How to make dictionary read-only?

Question

I have a class like:

class A:
    def __init__(self):
        self.data = {}

and at some moment I want to prohibit self.data fields modification.

I've read in PEP-416 rejection notice that there are a lot of ways to do it. So I'd like to find what they are.

I tried this:

a = A()
a.data = types.MappingProxyType(a.data)

That should work but first, its python3.3+ and second, when I do this "prohibition" multiple times I get this:

>>> a.data = types.MappingProxyType(a.data)
>>> a.data = types.MappingProxyType(a.data)
>>> a.data
mappingproxy(mappingproxy({}))

though it would be much better to get just mappingproxy({}) as I am going to "prohibit" a lot of times. Check of isinstance(MappingProxyType) is an option, but I think that other options can exist.

Thanks

A frozendict implementation looks rather trivial: github.com/slezica/python-frozendict/blob/master/frozendict/… — georg
– georg, Commented Sep 26, 2013 at 8:36
@codelover Thats right, I can override __setitem__ with throw NotImplemented. Is there guarantee that there will be no way to modify keys or values via any standard dict attributes in any python implementation? — sshilovsky
– sshilovsky, Commented Sep 27, 2013 at 6:28
@sshilovsky Nope, instead you could derive an user dict and override whatever & you can use where ever it is applicable. — Garfield
– Garfield, Commented Sep 27, 2013 at 12:52
Does this answer your question? What would a "frozen dict" be? — Seanny123
– Seanny123, Commented Feb 3, 2021 at 23:18

NoDataDumpNoContribution · Accepted Answer · 2020-12-04 20:52:45Z

47

Use collections.Mapping e.g.

import collections

class DictWrapper(collections.Mapping):

    def __init__(self, data):
        self._data = data

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

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

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

edited Dec 4, 2020 at 20:52

NoDataDumpNoContribution

10.9k9 gold badges70 silver badges113 bronze badges

answered Feb 11, 2015 at 10:58

mouad

70.5k18 gold badges117 silver badges106 bronze badges

Sign up to request clarification or add additional context in comments.

3 Comments

Martin Thoma Over a year ago

Can it be that this only makes the dict itself, but not necessarilty contained dicts immutable? I currently want to make a nested dict representing a config immuatable.

sshilovsky Over a year ago

I think it works exactly like you describe already. dict_wrapper[key] returns whatever you've put there, which might as well be a raw dict.

Jonathan Mayer Over a year ago

Great answer! It's worth mentioning: (1) Modern python wants collections.abc.Mapping, not collections.Mapping which has been deprecated. (2) collections.abc.Mapping simply doesn't have __setitem__ or ilk methods defined, so attempts to store data to the mapping will result in a TypeError exception. (3) If you want something to make an entire tree of dictionaries read-only, you need to make __gititem__ return a read-only wrapper for lists and dicts. Additionally, you'll need to define a read-only iterator for __iter__ to return.

Marcin Raczyński · Accepted Answer · 2022-07-08 10:57:11Z

41

This is the full implementation of fast (shallow-)read-only dict:

def _readonly(self, *args, **kwargs):
    raise RuntimeError("Cannot modify ReadOnlyDict")

class ReadOnlyDict(dict):
    __setitem__ = _readonly
    __delitem__ = _readonly
    pop = _readonly
    popitem = _readonly
    clear = _readonly
    update = _readonly
    setdefault = _readonly

My previous (worse) implementation was as follows (thanks @mtraceur for the great remarks!):

class ReadOnlyDict(dict):
    def __readonly__(self, *args, **kwargs):
        raise RuntimeError("Cannot modify ReadOnlyDict")
    __setitem__ = __readonly__
    __delitem__ = __readonly__
    pop = __readonly__
    popitem = __readonly__
    clear = __readonly__
    update = __readonly__
    setdefault = __readonly__
    del __readonly__

edited Jul 8, 2022 at 10:57

answered Jun 25, 2015 at 12:05

Marcin Raczyński

9851 gold badge10 silver badges14 bronze badges

6 Comments

blueyed Over a year ago

To allow for copying them using copy.copy and copy.deepcopy you might want to add the following two lines: __copy__ = dict.copy and __deepcopy__ = copy._deepcopy_dispatch.get(dict). Not much tested though.

mari.mts Over a year ago

right, but how to initialize this or how to make existing dict use this?

