pivoting to a series of 3 line charts.py

#Create a new field 'datetime', which converts strings to a datetime object
# datetime object will make our x-axis look good, showing appropriate labels like "Feb 2018" etc...
pyber_data_df['datetime'] = pd.to_datetime(pyber_data_df['date'])
# a datetime object *also* allows us to access the date part (2019-01-04) , ignoring the time-of-date part (05:39:20)
#Create another new field 'datedate' to store it
pyber_data_df['datedate'] = pyber_data_df['datetime'].dt.floor('d')

avg_fare_by_type_date = (pyber_data_df
                             #group by multiple fields by passing a list of those fields ; 
                             #each field will become a "level"  in the key which is distinct for each group
                             .groupby(['type', 'datedate'])
                             # get the average fare
                             .mean()['fare'])

(avg_fare_by_type_date
    #Pivot the dates from rows to columns; pandas needs diff columns to plot;
    #level 0 refers to unstacking the  first level in our grouped-by-results; 
    #the first level (level=0) is the `type` of city; we want to pivot aka `unstack` city type to new columns
    .unstack(level=0)
    #Some dates are missing values; use pandas 'fillna' or 'interpolate' to decide what to do with those missing values
    .fillna(method='ffill')
    #This data is very specific, take a "rolling average" of each column; with a "window size" of 10 days:
    .rolling(10).mean()
    .plot())