Skip to main content

Module enablig a sh like infix syntax (using pipes)

Project description

Infix programming toolkit

Module enabling a sh like infix syntax (using pipes).

Introduction

As an example, here is the solution for the 2nd Euler Project exercise:

Find the sum of all the even-valued terms in Fibonacci which do not exceed four million.

Given fib a generator of Fibonacci numbers:

sum(fib() | where(lambda x: x % 2 == 0) | take_while(lambda x: x < 4000000))

Each pipes is lazy evalatated, can be aliased, and partially initialized, so it could be rewritten as:

is_even = where(lambda x: x % 2 == 0)
sum(fib() | is_even | take_while(lambda x: x < 4000000)

Installing

To install the library, you can just run the following command:

# Linux/macOS
python3 -m pip install pipe

# Windows
py -3 -m pip install pipe

Using

The basic syntax is to use a Pipe like in a shell:

>>> from itertools import count
>>> from pipe import select, take
>>> sum(count() | select(lambda x: x ** 2) | take(10))
285

Some pipes take an argument:

>>> from pipe import where
>>> sum([1, 2, 3, 4] | where(lambda x: x % 2 == 0))
6

Some do not need one:

>>> from pipe import traverse
>>> for i in [1, [2, 3], 4] | traverse:
...     print(i)
1
2
3
4

In which case it's allowed to use the calling parenthesis:

>>> from pipe import traverse
>>> for i in [1, [2, 3], 4] | traverse():
...     print(i)
1
2
3
4

Constructing your own

You can construct your pipes using the Pipe class like:

from pipe import Pipe
square = Pipe(lambda iterable: (x ** 2 for x in iterable))
map = Pipe(lambda iterable, fct: builtins.map(fct, iterable)

As you can see it's often very short to write, and with a bit of luck the function you're wrapping already takes an iterable as the first argument, making the wrapping straight forward:

>>> from collections import deque
>>> from pipe import Pipe
>>> end = Pipe(deque)

and that's it itrable | end(3) is deque(iterable, 3):

>>> list(range(100) | end(3))
[97, 98, 99]

In case it gets more complicated one can use Pipe as a decorator to a function taking an iterable as the first argument, and any other optional arguments after:

>>> from statistics import mean

>>> @Pipe
... def running_average(iterable, width):
...     items = deque(maxlen=width)
...     for item in iterable:
...         items.append(item)
...         yield mean(items)

>>> list(range(20) | running_average(width=2))
[0, 0.5, 1.5, 2.5, 3.5, 4.5, 5.5, 6.5, 7.5, 8.5, 9.5, 10.5, 11.5, 12.5, 13.5, 14.5, 15.5, 16.5, 17.5, 18.5]
>>> list(range(20) | running_average(width=10))
[0, 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5.5, 6.5, 7.5, 8.5, 9.5, 10.5, 11.5, 12.5, 13.5, 14.5]

Partial Pipes

A pipe can be parametrized without being evaluated:

>>> running_average_of_two = running_average(2)
>>> list(range(20) | running_average_of_two)
[0, 0.5, 1.5, 2.5, 3.5, 4.5, 5.5, 6.5, 7.5, 8.5, 9.5, 10.5, 11.5, 12.5, 13.5, 14.5, 15.5, 16.5, 17.5, 18.5]

For multi-argument pipes then can be partially initialized, you can think of curying:

some_iterable | some_pipe(1, 2, 3)

is strictly equivalent to:

some_iterable | some_pipe(1)(2)(3)

So it can be used to specialize pipes, first a dummy example:

>>> @Pipe
... def addmul(iterable, to_add, to_mul):
...     """Computes (x + to_add) * to_mul to every items of the input."""
...     for i in iterable:
...         yield (i + to_add) * to_mul

>>> mul = addmul(0)  # This partially initialize addmul with to_add=0
>>> list(range(10) | mul(10))
[0, 10, 20, 30, 40, 50, 60, 70, 80, 90]

