Homec4science
Diffusion SP4E-TB-TL-FR (w12ex2_FR) Browse

R7571/work/week5/series/dumper.py
7c0ce725b1b5w12ex2_FR

Last Change
Raw File

dumper.py
View Options

# -*- coding: utf-8 -*-
# @Author: Theo Lemaire
# @Date: 2018-10-16 11:52:15
# @Last Modified by: Theo Lemaire
# @Last Modified time: 2018-10-16 14:31:19
from argparse import ArgumentParser
import matplotlib.pyplot as plt
from series import ComputeArithmetic, ComputePi
class DumperSeries:
def __init__(self, series):
self.series = series
def dump(self):
raise Exception("pure virtual function")
class PrintSeries(DumperSeries):
def __init__(self, series, frequency, maxiter):
super().__init__(series)
self.frequency = frequency
self.maxiter = maxiter
def dump(self):
for n in range(1, self.maxiter + 1, self.frequency):
s = self.series.compute(n)
try:
delta = self.series.getAnalyticPrediction() - s
delta_str = ' (Delta = {:.2e})'.format(delta)
except Exception as e:
delta_str = ''
print('{}({}) = {:.5f} {}'.format(self.series, n, s, delta_str))
class PlotSeries(DumperSeries):
def __init__(self, series, maxiter):
super().__init__(series)
self.maxiter = maxiter
def dump(self, fs=15, lw=2):
iterations = range(1, self.maxiter + 1)
fig, ax = plt.subplots(figsize=(10, 6))
ax.set_title('{} convergence'.format(self.series), fontsize=fs)
ax.plot(iterations, [self.series.compute(n) for n in iterations],
label='series', linewidth=lw)
try:
ax.axhline(self.series.getAnalyticPrediction(),
linestyle='--', linewidth=lw, c='k', label='analytic prediction')
except Exception as e:
print('series does not have constant analytical prediction')
ax.set_xlabel('# iterations', fontsize=fs)
ax.set_ylabel('{} value []'.format(self.series), fontsize=fs)
for item in ax.get_xticklabels() + ax.get_yticklabels():
item.set_fontsize(fs)
ax.legend(frameon=False, fontsize=fs)
plt.show()
def main():
# Define argument parser
ap = ArgumentParser()
ap.add_argument('-n', '--niter', type=int, default=5, help='Number of iterations')
ap.add_argument('-s', '--seriestype', type=str, default='Ar', help='Series type (Ar or Pi)')
ap.add_argument('-f', '--frequency', type=int, default=1, help='Dumping frequency')
ap.add_argument('-p', '--plot', default=False, action='store_true', help='Plot results')
# Parse arguments
args = ap.parse_args()
n = args.niter
f = args.frequency
s = args.seriestype.lower()
if n < 0:
raise ValueError('N must be positive')
# Instantiate appropriate series
if s == 'ar':
series = ComputeArithmetic()
elif s == 'pi':
series = ComputePi()
else:
raise ValueError('Unknown series type')
# Instantiate appropriate dumper series
if args.plot:
ps = PlotSeries(series, n)
else:
ps = PrintSeries(series, f, n)
# Call dump method
ps.dump()
# Initial complexity:
# 1 series output at specific N: S = sum(k=1 -> N, ...) -> O(N)
# series output from 1 to maxiter: sum(N=1 -> maxiter, sum(k=1 -> N, ...))
# -> O(sum(N=1 -> maxiter, O(N))
# -> O(maxiter*(maxiter + 1) / 2)
# New complexity: O(maxiter) !
# With reversed computation of Pi, we cannot optimize the complexity of the dump method
# by storing intermediate values, hence complexity is bound to O(maxiter*(maxiter + 1) / 2)
if __name__ == '__main__':
main()
c4science · Help