#!/usr/bin/env python
# coding: utf-8

# Visualizing my step data, from a Garmin [vívofit](https://buy.garmin.com/en-US/US/p/143405) (until about May 2016),  a [Forerunner 235](https://buy.garmin.com/en-US/US/p/529988) (May 2016 - June 2019) and a [Forerunner 935](https://buy.garmin.com/en-US/US/p/564291) (since June 2019).
# The data has been recorded to the [Apples Health](https://www.apple.com/ios/health/) app by [Garmin Connect](https://connect.garmin.com/).
# I exported the data from Apple Health to a [CSV](https://en.wikipedia.org/wiki/Comma-separated_values) file with the [QS Access](http://quantifiedself.com/qs-access-app/) app.

# In[1]:


# Load some modules we need
import pandas  # for data manipulation
import matplotlib.pylab as plt  # for plotting
import seaborn  # for plot styling
import dateutil.relativedelta  # for date calculations
import calendar  # for easy day and month names in the tick labels


# In[2]:


# Set some defaults
seaborn.set_style('dark')
plt.rcParams['figure.figsize'] = (16, 9)  # Size up figures a bit


# In[3]:


# Read in data
data = pandas.read_csv('steps.csv')


# In[4]:


# How does it look like?
data.head()


# In[5]:


data.tail()


# In[6]:


# Massage data
data.rename(columns={'Start': 'Date'}, inplace=True)
data.rename(columns={'Steps (count)': 'Steps'}, inplace=True)
data.Steps = data.Steps.astype(int)
data.drop('Finish', axis=1, inplace=True)


# In[7]:


# How doed it look now?
data.head()


# In[8]:


data.tail()


# In[9]:


# Make us a proper date column, based on https://stackoverflow.com/a/26763793
data['Date'] = pandas.to_datetime(data['Date'])


# In[10]:


# Make us a year, month and weekday colum, based on https://stackoverflow.com/q/48623332
data['Year'] = data.Date.dt.year
data['Month'] = data.Date.dt.month
data['Day'] = data.Date.dt.day
data['Weekday'] = data.Date.dt.dayofweek


# In[11]:


# How doed it look now?
data.head()


# In[12]:


yearwelookat = 2021


# In[13]:


# Give out totals to see if we have approximately the same as shown in Garmin Connect
print('In %s I made a total of %0.2f Million steps' % (yearwelookat,
                                                       data[(data.Year == yearwelookat)]['Steps'].sum() * 1e-6))


# In[14]:


if calendar.isleap(yearwelookat):
    dpy = 366
else:
    dpy = 365


# In[15]:


# Give out totals to see if we have approximately the same as shown in Garmin Connect
print('In %s I made %s steps per day' % (yearwelookat,
                                         int(round(data[(data.Year == yearwelookat)]['Steps'].sum() / dpy))))


# In[16]:


# Give out totals to see if we have approximately the same as shown in Garmin Connect
# In Garmin connect, go go 'Reports' in the sidebar, then 'Health & Fitness' > 'Steps'¨
# You can also try to mangle the export of *all* the Garmin data you get from 'Export your Data' here: https://www.garmin.com/en-US/account/datamanagement/
for month in sorted(data.Date.dt.month.unique()):
    print('In %s %s I made a total of %s steps' % (calendar.month_name[month],
                                                   yearwelookat,
                                                   data[(data.Year == yearwelookat) &
                                                        (data.Month == month)]['Steps'].sum()))


# In[17]:


# When were the five highest step counts?
data[data.Year == yearwelookat].sort_values('Steps', ascending=False).head()


# In[18]:


# When were the five lowest step counts?
data[data.Year == yearwelookat].sort_values('Steps').head()


# In[19]:


# Generate us a nice plot title, with help from https://stackoverflow.com/a/32083946/323100
Duration = dateutil.relativedelta.relativedelta(data[data.Year == yearwelookat].Date.max(),
                                                data[data.Year == yearwelookat].Date.min())


# In[20]:


# Generate us some nice tick labels
Ticklabels_weekday = ['%s\n~%g steps' % (d, round(s / 1e3, 1) * 1e3)
                      for (s, d) in zip(data[data.Year == yearwelookat].groupby('Weekday').Steps.mean(),
                                        list(calendar.day_name))]
# Generate us some nice tick labels
Ticklabels_month = ['%s\n~%g steps' % (m, round(s / 1e3, 1) * 1e3)
                    for (s, m) in zip(data[data.Year == yearwelookat].groupby('Month').Steps.mean(),
                                      list(calendar.month_name[1:]))]
# Generate us some nice tick labels
Ticklabels_year = ['%s\n~%g steps' % (y, round(s / 1e3, 1) * 1e3)
                   for (s, y) in zip(data[data.Year == yearwelookat].groupby('Year').Steps.mean(),
                                     data[data.Year == yearwelookat].Year.unique())]


# In[21]:


# Plot the steps per weekday
seaborn.boxplot(data=data[data.Year == yearwelookat],
                x='Weekday',
                y='Steps',
                fliersize=0)
seaborn.stripplot(data=data[data.Year == yearwelookat],
                  x='Weekday',
                  y='Steps',
                  jitter=True,
                  linewidth=1)
plt.gca().set_xticklabels(Ticklabels_weekday)
plt.title('Weekday steps for %s months and %s days in %s' % (Duration.months,
                                                             Duration.days,
                                                             yearwelookat))
plt.gca().xaxis.label.set_visible(False)
seaborn.despine()
plt.show()


# In[22]:


# Plot the steps per month
seaborn.boxplot(data=data[data.Year == yearwelookat],
                x='Month',
                y='Steps',
                fliersize=0)
seaborn.stripplot(data=data[data.Year == yearwelookat],
                  x='Month',
                  y='Steps',
                  jitter=True,
                  linewidth=1)
plt.gca().set_xticklabels(Ticklabels_month)
plt.title('Monthly steps for %s months and %s days in %s' % (Duration.months,
                                                             Duration.days,
                                                             yearwelookat))
plt.gca().xaxis.label.set_visible(False)
seaborn.despine()
plt.show()


# In[ ]: