Skip to content

Latest commit

 

History

History
211 lines (146 loc) · 6.02 KB

README.rst

File metadata and controls

211 lines (146 loc) · 6.02 KB

Parameterize Jobs

https://codecov.io/gh/ClimateImpactLab/parameterize_jobs/branch/master/graph/badge.svg?token=DUDCDOPYYC Documentation Status

parameterize_jobs is a lightweight, pure-python toolkit for concisely and clearly creating large, parameterized, mapped job specifications.

Features

  • Expand a job's dimensionality by multiplying ComponentSet, Constant, or ParallelComponentSet objects
  • Extend the number of jobs by adding ComponentSet, Constant, or ParallelComponentSet objects
  • Jobs are provided to functions as dictionaries of parameters
  • The helper decorator @expand_kwargs turns these kwarg dictionaries into named argument calls
  • Works seamlessly with many task running frameworks, including dask's client.map and profiling tools

TODOs

View and submit issues on the issues page.

Quickstart

ComponentSet objects are the base objects, and can be defined with any number of named iterables:

>>> import parameterize_jobs as pjs

>>> a = pjs.ComponentSet(a=range(5))
>>> a
<ComponentSet {'a': 5}>

These objects have defined lengths (if the provided iterable has a defined length), and can be indexed and iterated over:

>>> len(a)
5

>>> a[0]
{'a': 0}

>>> list(a)
[{'a': 0},
 {'a': 1},
 {'a': 2},
 {'a': 3},
 {'a': 4}]

Adding two ComponentSet objects extends the total job length

>>> a2 = pjs.ComponentSet(a=range(3))

>>> a+a2
<MultiComponentSet [{'a': 5}, {'a': 3}]>

>>> len(a+a2)
8

>>> list(a+a2)

[{'a': 0},
 {'a': 1},
 {'a': 2},
 {'a': 3},
 {'a': 4},
 {'a': 0},
 {'a': 1},
 {'a': 2}]

Multiplying two ComponentSet objects expands their dimensionality:

>>> b = pjs.ComponentSet(b=range(3))

>>> a*b
<ComponentSet {'a': 5, 'b': 3}>

>>> len(a*b)
15

>>> (a*b)[-1]
{'a': 4, 'b': 2}

>>> list(a*b)
[{'a': 0, 'b': 0},
 {'a': 0, 'b': 1},
 {'a': 0, 'b': 2},
 {'a': 1, 'b': 0},
 {'a': 1, 'b': 1},
 {'a': 1, 'b': 2},
 {'a': 2, 'b': 0},
 {'a': 2, 'b': 1},
 {'a': 2, 'b': 2},
 {'a': 3, 'b': 0},
 {'a': 3, 'b': 1},
 {'a': 3, 'b': 2},
 {'a': 4, 'b': 0},
 {'a': 4, 'b': 1},
 {'a': 4, 'b': 2}]

These parameterized job specifications can be used in mappable jobs. The helper decorator expand_kwargs modifies a function to accept a dictionary and expands them into keyword arguments:

>>> @pjs.expand_kwargs
... def my_simple_func(a, b, c=1):
...     return a * b * c

>>> list(map(my_simple_func, a*b))
[0, 0, 0, 0, 0, 0, 1, 2, 3, 4, 0, 2, 4, 6, 8, 0, 3, 6, 9, 12]

Jobs do not have to be the combinatorial product of all components:

>>> ab1 = pjs.ComponentSet(a=[0, 1], b=[0, 1])
>>> ab2 = pjs.ComponentSet(a=[10, 11], b=[-1, 1])

>>> list(map(my_simple_func, ab1 + ab2))
[0, 0, 0, 1, -10, -11, 10, 11]

A Constant object is simply a ComponentSet object defined with single values passed as keyword arguments rather than iterables passed as keyword arguments:

>>> c = pjs.Constant(c=5)

>>> list(map(my_simple_func, (ab1 + ab2) * c))
[0, 0, 0, 5, -50, -55, 50, 55]

A ParallelComponentSet object is simply a MultiComponentSet object where each Component is a Constant object.

>>> pcs = pjs.ParallelComponentSet(a = [1, 2],
                           b = [10, 20])

>>> list(map(my_simple_func, pcs))
[10, 40]

Arbitrarily complex combinations of ComponentSets can be created:

>>> c1 = pjs.Constant(c=1)
>>> c2 = pjs.Constant(c=2)

>>> list(map(my_simple_func, (ab1 + ab2) * c1 + (ab1 + ab2) * c2))
[0, 0, 0, 1, -10, -11, 10, 11, 0, 0, 0, 2, -20, -22, 20, 22]

Anything can be inside a ComponentSet iterable, including data, functions, or other objects:

>>> transforms = (
...     pjs.Constant(transform=lambda x: x, transform_name='linear')
...     + pjs.Constant(transform=lambda x: x**2, transform_name='quadratic'))
...

>>> fps = pjs.Constant(
...     read_pattern='source/my-fun-data_{year}.csv',
...     write_pattern='transformed/my-fun-data_{transform_name}_{year}.csv')

>>> years = pjs.ComponentSet(year=range(1980, 2018))

>>> @pjs.expand_kwargs
... def process_data(read_pattern, write_pattern, transform, transform_name, year):
...
...     df = pd.read_csv(read_pattern.format(year=year))
...
...     transformed = transform(df)
...
...     transformed.to_csv(
...         write_pattern.format(
...             transform_name=transform_name,
...             year=year))
...

>>> _ = list(map(process_data, transforms * fps * years))

This works seamlessly with dask's client.map to provide intuitive job parameterization:

>>> import dask.distributed as dd
>>> client = dd.LocalClient()
>>> futures = client.map(my_simple_func, (ab1 + ab2) * c1 + (ab1 + ab2) * c2)
>>> dd.progress(futures)