Which also works with keyword arguments:

>>> add = addmul(to_mul=1)  # This partially initialize addmul with `to_mul=1`
>>> list(range(10) | add(10))
[10, 11, 12, 13, 14, 15, 16, 17, 18, 19]

But now for something interesting:

>>> import re
>>> @Pipe
... def grep(iterable, pattern, flags=0):
...     for line in iterable:
...         if re.match(pattern, line, flags=flags):
...             yield line
...
>>> lines = ["Hello", "hello", "World", "world"]
>>> for line in lines | grep("H"):
...     print(line)
Hello

Now let's reuse it in two ways, first with a pattern:

>>> lowercase_only = grep("[a-z]+$")
>>> for line in lines | lowercase_only:
...     print(line)
hello
world

Or now with a flag:

>>> igrep = grep(flags=re.IGNORECASE)
>>> for line in lines | igrep("hello"):
...    print(line)
...
Hello
hello

Deprecations of pipe 1.x

In pipe 1.x a lot of functions were returning iterables and a lot other functions were returning non-iterables, causing confusion. The one returning non-iterables could only be used as the last function of a pipe expression, so they are in fact useless:

range(100) | where(lambda x: x % 2 == 0) | add

can be rewritten with no less readability as:

sum(range(100) | where(lambda x: x % 2 == 0))

so all pipes returning non-iterables are now deprecated and were removed in pipe 2.0.

What shoud I do?

Oh, you just upgraded pipe, got an exception, and landed here? You have three solutions:

  1. Stop using closing-pipes, replace ...|...|...|...|as_list to list(...|...|...|), that's it, it's even shorter.

  2. If "closing pipes" are not an issue for you, and you really like them, just reimplement the few you really need, it often take a very few lines of code, or copy them from here.

  3. If you still rely on a lot of them and are in a hurry, just pip install pipe<2.

And start testing your project using the Python Development Mode so you catch those warnings before they bite you.

But I like them, pleassssse, reintroduce them!

This has already been discussed in #74.

An @Pipe is often easily implemented in a 1 to 3 lines of code function, and the pipe module does not aim at giving all possibilities, it aims at giving the Pipe decorator.

So if you need more pipes, closing pipes, weired pipes, you-name-it, feel free to implement them on your project, and consider the already-implemented ones as examples on how to do it.

See the Constructing your own paragraph below.

Existing Pipes in this module

Alphabetical list of available pipes; when several names are listed for a given pipe, these are aliases.

chain

Chain a sequence of iterables:

>>> from pipe import chain
>>> list([[1, 2], [3, 4], [5]] | chain)
[1, 2, 3, 4, 5]

Warning : chain only unfold iterable containing ONLY iterables:

[1, 2, [3]] | chain

Gives a TypeError: chain argument #1 must support iteration Consider using traverse.

chain_with(other)

Like itertools.chain, yields elements of the given iterable, hen yields elements of its parameters

>>> from pipe import chain_with
>>> list((1, 2, 3) | chain_with([4, 5], [6]))
[1, 2, 3, 4, 5, 6]

dedup(key=None)

Deduplicate values, using the given key function if provided.

>>> from pipe import dedup
>>> list([-1, 0, 0, 0, 1, 2, 3] | dedup)
[-1, 0, 1, 2, 3]
>>> list([-1, 0, 0, 0, 1, 2, 3] | dedup(key=abs))
[-1, 0, 2, 3]

filter(predicate)

Alias for where(predicate), see where(predicate).

groupby(key=None)

Like itertools.groupby(sorted(iterable, key = keyfunc), keyfunc)

>>> from pipe import groupby, map
>>> items = range(10)
>>> ' / '.join(items | groupby(lambda x: "Odd" if x % 2 else "Even")
...                  | select(lambda x: "{}: {}".format(x[0], ', '.join(x[1] | map(str)))))
'Even: 0, 2, 4, 6, 8 / Odd: 1, 3, 5, 7, 9'

islice()

