Memory used increases with tkinter notebook and matplotlib - python-3.x

Question

I've got a small program that creates a GUI using tkinter. It contains a button that loads a .csv file, creates a notebook with as many tabs as columns in the csv file. Then, on every active tab (at least this is my intention) I have a plot created from a Figure.
The program works as expected , the only problem is that when switching Tabs, the memory used increases with each Tab click.
Memory usage was monitored using the Windows Task Manager.
After loading a csv file, I didn't see the used memory dropping when I chose not loading a new file.
If I don't call the plotting function, when creating only the Tabs, there is no memory issue.
I already tried to manually invoke the garbage collector with gc.collect(), but that didn't help. This is the code I have:
import matplotlib
matplotlib.use('TkAgg')
import pandas as pd
from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg, NavigationToolbar2Tk
from matplotlib.figure import Figure
import matplotlib.pyplot as plt
import os
import sys
import tkinter as tk
from tkinter import messagebox as msg
from tkinter import ttk, filedialog
##import gc
class Graphs(tk.Tk):
def __init__(self, *args, **kwargs):
tk.Tk.__init__(self, *args, **kwargs)
container = tk.Frame(self)
container.pack(side='top', fill='both', expand=1)
container.grid_rowconfigure(0, weight=1)
container.grid_columnconfigure(0, weight=1)
self.protocol('WM_DELETE_WINDOW', self._destroyWindow)
self.frames = {}
frame = StartPage(parent=container, controller=self)
self.frames[StartPage] = frame
frame.grid(row=0, column=0, sticky='nsew')
self.show_frame(StartPage)
def show_frame(self, cont):
frame = self.frames[cont]
frame.tkraise()
def _destroyWindow(self):
self.quit() # stops mainloop
self.destroy()
class StartPage(tk.Frame):
def __init__(self, parent, controller):
tk.Frame.__init__(self, parent)
# initialize lists
self.tabs_list = []
self.hdrs = []
self.figs_list = []
self.ax_list = []
self.canvas_list = []
self.toolbars_list = []
# initialize Data Frame
self.df = pd.DataFrame()
self.nb = None
self.canvas = None
self.toolbar = None
# create LOAD button
self.btn = tk.Button(self, text = 'Load file', command=self.load_csv)
self.btn.pack()
def load_csv(self):
'''
Reset Data Frame;
Destroy notebook if exists;
Load CSV file.
'''
# reset Data Frame
self.df = pd.DataFrame()
# destroy notebook if exists
if self.nb:
self.nb.pack_forget()
self.nb.destroy()
self.nb = None
## gc.collect()
# Select CSV file
self.file_path = filedialog.askopenfilename()
if not self.file_path:
msg.showinfo('Select CSV file', "No file chosen.")
return
try:
# read csv file (exemple.csv)
self.df = pd.read_csv(self.file_path, header=0)
except:
msg.showinfo('Select CSV file', 'Not a csv file / corrupt file.')
return
print(self.df.head())
print(self.df.shape)
# get dimensions
self.m, self.n = self.df.shape
