tobacco 1.0.3

Function pipelines for python

Tobacco: Your shiny new pipe >>!

Do you miss the %>% from R/dplyr? What about the |> from Julia? Now you can use a tobacco.pipe to chain together function calls! The @pipe decorator will take a function and repurpose the builtin right-bitshift operator >> to be a function pipe. I've hardly ever seen anyone using >>, so it doesn't seem likely many people will miss the default behavior.

A pipeline takes the value of a function call on the left-hand side, and passes it in as the first argument to the function on the right-hand side:

0 >> f >> g  # evaluates to g(f(0))

You can even specify keyword closures within the pipeline

# Evaluates to g(f(0, some_keyword=5), another="foo")
0 >> f(some_keyword=5) >> g(another="foo")

Installation

pip install tobacco

Usage

from tobacco import pipe


@pipe  # `pipe` overrides the builtin right-bitshift operator `>>`
def plus_one(a: int) -> int:
    return a + 1


@pipe
def minus_amount(b: int, amount: int = 2) -> int:
    return b - amount


@pipe
def plus_five(c: int) -> int:
    return c + 5


# 0 + 1 - 4 + 5 = 2
res = 0 >> plus_one >> minus_amount(amount=4) >> plus_five
print(res)  # 2

For very long data pipelines, this can clean up the readability a lot:

result = (
    load_data
    >> clean_data
    >> add_columns
    >> do_analysis
    >> upload_to_database
)

In this short example, result is the output of upload_to_database after all the prior steps completed.