import numpy as np
import matplotlib
import matplotlib.pyplot as plt

#S3  417 http://browser.geekbench.com/geekbench3/8570680
#S4: 716 https://browser.geekbench.com/android_devices/108
#S5: 931 https://browser.geekbench.com/android_devices/163
#S6: 1295 https://browser.geekbench.com/android_devices/209
#S7: 1811 https://browser.geekbench.com/android_devices/220
#S8: 1953 https://browser.geekbench.com/android_devices/376
# S9: 3739 https://www.tek.no/artikler/test-samsung-galaxy-s9/431900/2
samsung_scores = np.array([417, 716, 931, 1295, 1811, 1953, 3739])

# Release dates of Galaxy S (Rounded down to month)
# (source: google "Samsung Galaxy S<version> release date")
samsung_versions = np.array([2012+5./12, 2013+4/12., 2014+4/12., 2015+4/12.,
                             2016+3./12, 2017+4/12., 2018+3/12.])

p_samsung =  plt.plot(samsung_versions, samsung_scores, '*',
         label='Samsung Galaxy S <Year>')

# Now we fit the data, except the latest, and see how good a prediction it is
samsung_versions_shifted = samsung_versions - samsung_versions[0]*np.ones_like(samsung_versions)
samsung_poly_coeff = np.polyfit(samsung_versions_shifted[:-2],
    np.log(samsung_scores[:-2]), 1)

print(np.exp(samsung_poly_coeff[1])*np.exp(samsung_versions_shifted*samsung_poly_coeff[0]))

plt.plot(samsung_versions,
         np.exp(samsung_poly_coeff[1])*np.exp(samsung_versions_shifted*samsung_poly_coeff[0]),
         '--', color=p_samsung[0].get_color(),
         label='$%.2f (%.2f)^{\\mathrm{Year} - %.2f}$'
         % (np.exp(samsung_poly_coeff[1]),
          np.exp(samsung_poly_coeff[0]),
          samsung_versions[0]))
# iphone 4s 284 https://browser.geekbench.com/ios_devices/6
# iphone 5 754 https://browser.geekbench.com/ios_devices/20
# iphone 5s 1267 https://browser.geekbench.com/ios_devices/28
# iphone 6 1360 https://browser.geekbench.com/ios_devices/33
# iphone 6s 2214 https://browser.geekbench.com/ios_devices/38
# iphone 7 3388 https://browser.geekbench.com/ios_devices/44
# iphone 8 4217 https://browser.geekbench.com/ios_devices/50

iphone_scores = np.array([284, 754, 1267, 1360, 2214, 3388, 4217])

# Release dates of iPhone (Rounded down to month)
# (source: google "Iphone <version> release date")
iphone_versions = np.array(
    [2011. + 10/12., 2012 + 9/12., 2013+ 9/12., 2014+9/12.,
                   2015+9/12., 2016+9/12., 2017+9/12.])


p_apple = plt.plot(iphone_versions, iphone_scores,
                   'o', label='Apple iPhone <Year>')

iphone_versions_shifted = iphone_versions - iphone_versions[0]*np.ones_like(iphone_versions)
apple_poly_coeff = np.polyfit(iphone_versions_shifted[:-2],
    np.log(iphone_scores[:-2]), 1)


plt.plot(iphone_versions,
         np.exp(apple_poly_coeff[1])*np.exp(iphone_versions_shifted*apple_poly_coeff[0]),
         '--', color=p_apple[0].get_color(),
         label='$%.2f (%.2f)^{\\mathrm{Year} - %.2f}$'
         % (np.exp(apple_poly_coeff[1]),
          np.exp(apple_poly_coeff[0]),
          iphone_versions[0]))


plt.xlabel('Year')
plt.ylabel('Geekbench 4 Single score')
plt.legend()
plt.grid("on")
plt.show()