = -(_.count % 5 + 42) ** 3
>>> fn
_ -> -(_.count % 5 + 42) ** 3

Can your native lambdas do that?

This is especially useful for logging errors in functions that accept functions as arguments, enabling you to log something about the argument that isn't just <function <lambda>(x)>.

Renamable

Maybe you take umbrage with my aesthetic choice to use an underscore as my lambda identifier. That's fine. Rename it all you want:

>>> from onthelam import LambdaBuilder
>>> λx = LambdaBuilder("λx")
>>> [*map(λx // 2, range(10)]
[0, 0, 1, 1, 2, 2, 3, 3, 4, 4]

It uses whatever name you give it in its repr:

>>> λx // 2
λx -> λx // 2

Composable

Need to use the same argument twice in your lambda? No sweat.

>>> tetration_2 = _ ** _
>>> tetration_2(5)
3125

Need a lambda with more than one argument? Combining lambda builders with different names gets you a function that takes as many arguments as you have distinct names.

>>> _a = LambdaBuilder("_a")
>>> _b = LambdaBuilder("_b")
>>> _b = LambdaBuilder("_c")
>>> fn = _a[_b] + _c

You still get a useful repr:

>>> fn
_a, _b, _c -> _a[_b] + _c

And it works like you'd expect it:

>>> fn([1, 2, 3], 2, 4)
7

Calling by keyword is permitted as well:

>>> fn(_b=2, _c=4, _a=[1, 2, 3])
7

Designed with itertools in mind

The itertools module in the standard library is incredibly powerful, but using it often results in ugly code where you have to decide whether to use inline lambdas which add a lot of line noise or lots of one-time named function definition blocks that take up a lot of space relative to their importance.

Consider the following implementation of tetration, the mathematical operation of iterated exponentiation of a number with itself.

>>> from functools import reduce
>>> from itertools import repeat
>>> def tetration(x, n):
...     """Iterate `x ** x` n times"""
...     return reduce(_a ** _b, repeat(x, n))
...
>>> tetration(5, 1)
5
>>> tetration(5, 2)
3125
>>> tetration(5, 3)
298023223876953125

Onthelam clears out the clutter from using lambdas.

Limitations

Onthelam works by using the various special methods available to a class for customizing the behavior of an instance when it is operated on. In short: through lots of operator overloading. The limitation is that there are some expressions involving a LambdaBuilder instance that the object can't seamlessly transform into a lambda function. As a result, they are interpreted as attempts to use the defined lambda function as a lambda function. These are:

• Boolean contexts. Anything that tries to interpret the lambda argument's truthiness will fail entirely. E.g., the expression 1 if _ else 0 will fail.
• Use as the index to an object that is not itself a lambda argument. E.g., the expression [1, 2, 3][_] will fail.
• Tests of containment. E.g., the expression _ in [1, 2, 3] will fail.
• Using the lambda argument as an argument in a function call. E.g., the expression ord(_) will call the ord function with an identity lambda, not create a lambda which calls ord on its argument. Consider functools.partial for this case.