set_xticks in bar chart for groups of data categories

I have a dataframe with five columns [CAT_LET, fid, input, climate change, terrestrial acidification].

In my real dataset there are:

6 values for CAT_LET = ['veg', 'fruit',...],
3871 for fid= [1,2,...],
12 for input = ['kgN', 'urea',...],
and the corresponding values for two environmental impacts: climate change and terrestrial acidification.

The code makes a stacked bar graph scaled to 100% for each impact (with the contribution of each input for each fid) and should organize the bar per CAT_LET and fid.

Now I need to add the x labels to the graph where I can see each of the 6 CAT_LET values for groups of fid that vary in length. In my example I expect veg to appear in x for fids 1,3 and 5 and fruit for 2 and 4. Any idea how to add these ticks and labels?

In my real dataset I do not want all 3871 fid to get the category name on the x label but only the groups with the same category appear (they should be together).

Thanks for the help.

    import pylab as pb
    from glob import glob
    import pandas as pd
    import numpy as np

All_impacts_cat2 = pd.DataFrame()

All_impacts_cat2 = pd.DataFrame({'CAT_LET':['veg','veg','fruit','fruit','veg','veg','veg','fruit','veg','fruit','veg','fruit','fruit','veg','veg'], 
'fid': [5,1,4,2,3,1,5,2,3,4,1,2,4,3,5], 
'input':['urea','urea','kgN','urea','urea','kgN','kgN','kgN','kgN','urea','manure','manure','manure','manure','manure'], 
'climate change': [1,0,15,10,1,20,1,1,10,15,4,10,15,3,2], 
'terrestrial acidification': [0,1,10,10,2,1,2,10,1,15,4,20,25,3,2]})

impact_list = ['terrestrial acidification','climate change']
inputs = ['urea', 'kgN','manure']

for impact in impact_list[:1]:
values_list = []
for x in inputs:
    values_list.append(All_impacts_cat2.loc[All_impacts_cat2['input']==x].sort_values(by=['CAT_LET','fid'])[impact].values)

labels = [*range(0,len(values_list[0]))]
width = 0.35    

sum_values = np.zeros(np.shape(np.array(values_list[0]))) 
sum_total = np.zeros(np.shape(np.array(values_list[0])))
   
fig, ax = pb.subplots(figsize=(11,10))

for i in range(len(inputs)):
    sum_total = sum_total + np.array(values_list[i])
sum_total[sum_total ==0] = 1

for i in range(len(inputs)):
    data = values_list[i]/sum_total
    ax.bar(labels, data, width, label=inputs[i], bottom =sum_values)
    sum_values = sum_values + np.array(data)

ax.set_ylabel(impact)
ax.legend()
pb.show()

Solution

As a quick fix, you could collect the unique values for each fid and use them to label the x-positions of the bars.

from matplotlib import pyplot as plt
import pandas as pd
import numpy as np

All_impacts_cat2 = pd.DataFrame({'CAT_LET':['veg','veg','fruit','fruit','veg','veg','veg','fruit','veg','fruit','veg','fruit','fruit','veg','veg'], 
'fid': [5,1,4,2,3,1,5,2,3,4,1,2,4,3,5], 
'input':['urea','urea','kgN','urea','urea','kgN','kgN','kgN','kgN','urea','manure','manure','manure','manure','manure'], 
'climate change': [1,0,15,10,1,20,1,1,10,15,4,10,15,3,2], 
'terrestrial acidification': [0,1,10,10,2,1,2,10,1,15,4,20,25,3,2]})

impact_list = ['terrestrial acidification','climate change']
inputs = ['urea', 'kgN','manure']

for impact in impact_list[:1]:
    values_list = []
    for x in inputs:
        values_list.append(All_impacts_cat2.loc[All_impacts_cat2['input']==x].sort_values(by=['CAT_LET','fid'])[impact].values)

labels = [*range(0,len(values_list[0]))]
width = 0.35    

sum_values = np.zeros(np.shape(np.array(values_list[0]))) 
sum_total = np.zeros(np.shape(np.array(values_list[0])))
   
fig, ax = plt.subplots(figsize=(11,10))

for i in range(len(inputs)):
    sum_total = sum_total + np.array(values_list[i])
sum_total[sum_total ==0] = 1

for i in range(len(inputs)):
    data = values_list[i]/sum_total
    ax.bar(labels, data, width, label=inputs[i], bottom =sum_values)
    sum_values = sum_values + np.array(data)
    
#get all unique CAT_LET entries for each fid
cat_groups = All_impacts_cat2["CAT_LET"].groupby(All_impacts_cat2["fid"]).unique().apply(list)
#relabel the x-ticks
ax.set_xticks(range(cat_groups.size), cat_groups)

ax.set_ylabel(impact)
ax.legend()

plt.show()

Sample output:

This approach takes into account that (although implied by your sample dataframe) not one fid value is mapped to exactly one CAT_LET value (for instance, change the first veg to fruit). If each fid has exactly one CAT_LET entry assigned, then you could simplify your approach. Please also note that I have removed pylab from your code as its use is now discouraged by matplotlib.