# build the abscissa x from first column
self.x = self.df.iloc[:,0]
# create the notebook
self.nb = ttk.Notebook(self)
# allow Tab navigation
self.nb.enable_traversal()
# add Tabs
for k in range(1, self.n):
hdr = self.df.columns[k]
self.hdrs.append(hdr)
tab = tk.Frame(self.nb, name=hdr.lower())
self.nb.add(tab, text=hdr)
self.tabs_list.append(tab)
self.nb.pack(fill='both', expand=1)
# virtual event after a new tab is selected
self.nb.bind("<<NotebookTabChanged>>", self.plotTH)
def plotTH(self, event):
'''
Plot each Column from Data Frame on its own Tab/Figure
'''
# get path of the selected Tab
tab_path = event.widget.nametowidget(event.widget.select())
# add selected Tab to the list of Tabs
self.tabs_list.append(tab_path)
# get the Tab index;
# When there are no tabs, .select() returns an empty string,
# but .index('current') throws an exception;
# nb.select() returns the Tab NAME (string) of the current selection
if self.nb.select():
i = self.nb.index('current')
# get the Tab text
tab_text = self.nb.tab(i)['text']
else:
return
# remove previous figures ... not sure...
# the used memory as seen in Task Manager still increases
if self.canvas_list:
for cnv in self.canvas_list:
cnv.figure.get_axes().clear()
cnv.get_tk_widget().pack_forget()
cnv.get_tk_widget().destroy()
cnv._tkcanvas.pack_forget()
cnv._tkcanvas.destroy()
cnv = None
if self.figs_list:
for fig in self.figs_list:
fig.delaxes(fig.gca())
plt.cla()
fig.clf()
fig.clear()
plt.close(fig)
self.figs_list = []
# remove toolbar
for widget in tab_path.winfo_children():
widget.pack_forget()
widget.destroy()
self.nb.update() #!!!!!!!!!!!!
######## gc.collect()
# prepare plotting
fig = Figure(figsize=(7, 5), dpi=100)
ax = fig.add_subplot(111)
ax.plot(self.x, self.df.iloc[:,i+1], 'b-', linewidth=1, label=tab_text)
ax.set_xlabel('index')
ax.set_title(self.hdrs[i], fontsize = 8)
ax.legend(loc='best')
ax.grid()
# add to list of figures
self.figs_list.append(fig)
# add to list of axes
self.ax_list.append(ax)
canvas = FigureCanvasTkAgg(fig, master=tab_path)
# add to list of canvases
self.canvas_list.append(canvas)
## self.canvas.draw()
canvas.draw_idle()
canvas.get_tk_widget().pack(side=tk.TOP, fill=tk.BOTH, expand=1)
toolbar = NavigationToolbar2Tk(canvas, tab_path)
# add to list of toolbars
self.toolbars_list.append(toolbar)
toolbar.update()
canvas._tkcanvas.pack(side=tk.TOP, fill=tk.BOTH, expand=1)
def clearPlot(self):
""" not used"""
pass
app = Graphs()
app.title('CSV Plots')
app.geometry('800x600+400+150')
app.resizable(True, True)
app.mainloop();
I created a working csv file using this code:
import numpy as np
import pandas as pd
np.random.seed(0)
df = pd.DataFrame(np.random.randn(50,4), columns=['ALPHA', 'BETA', 'GAMMA', 'DELTA'])
df.index.names = ['Rec']
df.index = df.index + 1
df.to_csv('example.csv', index=True)
print(df)
I'm sorry for the long post. I really don't know were to go from here, so any help would be greatly appreciated.

