Quickstart¶

partialwrap: a small Python library providing wrappers for external executables and Python functions so that they can easily be partialised with Python’s functools.partial.

partialwrap is a Python library providing easy wrapper functions to use with Python’s functools.partial(). They allow to use any external executable as well as any Python function with arbitrary arguments and keywords to be used with libraries that call functions simply in the form func(x). This includes the functions of the scipy.optimize package or external packages such as emcee or pyeee, and allows the use of distributed function evaluations with Python’s multiprocessing or via mpi4py.

The complete documentation for partialwrap is available from Read The Docs.

Quick usage guide¶

Context¶

The easiest wrapper in partialwrap is for Python functions.

Consider the Rastrigin function (https://en.wikipedia.org/wiki/Rastrigin_function), which is a popular function for performance testing of optimization algorithms: $$y = a \cdot n + \sum_i^n (x_i^2 - a \cos(b \cdot x_i))$$

import numpy as np
def rastrigin(x, a, b=2.*np.pi):
return a*len(x) + np.sum(x**2 - a*np.cos(b*x))


The Rastrigin function has a global minimum of $$0$$ at all $$x_i = 0$$. $$a$$ influences mainly the depths of the (local and global) minima, whereas $$b$$ influences mainly the size of the minima.

A common form uses $$a = 10$$ and $$b = 2 \pi$$. The parameters $$x_i$$ should then be in the interval $$[-5.12,5.12]$$.

import numpy as np
def rastrigin1(x):
return 10.*len(x) + np.sum(x**2 - 10.*np.cos(2.*np.pi*x))


One could use the scipy.optimize package to try to find the minimum of the Rastrigin function:

import scipy.optimize as opt
x0 = np.array([0.5, 0.5])
print(res.x)
# retruns: [0.99497463 0.99498333]
res = opt.minimize(rastrigin1, x0, method='BFGS')
print(res.x)
# returns: [-7.82960597e-09 -7.82960597e-09]


scipy.optimize allows passing arguments to the function to minimize. One could hence use the general Rastrigin function rastrigin (instead of rastrigin1) to get the same result:

res = opt.minimize(rastrigin, x0, args=(10.), method='BFGS')


Simple Python functions¶

Not all optimizers allow the passing of arguments. And notably scipy.optimize does not allow the passing of keyword arguments, such as $$b$$ in the case of rastrigin. One can use partial() of Python’s functools in this case:

from functools import partial

def call_func_arg_kwarg(func, a, b, x):
return func(x, a, b=b)

# Partialise function with fixed parameters
a = 5.
b = 4.*np.pi
partial_rastrigin = partial(call_func_arg_kwarg, rastrigin, a, b)

res = opt.minimize(partial_rastrigin, x0, method='BFGS')


Figuratively speaking, partial() passes $$a$$ and $$b$$ to the function call_func_arg_kwarg already during definition. minimize() can then simply call it as partial_rastrigin(x), which finalizes the call to rastrigin(x, a, b=b).

partialwrap provides a convenience function function_wrapper() passing all arguments, given as a list, and keyword arguments, given as a dict, to arbitrary functions:

from partialwrap import function_wrapper

args   = [20.]
kwargs = {'b': 1.*np.pi}
rastra = partial(function_wrapper, rastrigin, args, kwargs)

res = opt.minimize(rastra, x0, method='BFGS')


Or in short, of course:

from partialwrap import function_wrapper

rastra = partial(function_wrapper, rastrigin, [20.], {'b': 1.*np.pi})
res = opt.minimize(rastra, x0, method='BFGS')


A common case in numerical optimization are bound parameters and specifically the exclusion of some well-known or correlated parameters from optimization. partialwrap provides a convenience function function_mask_wrapper() to include only the masked parameters in the function evaluation:

from partialwrap import function_mask_wrapper

x0      = np.array([0.5, 0.0001, 0.5])

xout       = x0.copy()


External executables¶

partialwrap provides two wrapper functions to work with external executables: partialwrap.exe_wrapper() and partialwrap.exe_mask_wrapper().

partialwrap writes the sampled parameter sets into files that can be read by the external program. The external program should write its result to a file that will then be read by partialwrap in return. That means partialwrap needs to have a function parameterwriter that writes the parameter file parameterfile needed by the executable exe. It then needs to have a function outputreader for reading the output file outputfile of the external executable exe.

Consider for simplicity an external Python program (e.g. rastrigin1.py) that calculates the Rastrigin function with $$a = 10$$ and $$b = 2 \pi$$, reading in an arbitrary number of parameters $$x_i$$ from a parameterfile = params.txt and writing its output into an outputfile = out.txt:

# File: rastrigin1.py

# Rastrigin function a=10, b=2*pi
import numpy as np
def rastrigin1(x):
return 10.*len(x) + np.sum(x**2 - 10.*np.cos(2.*np.pi*x))

# calc function
y = rastrigin1(x)

# write output file
with open('out.txt', 'w') as ff:
print(y, file=ff)


This program can be called on the command line (if params.txt is present) with:

python rastrigin1.py


The external program can be used with Python’s functools.partial() and the wrapper function partialwrap.exe_wrapper():

from functools import partial
from partialwrap import exe_wrapper, standard_parameter_writer, standard_output_reader

rastrigin_exe  = ['python3', 'rastrigin1.py']
parameterfile  = 'params.txt'
outputfile     = 'out.txt'
rastrigin_wrap = partial(exe_wrapper, rastrigin_exe,
parameterfile, standard_parameter_writer,

x0  = [0.1, 0.2, 0.3]
res = opt.minimize(rastrigin_wrap, x0, method='BFGS')


partialwrap.standard_parameter_reader() and partialwrap.standard_parameter_writer() are convenience functions that read and write one parameter per line in a file without a header. The function partialwrap.standard_output_reader() simply reads one value from a file without header. The empty dictionary at the end is explained in the userguide.

One can easily imagine to replace the Python program rastrigin1.py by any compiled executable from C, Fortran or alike. See the userguide for details.

Installation¶

The easiest way to install is via pip:

pip install partialwrap


Requirements¶

partialwrap is distributed under the MIT License. See the LICENSE file for details.