Just the itertools.islice function as a Pipe:

>>> from pipe import islice
>>> list((1, 2, 3, 4, 5, 6, 7, 8, 9) | islice(2, 8, 2))
[3, 5, 7]

izip()

Just the itertools.izip function as a Pipe:

>>> from pipe import izip
>>> list(range(0, 10) | izip(range(1, 11)))
[(0, 1), (1, 2), (2, 3), (3, 4), (4, 5), (5, 6), (6, 7), (7, 8), (8, 9), (9, 10)]

map(), select()

Apply a conversion expression given as parameter to each element of the given iterable

>>> list([1, 2, 3] | map(lambda x: x * x))
[1, 4, 9]

>>> list([1, 2, 3] | select(lambda x: x * x))
[1, 4, 9]

netcat

The netcat Pipe sends and receive bytes over TCP:

data = [
    b"HEAD / HTTP/1.0\r\n",
    b"Host: python.org\r\n",
    b"\r\n",
]
for packet in data | netcat("python.org", 80):
    print(packet.decode("UTF-8"))

Gives:

HTTP/1.1 301 Moved Permanently
Content-length: 0
Location: https://python.org/
Connection: close

permutations(r=None)

Returns all possible permutations:

>>> from pipe import permutations
>>> for item in 'ABC' | permutations(2):
...     print(item)
('A', 'B')
('A', 'C')
('B', 'A')
('B', 'C')
('C', 'A')
('C', 'B')
>>> for item in range(3) | permutations:
...     print(item)
(0, 1, 2)
(0, 2, 1)
(1, 0, 2)
(1, 2, 0)
(2, 0, 1)
(2, 1, 0)

reverse

Like Python's built-in reversed function.

>>> from pipe import reverse
>>> list([1, 2, 3] | reverse)
[3, 2, 1]

select(fct)

Alias for map(fct), see map(fct).

skip()

Skips the given quantity of elements from the given iterable, then yields

>>> from pipe import skip
>>> list((1, 2, 3, 4, 5) | skip(2))
[3, 4, 5]

skip_while(predicate)

Like itertools.dropwhile, skips elements of the given iterable while the predicate is true, then yields others:

>>> from pipe import skip_while
>>> list([1, 2, 3, 4] | skip_while(lambda x: x < 3))
[3, 4]

sort(key=None, reverse=False)

Like Python's built-in "sorted" primitive.

>>> from pipe import sort
>>> ''.join("python" | sort)
'hnopty'
>>> [5, -4, 3, -2, 1] | sort(key=abs)
[1, -2, 3, -4, 5]

t

Like Haskell's operator ":":

>>> from pipe import t
>>> for i in 0 | t(1) | t(2):
...     print(i)
0
1
2

tail(n)

Yields the given quantity of the last elements of the given iterable.

>>> from pipe import tail
>>> for i in (1, 2, 3, 4, 5) | tail(3):
...     print(i)
3
4
5

take(n)

Yields the given quantity of elements from the given iterable, like head in shell script.

>>> from pipe import take
>>> for i in count() | take(5):
...     print(i)
0
1
2
3
4

take_while(predicate)

Like itertools.takewhile, yields elements of the given iterable while the predicate is true:

>>> from pipe import take_while
>>> for i in count() | take_while(lambda x: x ** 2 < 100):
...     print(i)
0
1
2
3
4
5
6
7
8
9

tee

tee outputs to the standard output and yield unchanged items, useful for debugging a pipe stage by stage:

>>> from pipe import tee
>>> sum(["1", "2", "3", "4", "5"] | tee | map(int) | tee)
'1'
1
'2'
2
'3'
3
'4'
4
'5'
5
15

The 15 at the end is the sum returning.

transpose()

Transposes the rows and columns of a matrix.

