How to use dot in field name?

Question

How to use dot in field name ?

I see error in example:

db.test2.insert({ "a.a" : "b" })

can't have . in field names [a.a]

Answer 1

You can replace dot symbols of your field name to Unicode equivalent of \uff0E

db.test.insert({"field\uff0ename": "test"})
db.test.find({"field\uff0ename": "test"}).forEach(printjson)
{ "_id" : ObjectId("5193c053e1cc0fd8a5ea413d"), "field．name" : "test" }

See more:

1. http://docs.mongodb.org/manual/faq/developers/#faq-dollar-sign-escaping
2. http://docs.mongodb.org/manual/core/document/#dot-notation

Answer 2

You can also write a SONManipulator using the pymongo library that transforms the data going to and back out of mongodb. There are downsides; there is a performance hit (impact depends on your use case) and you have to transform your keys when you do searches using find.

Here's code with an example of how to use it in the comment for the KeyTransform class:

from pymongo.son_manipulator import SONManipulator

class KeyTransform(SONManipulator):
    """Transforms keys going to database and restores them coming out.

    This allows keys with dots in them to be used (but does break searching on
    them unless the find command also uses the transform).

    Example & test:
        # To allow `.` (dots) in keys
        import pymongo
        client = pymongo.MongoClient("mongodb://localhost")
        db = client['delete_me']
        db.add_son_manipulator(KeyTransform(".", "_dot_"))
        db['mycol'].remove()
        db['mycol'].update({'_id': 1}, {'127.0.0.1': 'localhost'}, upsert=True,
                           manipulate=True)
        print db['mycol'].find().next()
        print db['mycol'].find({'127_dot_0_dot_0_dot_1': 'localhost'}).next()

    Note: transformation could be easily extended to be more complex.
    """

    def __init__(self, replace, replacement):
        self.replace = replace
        self.replacement = replacement

    def transform_key(self, key):
        """Transform key for saving to database."""
        return key.replace(self.replace, self.replacement)

    def revert_key(self, key):
        """Restore transformed key returning from database."""
        return key.replace(self.replacement, self.replace)

    def transform_incoming(self, son, collection):
        """Recursively replace all keys that need transforming."""
        for (key, value) in son.items():
            if self.replace in key:
                if isinstance(value, dict):
                    son[self.transform_key(key)] = self.transform_incoming(
                        son.pop(key), collection)
                else:
                    son[self.transform_key(key)] = son.pop(key)
            elif isinstance(value, dict):  # recurse into sub-docs
                son[key] = self.transform_incoming(value, collection)
        return son

    def transform_outgoing(self, son, collection):
        """Recursively restore all transformed keys."""
        for (key, value) in son.items():
            if self.replacement in key:
                if isinstance(value, dict):
                    son[self.revert_key(key)] = self.transform_outgoing(
                        son.pop(key), collection)
                else:
                    son[self.revert_key(key)] = son.pop(key)
            elif isinstance(value, dict):  # recurse into sub-docs
                son[key] = self.transform_outgoing(value, collection)
        return son

Answer 3

Actualy you may use dots in queries. See: http://www.mongodb.org/display/DOCS/Dot+Notation+%28Reaching+into+Objects%29

Because of this special dot symbol mean you cannot use it in field names. Like you cannot use dot symbol in identifiers in most of programming languages.

You may write query db.test2.find({ "a.a" : "b" }) but if you want to be able to write such a query you need to insert your object like so: db.test2.insert({"a": {"a": "b"}}). This will create document with the field named "a" with the value of embeded document containing the field named "a" (again) with the value "b".

Answer 4

def remove_dots(data):
    for key in data.keys():
        if type(data[key]) is dict: data[key] = remove_dots(data[key])
        if '.' in key:
            data[key.replace('.', '\uff0E')] = data[key]
            del data[key]
    return data

this recursive method replaces all dot characters from keys of a dict with \uff0E as suggested by Fisk

Answer 5

I replaced the key value using myString.replace(".","\u2024") before inserting it into the JsonObject.

Answer 6

Initially I used a simple recursion to replace all "." characters with its unicode equivalent but figured it out that even the dots in the values was getting replaced. So I thought that we should replace the dots only from keys and made the changes accordingly in case "if isinstance(input, dict)". I thought it should be a sufficient condition to do the magic but I forgot that dict value can also be a dict or a list and then I finally added that check that if value of a dict was not string then, go inside recursively and was finally able to come up with this solution which eventually did the trick.

def remove_dots(data):
    if isinstance(data, dict):
            return {remove_dots(key): value if isinstance(value, str) else remove_dots(value) for key,value in data.iteritems()}
    elif isinstance(data, list):
            return [remove_dots(element) for element in data]
    elif isinstance(data, str):
            return data.replace('.','\u002e')
    else:                                                                                             
            return data

Answer 7

I've only really come across this problem when trying to serialize Dictionaries and such where the offending dot can appear as a key name. Edited to show the references.

The quick and dirty C# approach:

using MongoDB.Bson;
using Newtonsoft.Json.Linq;
using System.Text.RegularExpressions;

public static T Sanitize<T>(T obj)
{
      var str = JObject.FromObject(obj).ToJson();
      var parsed = Regex.Replace(str, @"\.(?=[^""]*"":)", "_");   //i.e. replace dot with underscore when found as a json property name { "property.name": "don't.care.what.the.value.is" }
      return JObject.Parse(parsed).ToObject<T>();
}