Overriding dict.update() method in subclass to prevent overwriting dict keys

Question

Earlier today, I read the question "Raise error if python dict comprehension overwrites a key" and decided to try my hand at an answer. The method that naturally occurred to me was to subclass dict for this. However, I got stuck on my answer, and now I'm obsessed with getting this worked out for myself.

Notes:

• No - I do not plan on turning in the answer to this question as an answer to the other question.
• This is purely an intellectual exercise for me at this point. As a practical matter, I would almost certainly use a namedtuple or a regular dictionary wherever I have a requirement for something like this.

My (not quite working) Solution:

class DuplicateKeyError(KeyError):
    pass



class UniqueKeyDict(dict):
    def __init__(self, *args, **kwargs):
        self.update(*args, **kwargs)


    def __setitem__(self, key, value):
        if key in self:  # Validate key doesn't already exist.
            raise DuplicateKeyError('Key \'{}\' already exists with value \'{}\'.'.format(key, self[key]))
        super().__setitem__(key, value)


    def update(self, *args, **kwargs):
        if args:
            if len(args) > 1:
                raise TypeError('Update expected at most 1 arg.  Got {}.'.format(len(args)))
            else:
                try:
                    for k, v in args[0]:
                        self.__setitem__(k, v)
                except ValueError:
                    pass

        for k in kwargs:
            self.__setitem__(k, kwargs[k])

My Tests and Expected Results

>>> ukd = UniqueKeyDict((k, int(v)) for k, v in ('a1', 'b2', 'c3', 'd4'))  # Should succeed.
>>> ukd['e'] = 5  # Should succeed.
>>> print(ukd)
{'a': 1, 'b': 2, 'c': 3, d: 4, 'e': 5}
>>> ukd['a'] = 5  # Should fail.
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "<stdin>", line 8, in __setitem__
__main__.DuplicateKeyError: Key 'a' already exists with value '1'.
>>> ukd.update({'a': 5})  # Should fail.
>>> ukd = UniqueKeyDict((k, v) for k, v in ('a1', 'b2', 'c3', 'd4', 'a5'))  # Should fail.
>>>

I'm certain the issue is in my update() method, but I'm not able to determine just what I'm doing wrong.

Below is the original version of my update() method. This version fails as expected on duplicates when calling my_dict.update({k: v}) for a key/value pair already in the dict, but does not fail when including a duplicate key while creating the original dict, due to the fact that converting the args to a dict results in default behavior for a dictionary, i.e., overwriting the duplicate key.

def update(self, *args, **kwargs):
    for k, v in dict(*args, **kwargs).items():
        self.__setitem__(k, v)

Answer 1

It's interesting that simply overriding __setitem__ is not enough to change the behavior of update in dict. I would have expected that dict would use its __setitem__ method when it's being updated using update. In all cases, I think it's better to implement collections.MutableMapping to achieve the desired result without touching update:

import collections

class UniqueKeyDict(collections.MutableMapping, dict):

    def __init__(self, *args, **kwargs):
        self._dict = dict(*args, **kwargs)

    def __getitem__(self, key):
        return self._dict[key]

    def __setitem__(self, key, value):
        if key in self:
            raise DuplicateKeyError("Key '{}' already exists with value '{}'.".format(key, self[key]))
        self._dict[key] = value

    def __delitem__(self, key):
        del self._dict[key]

    def __iter__(self):
        return iter(self._dict)

    def __len__(self):
        return len(self._dict)

Edit: included dict as base class to satisfy the isinstance(x, dict) check.

Answer 2

Note that, per the documentation:

• dict.update takes a single other parameter, "either another dictionary object or an iterable of key/value pairs" (I've used collections.Mapping to test for this) and "If keyword arguments are specified, the dictionary is then updated with those key/value pairs"; and
• dict() takes a single Mapping or Iterable along with optional **kwargs (the same as update accepts...).

This is not quite the interface you have implemented, which is leading to some issues. I would have implemented this as follows:

from collections import Mapping


class DuplicateKeyError(KeyError):
    pass


class UniqueKeyDict(dict):

    def __init__(self, other=None, **kwargs):
        super().__init__()
        self.update(other, **kwargs)

    def __setitem__(self, key, value):
        if key in self:
            msg = 'key {!r} already exists with value {!r}'
            raise DuplicateKeyError(msg.format(key, self[key]))
        super().__setitem__(key, value)

    def update(self, other=None, **kwargs):
        if other is not None:
            for k, v in other.items() if isinstance(other, Mapping) else other:
                self[k] = v
        for k, v in kwargs.items():
            self[k] = v

In use:

>>> UniqueKeyDict((k, v) for k, v in ('a1', 'b2', 'c3', 'd4'))
{'c': '3', 'd': '4', 'a': '1', 'b': '2'}
>>> UniqueKeyDict((k, v) for k, v in ('a1', 'b2', 'c3', 'a4'))
Traceback (most recent call last):
  File "<pyshell#8>", line 1, in <module>
    UniqueKeyDict((k, v) for k, v in ('a1', 'b2', 'c3', 'a4'))
  File "<pyshell#7>", line 5, in __init__
    self.update(other, **kwargs)
  File "<pyshell#7>", line 15, in update
    self[k] = v
  File "<pyshell#7>", line 10, in __setitem__
    raise DuplicateKeyError(msg.format(key, self[key]))
DuplicateKeyError: "key 'a' already exists with value '1'"

and:

>>> ukd = UniqueKeyDict((k, v) for k, v in ('a1', 'b2', 'c3', 'd4'))
>>> ukd.update((k, v) for k, v in ('e5', 'f6'))  # single Iterable
>>> ukd.update({'h': 8}, g='7')  # single Mapping plus keyword args
>>> ukd
{'e': '5', 'f': '6', 'a': '1', 'd': '4', 'c': '3', 'h': 8, 'b': '2', 'g': '7'}

If you ever end up using this, I'd be inclined to give it a different __repr__ to avoid confusion!

