default argument in function for creating a dictionary creates an infinete loop of self references

Question

a have a problem with the following with the following code:

from lxml import etree as ET

class test(object):
    def __init__(self, **args):
        self.tag = []
        self.meshfiles = []
        self.root = None
    def addMesh(self, **args):
        if not "filename" in args:
            raise KeyError("No filename given")
        else:
            self.meshfiles.append(args["filename"])
    def populateTree(self, tag, text='', attr={}, children={}):
        return {'tag': tag, 'text': text, 'attr': attr, 'children': children}

    @property
    def tree(self):
        baum = {'meshes': {}}
        if len(self.meshfiles) == 1:
            baum['meshes'] = self.populateTree('mesh', text=self.meshfiles[0])
        else:
            baum['meshes'] = self.populateTree('meshes')
            for i, meshfile in enumerate(self.meshfiles):
                # works:
                baum['meshes']['children'][i] = self.populateTree('mesh', text=meshfile)
                # not working:
                # baum['meshes']['children'][i] = self.populateTree('mesh', text=meshfile, children={})
        return baum

    def createXML(self):
        self.root = ET.Element("rootelement")
        self.dict2xml(self.root, self.tree)
        return ET.tostring(self.root, pretty_print=True)
    def dict2xml(self, parent, dictionary):
        for entry in dictionary:
            self.tag.append(ET.SubElement(parent, dictionary[entry]['tag']))
            self.tag[-1].text = str(dictionary[entry]['text'])
            for attr in dictionary[entry]['attr']:
                self.tag[-1].set(attr, dictionary[entry]['attr'][attr])
            if len(dictionary[entry]['children']) > 0:
                self.dict2xml(self.tag[-1], dictionary[entry]['children'])
if __name__ == "__main__":
    t = test()
    t.addMesh(filename="a")
    t.addMesh(filename="c")
    t.addMesh(filename="b")
    print(t.tree)
    print(t.createXML())

While this example gives a recursion error:

    {'meshes': {'tag': 'meshes', 'text': '', 'attr': {}, 'children': {0: {'tag': 'mesh', 'text': 'a', 'attr': {}, 'children': {...}}, 1: {'tag': 'mesh', 'text': 'c', 'attr': {}, 'children': {...}}, 2: {'tag': 'mesh', 'text': 'b', 'attr': {}, 'children': {...}}}}}
Traceback (most recent call last):
  File "/home/buchwalj/temp/bug-python/test.py", line 48, in <module>
    print(t.createXML())
  File "/home/buchwalj/temp/bug-python/test.py", line 32, in createXML
    self.dict2xml(self.root, self.tree)
  File "/home/buchwalj/temp/bug-python/test.py", line 41, in dict2xml
    self.dict2xml(self.tag[-1], dictionary[entry]['children'])
  File "/home/buchwalj/temp/bug-python/test.py", line 41, in dict2xml
    self.dict2xml(self.tag[-1], dictionary[entry]['children'])
  File "/home/buchwalj/temp/bug-python/test.py", line 41, in dict2xml
    self.dict2xml(self.tag[-1], dictionary[entry]['children'])
  [Previous line repeated 994 more times]
  File "/home/buchwalj/temp/bug-python/test.py", line 36, in dict2xml
    self.tag.append(ET.SubElement(parent, dictionary[entry]['tag']))
 RecursionError: maximum recursion depth exceeded while calling a Python object

Commenting out line 27 instead of 25 prints out the correct dictionary with the corresponding xml. The only difference there is, that the working example uses the same argument as the default argument in the function call instead of the default argument itself. Is this a bug or am I doing something wrong here?

I didn't think through all of this code, but the difference is that in the first line (that as not commented out above), the *same* dictionary (the default argument) gets reused every time, whereas in the second line (that is commented out above), a *new* dictionary is created each time. — mkrieger1, Apr 10 '21 at 22:18
Obviously. However, it is not clear to me that it should work that way. These are different function calls and dictionaries are mutable objects, so this behavior does not make much sense. — foobar, Apr 10 '21 at 22:45
Please try to create a minimal example, by getting rid of functionality not related to the problem. — Karl Knechtel, Apr 10 '21 at 23:12

score 0 · Accepted Answer · answered Apr 10 '21 at 22:52

Using empty dictionary as defaults are actually bad ideas, mainly because dictionaries are mutable objects, meaning if you accidentally changing the content of the dictionary, it is there to stay (cause reference to object stays the same). See: Why is the empty dictionary a dangerous default value in Python?

I would suggest changing your populateTree function to something like:

def populateTree(self, tag, text='', attr=None, children=None):
        if attr is None:
            attr = {}
        if children is None:
            children = {}
        return {'tag': tag, 'text': text, 'attr': attr, 'children': children}

so it creates a new empty dictionary everything populateTree is called without those arguments.

default argument in function for creating a dictionary creates an infinete loop of self references

1 Answers1