Question

I have a custom gtk.treeview wrapper class that manages its own liststore. The class has its own method that clears the liststore’s data and overwrites it with new data. For the purpose of testing memory consumption,the treeview just displays 50,000 rows of integers in 8 columns.

I’ve noticed that every call to my function that updates the liststore with data seems to increase the memory the application uses, and it never goes back down after data is cleared.

Here is my simple treeview implementation:

class TreeViewPrototype(gtk.TreeView):
    def __init__(self):
        gtk.TreeView.__init__(self)

        self.columns = [str, str, str, str, str, str, str, str]
        # Initialize Treeview  and TreeViewColumns here
        # Setup columns
        self.liststore = gtk.ListStore(*self.columns)
        self.set_model(self.liststore)

    def set_list_model(self):
        self.liststore.clear()
        self.liststore = gtk.ListStore(*self.columns)
        # Populate liststore with dummy data
        for i in range(50000):
            row = []
            for j in range(len(self.columns)):
                row.append("%d[%d]"%(i,j))
            self.liststore.append(row)

It seems to me that every call to set_list_model, when I overwrite the liststore with a new one, it never unallocates the memory of the data. What am I doing wrong?

Answer 1

The problem seems to be with the reference to the liststore in the
treeview. Because the treeview is a container that is present
throughout the entire lifecycle of the application, the old liststore
hangs around, being still referenced by the treeview’s constructor. If
a new liststore is created everytime and no reference is held by the
treeview, that problem seems to go away.

My Solution:

class TreeViewPrototype( gtk.TreeView ):
   def __init__(self):
       gtk.TreeView.__init__(self)
       self.columns = [str,str,str,str,str,str,str,str]
       #TreeView Initialization

   def set_list_model( self ):
       self.set_model(None)
       liststore = gtk.ListStore( *self.columns)
       for i in range( 50000 ):
           row = []
           for j in range( len( self.columns)):
               row.append( "%d[%d]"%(i,j))
           liststore.append( row )
       self.set_model( liststore )

Using set_model(None) seems to correctly dereference the current
liststore in the model so it doesn’t hang around in memory.