Using_plotly.py

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

# In[1]:


import numpy as np
import pandas as pd
import plotly

from plotly.subplots import make_subplots

plotly.__version__


# ***See Returns.ipynb for an example of using plotly with pandas Series and DataFrames***

# ***Reading https://plot.ly/python/creating-and-updating-figures/ is worthwhile***

# ## Simple example

# In[2]:


import plotly
import plotly.graph_objs as go


trace0 = go.Scatter(
    x=[1, 2, 3, 4],
    y=[10, 15, 13, 17]
)
trace1 = go.Scatter(
    x=[1, 2, 3, 4],
    y=[16, 5, 11, 9]
)
data = [trace0, trace1]

go.Figure(data).show()


# ## Wrapping matplotlib

# In[3]:




# Matplotlib
import matplotlib.pyplot as plt
import plotly.tools as tls

## Generating the data..
x =  np.linspace(-2*np.pi, 2*np.pi, 1000)


fig = plt.figure()                     # Initiate a figure

a1 = fig.add_subplot(221)              # Figure has 1 row 1 column ; choose the first subplot
a1.plot(x,np.sin(x))                   # Plot x and y in the chosen subplot
a1.grid()

a2 = fig.add_subplot(224)
a2.plot(x,np.cos(x))
a2.grid()


# In[4]:


# Now we make the matplotlib figure interactive using plotly

plotly_fig = tls.mpl_to_plotly(fig)

plotly_fig.show()


# ## Plotting columns of a DataFrame

# In[5]:


def PlotData(df):
    pdata = [{'x':df.index,'y':df[col],'name':col} for col in df.columns]
    return(pdata)


# In[6]:


ETFRets = pd.read_pickle('/home/vpoduri/Site/ETFRets.pkl')
ETFRets.head()


# In[7]:


# Plotting
pdata = PlotData(ETFRets)
layout = go.Layout(title = 'ETF Returns',yaxis=dict(tickformat='.1%',title='Return'))

pFig = go.Figure(pdata,layout)   # creates a Figure object
pFig.show()


# ## Elaborate subplots

# In[8]:


# Rolling Correlations
RCorr = ETFRets.rolling(12).corr(pairwise=True) # Really nice syntax!

nr = 2 ; nc = 4
fig = make_subplots(rows=nr, cols=nc, start_cell="top-left",
                   subplot_titles=list(RCorr.columns))


# In[9]:


foo = RCorr.reset_index().rename(columns={"level_1":"ETF"})
foo.head(10)


# In[10]:


# Visualization with subplots. Quite finicky.
ETFList = foo.columns[2:]

for j in range(len(ETFList)):
    ticker = ETFList[j]
    df1 = foo.loc[foo["ETF"]==ticker,:]
    df1 = df1.drop(columns=["ETF",ticker])  # All values are 1
    
    fr,fc = np.unravel_index(j,(nr,nc),order='F')  # Calculates subplot index, Fortran style (column major order)
    
    plines = [go.Scatter(x=df1["Date"],y=df1[col],name=col) for col in ETFList.difference([ticker])]
    _=[fig.append_trace(pl,row=int(fr+1),col=int(fc+1)) for pl in plines]
    
                            # only append_trace seems to do the trick with subplots
                            # np.int64 for row and col cause errors, so have to use int
        
fig.show()


# ## plotly express

# Plotly express was inspired by ggplot and Seaborn. It is a wrapper around lower-level classes and methods. It works best if the data are in a long form DataFrame. 
# 
# See its [tutorial](https://plot.ly/python/plotly-express/) for details.

# In[11]:


import plotly.express as px

RC2 = RCorr.reset_index().rename(columns={'level_1':'ETF'})
RC2 = pd.melt(RC2,id_vars=['Date','ETF'],var_name='With',value_name='Corr')

px.line(RC2,x='Date',y='Corr',color='With',facet_col='ETF',facet_col_wrap=4)   # Remarkably elegant


# In[48]:


RC2.head()