Marcin Raczyński Over a year ago

ro_dict = ReadOnlyDict(rw_dict)

mtraceur Over a year ago

Doing del __readonly__ will mess with pickling of the methods, both unbound and bound, if that even comes up. Also all names of the form __*__ are reserved by Python, and are free to break at any time. So I would call it _readonly, define it outside the class, and just leave it in the module it's defined in instead of del-ing it. (Bound which are that function will still be unpickleable, but that's due to a weirdly broken approach in Python's pickling of bound methods - unbound references to those methods will be pickleable.)

mtraceur Over a year ago

Cheers, just noticed the edit, happy to help!

|

JadedTuna · Accepted Answer · 2013-09-26 13:24:23Z

11

Very easy, you just override default dict's methods!
Here is an example:

class ReadOnlyDict(dict):

    __readonly = False

    def readonly(self, allow=1):
        """Allow or deny modifying dictionary"""
        self.__readonly = bool(allow)

    def __setitem__(self, key, value):

        if self.__readonly:
            raise TypeError, "__setitem__ is not supported"
        return dict.__setitem__(self, key, value)

    def __delitem__(self, key):

        if self.__readonly:
            raise TypeError, "__delitem__ is not supported"
        return dict.__delitem__(self, key)

BTW, you can also remove .pop, .update and other methods you need. Just play around with it.

edited Sep 26, 2013 at 13:24

answered Sep 26, 2013 at 8:40

JadedTuna

1,8633 gold badges20 silver badges33 bronze badges

3 Comments

justhalf Over a year ago

How do you apply this "at some moment"? I guess OP needs the dict to be read-write, then make it frozen after some time.

JadedTuna Over a year ago

If he needs, he can modify them too

sshilovsky Over a year ago

This approach is obvious but is not simple as I have to override almost each dict method and even reimplement dict views like .keys(), .items() etc. And what if Guido adds some more in future python versions? Anyway, thanks for the point, but I wish there were more answers.

Ouroborus · Accepted Answer · 2021-08-06 01:56:06Z

4

The best way is to derive from UserDict like this:

from collections import UserDict

class MyReadOnlyDict(UserDict):
   def my_set(self, key, val, more_params):
       # do something special
       # custom logic etc
       self.data[key] = val

   def __setitem__(self, key, val):
       raise NotImplementedError('This dictionary cannot be updated')

   def __delitem__(self, key):
       raise NotImplementedError('This dictionary does not allow delete')

The advantage of this method is that you can still have internal methods in your class that can update dictionary by accessing self.data.

edited Aug 6, 2021 at 1:56

Ouroborus

17k8 gold badges42 silver badges65 bronze badges

answered May 9, 2020 at 17:13

Shital Shah

69.9k21 gold badges258 silver badges202 bronze badges

1 Comment

Vladimir Prudnikov Over a year ago

What about instance = MyReadOnlyDict(); instance.data['a'] = 'b'

Polv · Accepted Answer · 2018-07-11 16:25:43Z

How about? It is the update of @mouad 's answer.

import json
from collections import OrderedDict
from collections.abc import Mapping


class ReadOnlyJsonObject(Mapping):
    def __init__(self, data, dumps_kw: dict=None, loads_kw: dict=None):
        if dumps_kw is None:
            dumps_kw = dict()
        if loads_kw is None:
            self._loads_kw = dict(object_pairs_hook=OrderedDict)
        else:
            self._loads_kw = loads_kw

        if isinstance(data, str):
            self._json_string = data
        else:
            self._json_string = json.dumps(data, **dumps_kw)

    @property
    def _data(self):
        return json.loads(self._json_string, **self._loads_kw)

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

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

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

    def __str__(self):
        return self._json_string

Not sure about the performance, though. I use this one in a real project https://github.com/patarapolw/AnkiTools/blob/master/AnkiTools/tools/defaults.py

Hai Vu · Accepted Answer · 2023-07-27 18:58:10Z

Belated answer. You can use a combination of property and MappingProxyType to protect the .data field from modifications.

from types import MappingProxyType

class A:
    def __init__(self):
        self._data = {}

    @property
    def data(self):
        return MappingProxyType(self._data)

    def update(self, key, value):
        self._data[key] = value

a = A()
a.update("foo", 1)
a.update("bar", 2)
print(f"a.data={a.data}")

# Attempt to modify a.data directly will fail
try:
    a.data["bar"] = 20
except TypeError as error:
    print(f"Error: {error}")

# Attempt to assign something else to a.data, will also fail
try:
    a.data = {}
except AttributeError as error:
    print(f"Error: {error}")

Output:

a.data={'foo': 1, 'bar': 2}
Error: 'mappingproxy' object does not support item assignment
Error: property 'data' of 'A' object has no setter

Notes

The A.data is not a regular attribute, it is a property. Since there is no setter, it is also a read-only property.
A.data returns a MappingProxy instead of the dictionary itself. That means code outside of the class will not be able to modify the dictionary's contents.
In order to modify the dictionary, I added the method update() as an example
If people want to, they can modify A._data directly, there is nothing we can do to prevent that.

Collectives™ on Stack Overflow

How to make dictionary read-only?

6 Answers 6

3 Comments

6 Comments

3 Comments

1 Comment

Comments

Notes

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

6 Answers 6

3 Comments

6 Comments

3 Comments

1 Comment

Comments

Notes

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related