Answer 1

None of the following lists are ever purged; they hold references to large objects that cannot be garbage collected.
self.tabs_list = []
self.hdrs = []
self.figs_list = []
self.ax_list = []
self.canvas_list = []
self.toolbars_list = []
You probably should create an object that holds the data for each tab, and destroy/purge/reassign to None that when changing tabs.

Answer 2

I changed the code; I realized I was creating a new figure (with all the Artists) and a new canvas (FigureCanvasTkAgg) with every TabChanged event. Instead, I create a figure and its canvas at the moment of Tab addition to the notebook; I do the plot only when the TabChanged occurs and I take care to close all the plots and to destroy the previous notebook when performing a new CSV file load. Destroying the notebook removes the canvases too. Here is the code:
import matplotlib
import matplotlib.style as mplstyle
matplotlib.use('TkAgg')
import pandas as pd
from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg, NavigationToolbar2Tk
from matplotlib.figure import Figure
import matplotlib.pyplot as plt
import os, sys, time
import tkinter as tk
from tkinter import messagebox as msg
from tkinter import ttk, filedialog
mplstyle.use('fast')
class Graphs(tk.Tk):
def __init__(self, *args, **kwargs):
tk.Tk.__init__(self, *args, **kwargs)
container = tk.Frame(self)
container.pack(side='top', fill='both', expand=1)
container.grid_rowconfigure(0, weight=1)
container.grid_columnconfigure(0, weight=1)
self.protocol('WM_DELETE_WINDOW', self._destroyWindow)
self.frames = {}
frame = StartPage(parent=container, controller=self)
self.frames[StartPage] = frame
frame.grid(row=0, column=0, sticky='nsew')
self.show_frame(StartPage)
def show_frame(self, cont):
frame = self.frames[cont]
frame.tkraise()
def _destroyWindow(self):
self.quit() # stops mainloop
self.destroy()
class StartPage(tk.Frame):
def __init__(self, parent, controller):
tk.Frame.__init__(self, parent)
# initialize Data Frame
self.df = pd.DataFrame()
self.nb = None
self.fig = None
self.canvas = None
self.toolbar = None
# create LOAD button
self.btn = tk.Button(self, text = 'Load file', command=self.load_csv)
self.btn.pack()
def load_csv(self):
''' Close the plots;
Reset Data Frame;
Destroy notebook if exists;
Load CSV file.
'''
# Setting interactive mode off
if plt.isinteractive():
plt.ioff()
plt.close("all")
# reset Data Frame
self.df = pd.DataFrame()
# initialize list
self.hdrs = []
try:
# destroy notebook if exists
self.nb.pack_forget()
self.nb.destroy()
except:
pass
self.nb = None
# Select CSV file
self.file_path = filedialog.askopenfilename()
if not self.file_path:
msg.showinfo('Select CSV file', "No file chosen.")
return
try:
# read csv file (exemple.csv)
self.df = pd.read_csv(self.file_path, header=0)
except:
msg.showinfo('Select CSV file', 'Not a csv file / corrupt file.')
return
# get dimensions
self.m, self.n = self.df.shape
# build the abscissa x from first column
self.x = self.df.iloc[:,0]
# create the notebook
self.nb = ttk.Notebook(self)
# allow Tab navigation
self.nb.enable_traversal()
# add Tabs
for k in range(1, self.n):
hdr = self.df.columns[k]
self.hdrs.append(hdr)
tab = tk.Frame(self.nb, name=hdr.lower())
self.nb.add(tab, text=hdr)
self.fig = plt.figure(num=hdr.lower(), clear=True, figsize=(7, 5), dpi=100)
# self.ax = self.fig.add_subplot()
self.canvas = FigureCanvasTkAgg(self.fig, master=tab)
self.canvas.get_tk_widget().pack(side=tk.TOP, fill=tk.BOTH, expand=1)
toolbar = NavigationToolbar2Tk(self.canvas, tab)
toolbar.update()
self.canvas._tkcanvas.pack(side=tk.TOP, fill=tk.BOTH, expand=1)
self.nb.pack(fill='both', expand=1)
# virtual event after a new tab is selected
self.nb.bind("<<NotebookTabChanged>>", self.plotTH)
def plotTH(self, event):
'''
Plot each Column from Data Frame on its own Tab/Figure
'''
# Setting interactive mode on is essential: plt.ion()
if not plt.isinteractive():
plt.ion()
# tab index
i = self.nb.index('current')
# tab text
tab_text = self.nb.select().split('.')[-1]
# set current figure
cf = plt.figure(tab_text)
plt.clf()
# plotting
ax = plt.subplot(111)
ax.plot(self.x, self.df.iloc[:,i+1], 'b-', linewidth=1, label=tab_text)
ax.set_xlabel('index')
ax.set_title(self.hdrs[i], fontsize = 8)
ax.legend(loc='best')
ax.grid()
cf.canvas.draw()
app = Graphs()
app.title('CSV Plots')
app.geometry('800x600+400+150')
app.resizable(True, True)
app.mainloop()