Add draw_interactive() method

copulogram/Copulogram.py

@@ -6,16 +6,27 @@
 @author: Elias Fekhari
 """
 
-
+#%%
 import numpy as np
 import pandas as pd
 import seaborn as sns
 from matplotlib import rc
 import matplotlib.pyplot as plt
 
+# Interactive imports
+from itertools import product
+from bokeh.io import show
+from bokeh.layouts import gridplot
+from bokeh.models import (BasicTicker, Circle, ColumnDataSource,
+                          DataRange1d, Grid, LassoSelectTool, LinearAxis,
+                          Plot, ResetTool)
+from bokeh.transform import factor_cmap, linear_cmap
+from bokeh.palettes import Category10, Viridis
+
+
 class Copulogram:
     """
-    Interactive plot for multivariate distributions.
+    Draws a plot for multivariate distributions.
     The lower triangle is a matrixplot of the data (without transformation), 
     while the upper triangle is a matrixplot of the ranked data.
 
@@ -143,28 +154,123 @@ def draw(self,
         return copulogram
 
     def draw_interactive(self, 
-            kde_on_marginals=True,
-            quantile_contour_levels=None):
+            color='navy',
+            alpha=1.,
+            hue=None,
+            hue_palette=None,
+            marker='o',
+            subplot_size=5
+            ):
         """
         Draws the copulogram plot with a static or interactive option. 
 
         Parameters
         ----------
-        kde_on_marginals : Bool
-            Defines the type of plot on the diagonal. Histogram when 
-            the variable is set to False, kernel density estimation otherwise.
-        quantile_contour_levels : 1-d list of floats
-            When the variable takes a value, the contours of the quantiles 
-            defined by the variable are plotted.  
-
+        TBD
+        
         Returns
         -------
         copulogram : TBD
         """
 
+        df = self.data.copy(deep=True)
+        df_numeric = df._get_numeric_data()
+        rdf = df_numeric.rank()
+
+        plotted_cols = np.array(df_numeric.columns)
+        plotted_cols = np.delete(plotted_cols, np.where(plotted_cols == hue)).tolist()
+        dim = len(plotted_cols)
+
+        if hue is not None: 
+            df_numeric[hue] = df[hue]
+            rdf[hue] = df[hue]
+        source = ColumnDataSource(data=df_numeric)
+        rsource = ColumnDataSource(data=rdf)
+
+        plot_list = []
+        for i, (y, x) in enumerate(product(plotted_cols, plotted_cols)):
+            # Scatter plot
+            if hue is None: 
+                scatter_color = color
+            else: 
+                if df[hue].dtype =='O':
+                    if hue_palette is None:
+                        hue_palette = "Category10_3"
+                    scatter_color = factor_cmap(hue, hue_palette, sorted(df[hue].unique()))
+                else:
+                    if hue_palette is None:
+                        hue_palette="Spectral6"
+                    scatter_color = linear_cmap(hue, hue_palette, low=df[hue].min(), high=df[hue].max())
+
+
+            circle = Circle(x=x, y=y, fill_alpha=alpha, size=5, line_color=None,
+                            fill_color=scatter_color)
+            # Lower triangle
+            if (i%dim) <= (i//dim): # Column index smaller than row index (i.e., lower triangle)
+                # Define one empty plot
+                p = Plot(x_range=DataRange1d((df[x].min(), df[x].max())), y_range=DataRange1d((df[y].min(), df[y].max())),
+                        background_fill_color="#fafafa",
+                        border_fill_color="white", width=200, height=200, min_border=subplot_size)
+                r = p.add_glyph(source, circle)
+                # Delete diagonal plot
+                if (i%dim) == (i//dim):
+                    r.visible = False
+                    p.grid.grid_line_color = None
+            # Upper triangle
+            elif (i%dim) > (i//dim): 
+                # Define one empty plot
+                p = Plot(x_range=DataRange1d((rdf[x].min(), rdf[x].max())), y_range=DataRange1d((rdf[y].min(), rdf[y].max())),
+                        background_fill_color="#fafafa",
+                        border_fill_color="white", width=200, height=200, min_border=5)
+                r = p.add_glyph(rsource, circle)
+            p.x_range.renderers.append(r)
+            p.y_range.renderers.append(r)
+            # First column ticks
+            if i % dim == 0:
+                p.min_border_left = p.min_border + 4
+                p.width += 40
+                yaxis = LinearAxis(axis_label=y)
+                yaxis.major_label_orientation = "vertical"
+                p.add_layout(yaxis, "left")
+                yticker = yaxis.ticker
+            else:
+                yticker = BasicTicker()
+            p.add_layout(Grid(dimension=1, ticker=yticker))
+
+            # Last row ticks
+            if i >= dim * (dim-1):
+                p.min_border_bottom = p.min_border + 40
+                p.height += 40
+                xaxis = LinearAxis(axis_label=x)
+                p.add_layout(xaxis, "below")
+                xticker = xaxis.ticker
+            else:
+                xticker = BasicTicker()
+            p.add_layout(Grid(dimension=0, ticker=xticker))
+            p.add_tools(LassoSelectTool(), ResetTool())
+            plot_list.append(p)
+
+        gridp = gridplot(plot_list, ncols=dim)
+        show(gridp)
+
         return None
 
-## TODO
-# Include contours
-# Add interactive aspect
-# 
+##TODO:
+# Add docstrings
+# Remove the misleading yticks from the top left plot? Ideally we should add the index ticks on the top left of the plot
+# Add color bar on the interactive method using : https://docs.bokeh.org/en/latest/docs/examples/basic/data/color_mappers.html
+
+
+#%%
+if __name__ == "__main__":
+    #data = sns.load_dataset('iris')
+    import pandas as pd
+    data = pd.read_csv("../examples/data/wind_waves_ANEMOC_1H.csv", index_col=0)
+    data = data.iloc[:1000]
+
+    output = data["U_hub (m/s)"] ** 3 * ((np.pi / 180) * data["θ_wind (deg)"]) + (data["Hs (m)"] ** 2 * data["Tp (s)"]) / ((np.pi / 180) * data["θ_wave_new (deg)"])  
+    data['output'] = np.log10(output)
+
+    copulogram = Copulogram(data)
+    copulogram.draw_interactive(hue="output")
+# %%