pipelining generators like python, not like shell
Project description
Rohrleitung
===========
Rohrleitung (ger.): _pipeline_
Didn't like the shell-like approach the other pipelining packages took, where you
would concatenate specially written generators with the UNIX pipe `|`. While that's
great for the CLI and short shell scripts, I wanted a more programmable way.
Worth mentioning is surely [genpipeline](https://github.com/fkarb/genpipeline) that
implements coroutine-based pipelines as presented by [David Beazley's Coroutines
intro](http://www.dabeaz.com/coroutines/). But this means you have to do everything
in coroutines, I built Rohrleitung so I could just use classic generator-based
filters. Look into both and decide which way is cleaner for your environment.
ISC licenced, see the LICENCE file.
Examples:
---------
from functools import partial
# pip install toolz
from toolz.curried import interpose
from rohrleitung import Pipeline, L
def three_n_plus_one(n):
if n % 2:
return 3 * n + 1
else:
return n / 2
@L
def collatz_length(n, l=0):
if int(n) < 1:
raise ValueError('Nope')
if n == 1:
return l
else:
# L changes call signatures of decorated function, so in recursive ones
# you have to adapt.
return collatz_length.__wrapped__(three_n_plus_one(n), l + 1)
# building pipelines in a programmatic way, without immediately executing them
# standard list manipulation can change pipeline on the fly
pipeline = [
partial(filter, lambda x: x % 2), # Standard python's filter function
lambda y: (2 * x for x in y), # Like using L, but manually
L(lambda x: x ** 3), # Using L helper function
collatz_length,
]
pipeline.append(interpose('a'))
print(list(Pipeline(pipeline, range(10))))
a = Pipeline([L(lambda x: 2 * x), L(lambda x: x + 1), L(lambda x: x ** 2)])
for i, k in enumerate(a(range(3))):
print("(2x+1)^2, x={}: {}".format(i, k))
# With toolz' curried filter function we don't need partial
from toolz.curried import filter
newpipeline = [
collatz_length,
filter(lambda x: not(x % 2)),
L(lambda x: bin(x)),
L(lambda x: x.count('1'))
]
p1 = Pipeline(newpipeline)
for i in p1(range(1, 10)):
print(i)
# Reuse same pipeline
for i in p1(range(1, 3)):
print(i)
# Or modify it, then use again
p1.pipeline.insert(2, L(lambda x: x + 1))
for i in p1(range(1, 10)):
print(i)
p2 = Pipeline(pipeline)
p3 = 2 * (p2 + [L(lambda x: int(str(x), 16))]) + p1
# Alternativly 2 * p2 | p1 if you prefer shell syntax
for i in p3(range(1, 10)):
print(i)
# Slicing and cutting pipelines on the fly
for i in p1[1:4:2](range(10)):
print(i)
===========
Rohrleitung (ger.): _pipeline_
Didn't like the shell-like approach the other pipelining packages took, where you
would concatenate specially written generators with the UNIX pipe `|`. While that's
great for the CLI and short shell scripts, I wanted a more programmable way.
Worth mentioning is surely [genpipeline](https://github.com/fkarb/genpipeline) that
implements coroutine-based pipelines as presented by [David Beazley's Coroutines
intro](http://www.dabeaz.com/coroutines/). But this means you have to do everything
in coroutines, I built Rohrleitung so I could just use classic generator-based
filters. Look into both and decide which way is cleaner for your environment.
ISC licenced, see the LICENCE file.
Examples:
---------
from functools import partial
# pip install toolz
from toolz.curried import interpose
from rohrleitung import Pipeline, L
def three_n_plus_one(n):
if n % 2:
return 3 * n + 1
else:
return n / 2
@L
def collatz_length(n, l=0):
if int(n) < 1:
raise ValueError('Nope')
if n == 1:
return l
else:
# L changes call signatures of decorated function, so in recursive ones
# you have to adapt.
return collatz_length.__wrapped__(three_n_plus_one(n), l + 1)
# building pipelines in a programmatic way, without immediately executing them
# standard list manipulation can change pipeline on the fly
pipeline = [
partial(filter, lambda x: x % 2), # Standard python's filter function
lambda y: (2 * x for x in y), # Like using L, but manually
L(lambda x: x ** 3), # Using L helper function
collatz_length,
]
pipeline.append(interpose('a'))
print(list(Pipeline(pipeline, range(10))))
a = Pipeline([L(lambda x: 2 * x), L(lambda x: x + 1), L(lambda x: x ** 2)])
for i, k in enumerate(a(range(3))):
print("(2x+1)^2, x={}: {}".format(i, k))
# With toolz' curried filter function we don't need partial
from toolz.curried import filter
newpipeline = [
collatz_length,
filter(lambda x: not(x % 2)),
L(lambda x: bin(x)),
L(lambda x: x.count('1'))
]
p1 = Pipeline(newpipeline)
for i in p1(range(1, 10)):
print(i)
# Reuse same pipeline
for i in p1(range(1, 3)):
print(i)
# Or modify it, then use again
p1.pipeline.insert(2, L(lambda x: x + 1))
for i in p1(range(1, 10)):
print(i)
p2 = Pipeline(pipeline)
p3 = 2 * (p2 + [L(lambda x: int(str(x), 16))]) + p1
# Alternativly 2 * p2 | p1 if you prefer shell syntax
for i in p3(range(1, 10)):
print(i)
# Slicing and cutting pipelines on the fly
for i in p1[1:4:2](range(10)):
print(i)
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
rohrleitung-0.1.1.tar.gz
(3.9 kB
view hashes)