>>> from pipe import transpose
>>> [[1, 2, 3], [4, 5, 6], [7, 8, 9]] | transpose
[(1, 4, 7), (2, 5, 8), (3, 6, 9)]

traverse

Recursively unfold iterables:

>>> list([[1, 2], [[[3], [[4]]], [5]]] | traverse)
[1, 2, 3, 4, 5]
>>> squares = (i * i for i in range(3))
>>> list([[0, 1, 2], squares] | traverse)
[0, 1, 2, 0, 1, 4]

uniq(key=None)

Like dedup() but only deduplicate consecutive values, using the given key function if provided (or else the identity).

>>> from pipe import uniq
>>> list([1, 1, 2, 2, 3, 3, 1, 2, 3] | uniq)
[1, 2, 3, 1, 2, 3]
>>> list([1, -1, 1, 2, -2, 2, 3, 3, 1, 2, 3] | uniq(key=abs))
[1, 2, 3, 1, 2, 3]

where(predicate), filter(predicate)

Only yields the matching items of the given iterable:

>>> list([1, 2, 3] | where(lambda x: x % 2 == 0))
[2]

Don't forget they can be aliased:

>>> positive = where(lambda x: x > 0)
>>> negative = where(lambda x: x < 0)
>>> sum([-10, -5, 0, 5, 10] | positive)
15
>>> sum([-10, -5, 0, 5, 10] | negative)
-15

Euler project samples

Find the sum of all the multiples of 3 or 5 below 1000.

>>> euler1 = sum(count() | where(lambda x: x % 3 == 0 or x % 5 == 0) | take_while(lambda x: x < 1000))
>>> assert euler1 == 233168

Find the sum of all the even-valued terms in Fibonacci which do not exceed four million.

euler2 = sum(fib() | where(lambda x: x % 2 == 0) | take_while(lambda x: x < 4000000))

Find the difference between the sum of the squares of the first one hundred natural numbers and the square of the sum.

>>> square = map(lambda x: x ** 2)
>>> euler6 = sum(range(101)) ** 2 - sum(range(101) | square)
>>> assert euler6 == 25164150

Lazy evaluation

Using this module, you get lazy evaluation at two levels:

  • the object obtained by piping is a generator and will be evaluated only if needed,
  • within a series of pipe commands, only the elements that are actually needed will be evaluated.

To illustrate:

from itertools import count
from pipe import select, where, take


def dummy_func(x):
    print(f"processing at value {x}")
    return x


print("----- test using a generator as input -----")

print(f"we are feeding in a: {type(count(100))}")

res_with_count = (count(100) | select(dummy_func)
                             | where(lambda x: x % 2 == 0)
                             | take(2))

print(f"the resulting object is: {res_with_count}")
print(f"when we force evaluation we get:")
print(f"{list(res_with_count)}")

print("----- test using a list as input -----")

list_to_100 = list(range(100))
print(f"we are feeding in a: {type(list_to_100)} which has length {len(list_to_100)}")

res_with_list = (list_to_100 | select(dummy_func)
                             | where(lambda x: x % 2 == 0)
                             | take(2))

print(f"the resulting object is: {res_with_list}")
print(f"when we force evaluation we get:")
print(f"{list(res_with_list)}")

Which prints:

----- test using a generator as input -----
we are feeding in a: <class 'itertools.count'>
the resulting object is: <generator object take at 0x7fefb5e70c10>
when we force evaluation we get:
processing at value 100
processing at value 101
processing at value 102
processing at value 103
processing at value 104
[100, 102]
----- test using a list as input -----
we are feeding in a: <class 'list'> which has length 100
the resulting object is: <generator object take at 0x7fefb5e70dd0>
when we force evaluation we get:
processing at value 0
processing at value 1
processing at value 2
processing at value 3
processing at value 4
[0, 2]

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pipe-2.0.tar.gz (9.3 kB view hashes)

Uploaded source

Built Distribution

pipe-2.0-py3-none-any.whl (8.8 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page