Answer 3

I am not sure this is the problem but I just noticed that you are treating your args in the update method as a list of pairs:

for k, v in args[0]

while you are actually supplying a dictionary:

ukd.update({'a': 5})

Have you tried this:

try:
    for k, v in args[0].iteritems():
        self.__setitem__(k, v)
except ValueError:
    pass

EDIT: Probably this error went unnoticed because you are excepting a ValueError, which is what treating a dictionary as a list of pairs will raise.

Answer 4

I was able to achieve the goal with the following code:

class UniqueKeyDict(dict):
    def __init__(self, *args, **kwargs):
        self.update(*args, **kwargs)

    def __setitem__(self, key, value):
        if self.has_key(key):
            raise DuplicateKeyError("%s is already in dict" % key)
        dict.__setitem__(self, key, value)

    def update(self, *args, **kwargs):
        for d in list(args) + [kwargs]:
            for k,v in d.iteritems():
                self[k]=v

Answer 5

This interesting question is a bit older and has already some solid answers (my favourite is the one from sirfz). Nevertheless, I would like to propose yet another one. You could use the dict-wrapper UserDict. If I'm not mistaken, this should do the job you were looking for:

from collections import UserDict

class DuplicateKeyError(KeyError):
    pass

class UniqueKeyDict(UserDict):

    def __setitem__(self, key, value):
        if key in self:
            raise DuplicateKeyError(f"Key '{key}' already exists with value '{self[key]}'")
        self.data[key] = value

As with the usage of collections.abc.MutableMapping the update method gets modified implicitly. But in contrast, you only have to (re)define the __setitem__ method. Since your modification is rather minor, the use of UserDict seems like an appropriate approach to me.

An instance of this class is not an instance of dict, but it is an instance of collections.abc.Mapping, which should be used for testing for dict-likeness.

Answer 6

Why not do something along the lines inspired by MultiKeyDict using setdefault? This leaves the update method as a way to override the currently stored values, breaking, I know, the intent that d[k] = v == d.update({k, v}). In my application the override was useful. So before flagging this as not answering the OP question, please consider this answer might be useful for someone else.

class DuplicateKeyError(KeyError):
    """File exception rasised by UniqueKeyDict"""
    def __init__(self, key, value):
        msg = 'key {!r} already exists with value {!r}'.format(key, value)
        super(DuplicateKeyError, self).__init__(msg)


class UniqueKeyDict(dict):
    """Subclass of dict that raises a DuplicateKeyError exception"""
    def __setitem__(self, key, value):
        if key in self:
            raise DuplicateKeyError(key, self[key])
        self.setdefault(key, value)


class MultiKeyDict(dict):
    """Subclass of dict that supports multiple values per key"""
    def __setitem__(self, key, value):
        self.setdefault(key, []).append(value)

Rather new to python so flame on, probably deserve it...

Answer 7

According to the given answers and help(dict.update) I did force_update method. I will appreciate it if you will post in comments a real python code of dict.update method (I found only C source code).

class UniqueKeyDict(UserDict):
    """
    Disable key overwriting if already exists
    """

    def __setitem__(self, key=None, value=None, **kwargs):
        if (key in self  # Different options to override key (just a fun)
                and not (isinstance(value, Iterable) and len(value) == 3 and self[key] in value and value[-1] is True)
                and not (isinstance(value, Iterable) and len(value) == 2 and value[-1] == '--force')):
            raise DuplicateKeyError(f"Key '{key}' already exists with value '{self[key]}'")
        self.data[key] = value

    def force_update(self, *a, **kw) -> None:
        """
        See help({}.update)
        """
        a = a[0] if len(a) == 1 else None  # *a is always tuple
        if a and hasattr(a, 'keys'):
            for k in a:
                self.pop(k, None)
                self[k] = a[k]
        elif a:
            for k, v in a:
                self.pop(k, None)
                self[k] = v
        for k in kw:
            self.pop(k, None)
            self[k] = kw[k]


# Check it, it should cover all the cases with regular dict.update method
du = UniqueKeyDict()
du['e'] = 3
du.force_update({'q': 1, 'qq': 2, 'qqq': 3})
du.update({'q': 1, 'qq': 2, 'qqq': 3})  # Error
du.force_update({'q': 1, 'qq': 2, 'qqq': 3})  # No error
du.force_update({})
du.force_update([])
du.force_update(w=2, ww=22, www=222)
du.force_update([[1, 2], [3, 4], [5, 6], [7, 8]])