[FIXED] Create heatmap graph in matplotlib

Issue

I have a two-dimensional numpy array that is a set of values ​​for some parameter. Part of the data in this array is missing and is of type numpy.nan. I would like to plot a "heatmap" type graph using a heatmap, pcolor or matshow or imshow, in this case it doesn't matter to me. The color will indicate the value of the parameter. But I faced with some difficulties making this task a time consuming job. The passing of a part of the data in the array is due to my data field for displaying the heatmap is not a rectangle. It is inscribed in a figure that approximates a circle. I assign NaN values ​​to those points in the data matrix that do not belong to the circle. Those in this matrix, the nan-values have points that are not conventionally a circle. Only some columns and rows are filled with data completely. This approach helps me to solve the problem how it is possible to display a heatmap of some arbitrary shape, and not just a rectangle. The figure shows what I want to get approximately.

In this regard, I have two questions:

1. When solving the problem with my approach, what heatmap setting can set the color for nan values, since white is not always what I want. I would like to be able to set it of my choice.
2. A more global question. Is there an easier way to plot a heatmap for which I can simply specify the x and y coordinates of the data points, and the parameter value (e.g. z) that will be converted to color. I would like not to create a two-dimensional array with part of the missing data, but simply pass array-like objects x, y and z and get such a heatmap. But I want this graph to look like a solid field of rectangular tiles, as shown in my picture, and not like some kind of scatterplot, with multi-colored marker points. In my case, the code that creates the graph looks like this:

    import numpy as np
    import pandas as pd
    import matplotlib as mpl
    import matplotlib.pyplot as plt
    import datetime
    %matplotlib inline
    
    #retrieving and transforming data... 
    
    #creating two-dimensional data array 
    #create_matrix_data_vect - some custom function
    data = create_matrix_data_vect(param_data=data.z_parameter, \
                                positions_data=positions_data_normal, \
                                data_shape=data_array_shape)
    data.shape

(12, 12) - the size can be much larger 

The data array looks like this:

array([[  nan,   nan,   nan,   nan, 201. , 195. , 203. , 196. ,   nan,
          nan,   nan,   nan],
       [  nan,   nan,   nan, 193. , 172. , 181. , 212. ,  91.3, 179. ,
          nan,   nan,   nan],
       [  nan,   nan, 188. , 208. , 180. , 201. , 239. , 243. , 254. ,
        253. ,   nan,   nan],
       [  nan, 232. , 234. , 224. , 266. , 263. , 258. , 218. , 215. ,
        200. , 192. ,   nan],
       [  nan, 211. , 201. , 194. , 190. , 199. , 196. , 187. , 200. ,
        199. , 192. ,   nan],
       [  nan, 200. , 202. , 207. , 201. , 195. , 203. , 196. , 193. ,
        172. , 181. ,   nan],
       [  nan, 212. ,  91.3, 179. , 188. , 208. , 180. , 201. , 239. ,
        243. , 254. ,   nan],
       [  nan, 253. , 232. , 234. , 224. , 266. , 263. , 258. , 218. ,
        215. , 200. ,   nan],
       [  nan, 192. , 211. , 201. , 194. , 190. , 199. , 196. , 187. ,
        200. , 199. ,   nan],
       [  nan,   nan, 192. , 200. , 202. , 207. , 201. , 195. , 203. ,
        196. ,   nan,   nan],
       [  nan,   nan,   nan, 193. , 172. , 181. , 212. ,  91.3, 179. ,
          nan,   nan,   nan],
       [  nan,   nan,   nan,   nan, 188. ,   nan,   nan, 208. ,   nan,
          nan,   nan,   nan]])

This code is used to draw the graph:

bounds = np.linspace(param_ng, param_vg,continious_color_map.N)\  
.reshape(continious_color_map.N,)
norm = mpl.colors.BoundaryNorm(bounds, continious_color_map.N)

fig, ax = plt.subplots(1, 1, figsize = (7, 7))
img = ax.matshow(data, cmap=continious_color_map, norm=norm)

divider = make_axes_locatable(ax)
cax = divider.append_axes("right", size='5%', pad='1%')
fig.colorbar(img, cax=cax, ax=ax, ticks=bounds)

I would be grateful for any help

Solution

You can set the color for nan values via .set_bad() to the colormap.

To create an image from ungridded x,y,z positions, you can use contourf():

import matplotlib.pyplot as plt
import matplotlib as mpl
import numpy as np
from numpy import nan

data = np.array([[nan, nan, nan, nan, 201., 195., 203., 196., nan, nan, nan, nan],
                 [nan, nan, nan, 193., 172., 181., 212., 91.3, 179., nan, nan, nan],
                 [nan, nan, 188., 208., 180., 201., 239., 243., 254., 253., nan, nan],
                 [nan, 232., 234., 224., 266., 263., 258., 218., 215., 200., 192., nan],
                 [nan, 211., 201., 194., 190., 199., 196., 187., 200., 199., 192., nan],
                 [nan, 200., 202., 207., 201., 195., 203., 196., 193., 172., 181., nan],
                 [nan, 212., 91.3, 179., 188., 208., 180., 201., 239., 243., 254., nan],
                 [nan, 253., 232., 234., 224., 266., 263., 258., 218., 215., 200., nan],
                 [nan, 192., 211., 201., 194., 190., 199., 196., 187., 200., 199., nan],
                 [nan, nan, 192., 200., 202., 207., 201., 195., 203., 196., nan, nan],
                 [nan, nan, nan, 193., 172., 181., 212., 91.3, 179., nan, nan, nan],
                 [nan, nan, nan, nan, 188., nan, nan, 208., nan, nan, nan, nan]])

continious_color_map = plt.get_cmap('Spectral', 20)
continious_color_map.set_bad('purple')

bounds = np.linspace(np.nanmin(data), np.nanmax(data), continious_color_map.N)
norm = mpl.colors.BoundaryNorm(bounds, continious_color_map.N)

fig, (ax1, ax2) = plt.subplots(ncols=2, figsize=(14, 6))

img1 = ax1.matshow(data, cmap=continious_color_map, norm=norm)
plt.colorbar(img1, ax=ax1, shrink=0.9)

xs = [i for row in data for i, x in enumerate(row) if not np.isnan(x)]
ys = [j for j, row in enumerate(data) for y in row if not np.isnan(y)]
zs = [d for row in data for d in row if not np.isnan(d)]
img2 = ax2.tricontourf(xs, ys, zs, levels=bounds, cmap=continious_color_map, norm=norm)
ax2.set_aspect('equal')
ax2.set_facecolor('magenta')
ax2.invert_yaxis()
plt.colorbar(img2, ax=ax2, shrink=0.9)

plt.tight_layout()
plt.show()

Answered By - JohanC