creating inset in matplot lib

I have created a plot in matplot lib and I wish to add an inset to that plot. The data I wish to plot is kept inside of a dictionary that I use in other figures. I'm finding this data inside a loop which I then run this loop again for the subplot. Here is the relevant segment:

leg = []
colors=['red','blue']
count = 0
for key in Xpr: #Xpr holds my data
#skipping over what I don't want to plot
if not key[0] == '5': continue
if key[1] == '0': continue
if key[1] == 'a': continue
leg.append(key)
x = Xpr[key]
y = Ypr[key] #Ypr holds the Y axis and is created when Xpr is created
plt.scatter(x,y,color=colors[count],marker='.')
count += 1

plt.xlabel(r'$z/\mu$')
plt.ylabel(r'$\rho(z)$')
plt.legend(leg)
plt.xlim(0,10)
#Now I wish to create the inset
a=plt.axes([0.7,0.7,0.8,0.8])
count = 0
for key in Xpr:
break
if not key[0] == '5': continue
if key[1] == '0': continue
if key[1] == 'a': continue
leg.append(key)
x = Xpr[key]
y = Ypr[key]
a.plot(x,y,color=colors[count])
count += 1
plt.savefig('ion density 5per Un.pdf',format='pdf')

plt.cla()


The strange thing is that when I tried to move the inset position, I still get the previous insets (those from the previous run of the code). I even tried to comment out the a=axes([]) line without any apparent. I attach en example file. Why is it acting in that way?

The simple answer is you should use plt.clf() which clears the figure, not the current axes. There is also a break in the inset loop which means none of that code will ever run.

When you start to do more complicated things than use a single axes, it is worth switching over to using the OO interface to matplotlib. It may seem more complicated at first, but you no longer have to worry about the hidden state of pyplot. Your code can be re-written as

fig = plt.figure()
ax = fig.add_axes([.1,.1,.8,.8]) # main axes
colors=['red','blue']
for key  in Xpr: #Xpr holds my data
#skipping over what I don't want to plot
if not key[0] == '5': continue
if key[1] == '0': continue
if key[1] == 'a': continue
x = Xpr[key]
y = Ypr[key] #Ypr holds the Y axis and is created when Xpr is created
ax.scatter(x,y,color=colors[count],marker='.',label=key)
count += 1

ax.set_xlabel(r'$z/\mu$')
ax.set_ylabel(r'$\rho(z)$')
ax.set_xlim(0,10)
leg = ax.legend()

#Now I wish to create the inset
count =0
for key  in Xpr: #Xpr holds my data
if not key[0] == '5': continue
if key[1] == '0': continue
if key[1] == 'a': continue
x = Xpr[key]
y = Ypr[key]
ax_inset.plot(x,y,color=colors[count],label=key)
count +=1

ax_inset.legend()