Interacting with IPython Notebook widgets

Interactive IPython notebook widgets are, at the time of writing (July 2015), an experimental feature. I, and as far as I know, many other people, hope that this feature will remain. In a nutshell, the widgets let you select values as you would with HTML forms. This includes sliders, drop-down boxes, and check boxes. As you can read, these widgets are very convenient for visualizing the weather data I introduced in Chapter 1, Laying the Foundation for Reproducible Data Analysis.

How to do it...

1. Import the following:

   import seaborn as sns
   import numpy as np
   import pandas as pd
   import matplotlib.pyplot as plt
   from IPython.html.widgets import interact
   from dautil import data
   from dautil import ts

2. Load the data and request inline plots:

   %matplotlib inline
   df = data.Weather.load()

3. Define the following function, which displays bubble plots:

   def plot_data(x='TEMP', y='RAIN', z='WIND_SPEED', f='A', size=10, cmap='Blues'):
       dfx = df[x].resample(f)
       dfy = df[y].resample(f)
       dfz = df[z].resample(f)
       
       bubbles = (dfz - dfz.min())/(dfz.max() - dfz.min())
       years = dfz.index.year
       sc = plt.scatter(dfx, dfy, s= size * bubbles + 9, c = years,
                   cmap=cmap, label=data.Weather.get_header(z), alpha=0.5)
       plt.colorbar(sc, label='Year')
       
       freqs = {'A': 'Annual', 'M': 'Monthly', 'D': 'Daily'}
       plt.title(freqs[f] + ' Averages')
       plt.xlabel(data.Weather.get_header(x))
       plt.ylabel(data.Weather.get_header(y))
       plt.legend(loc='best')

4. Call the function we just defined with the following code:

   vars = df.columns.tolist()
   freqs = ('A', 'M', 'D')
   cmaps = [cmap for cmap in plt.cm.datad if not cmap.endswith("_r")]
   cmaps.sort()
   interact(plot_data, x=vars, y=vars, z=vars, f=freqs, size=(100,700), cmap=cmaps)

5. This is one of the recipes where you really should play with the code to understand how it works. The following is an example bubble plot:
   
6. Define another function (actually, it has the same name), but this time the function groups the data by day of year or month:

   def plot_data(x='TEMP', y='RAIN', z='WIND_SPEED', groupby='ts.groupby_yday', size=10, cmap='Blues'):
       if groupby == 'ts.groupby_yday':
           groupby = ts.groupby_yday
       elif groupby == 'ts.groupby_month':
           groupby = ts.groupby_month
       else:
           raise AssertionError('Unknown groupby ' + groupby)
           
       dfx = groupby(df[x]).mean()
       dfy = groupby(df[y]).mean()
       dfz = groupby(df[z]).mean()
       
       bubbles = (dfz - dfz.min())/(dfz.max() - dfz.min())
       colors = dfx.index.values
       sc = plt.scatter(dfx, dfy, s= size * bubbles + 9, c = colors,
                   cmap=cmap, label=data.Weather.get_header(z), alpha=0.5)
       plt.colorbar(sc, label='Day of Year')
       
       by_dict = {ts.groupby_yday: 'Day of Year', ts.groupby_month: 'Month'}
       plt.title('Grouped by ' + by_dict[groupby])
       plt.xlabel(data.Weather.get_header(x))
       plt.ylabel(data.Weather.get_header(y))
       plt.legend(loc='best')

7. Call this function with the following snippet:

   groupbys = ('ts.groupby_yday', 'ts.groupby_month')
   interact(plot_data, x=vars, y=vars, z=vars, groupby=groupbys, size=(100,700), cmap=cmaps)

Refer to the following plot for the end result:

My first impression of this plot is that the temperature and wind speed seem to be correlated. The source code is in the Interactive.ipynb file in this book's code bundle.

See also

• The documentation on interactive IPython widgets at https://ipython.org/ipython-doc/dev/api/generated/IPython.html.widgets.interaction.html (retrieved July 2015)