Nested classes are not serializable in python when trying JSON dump

Question

I currently have two classes in Python like these ones

class person:
    age=""
    name=""
    ranking = {}

    def addRanking():
    #Do Whatever treatment and add to the ranking dict

class ranking:
    semester = ""
    position = ""
    gpa = ""

I have my list of person as a dictionary called dictP json.dumps() this dictionary but it seems that it doesn't work. Here is my function to dump to JSON

def toJson():
    jsonfile = open('dict.json', 'w')
    print(json.dump(listP, jsonfile))

I get the famous: is not JSON serializable.

Would you know what I can do to help this problem. I thought that having two dictionaries (which are serializable) would avoid this kind of issue, but apparently not.

Thanks in advance

Edit:

Here is an example (typed on my phone sorry for typos, I'm not sure it does run but it's so you get the idea):

class person:
    age=""
    name=""
    ranking = {}

    def __init__(self, age, name):
        self.age = age
        self.name = name
        self.ranking = {}

    def addRanking(self,semester,position,gpa):
        #if the semester is not already present in the data for that person
            self.ranking[semester] = make_ranking(semester,position,gpa)

class ranking:
    semester = ""
    position = ""
    gpa = ""

    def __init__(self, semester, position, gpa):
        self.semester = semester
        self.position = position
        self.gpa = gpa

dictP = {}


def make_person(age, name):
    # Some stuff happens there
    return person(age,name)

def make_ranking(semester,postion,gpa):
    #some computation there
    return ranking(semester,position,gpa)

def pretending_to_read_csv():
    age = 12
    name = "Alice"
    p = make_person(age, name)
    dictP["1"] = p

    age = 13
    name = "Alice"
    p = make_person(age, name)
    dictP["2"] = p

    #We read a csv for ranking that gives us an ID 
    semester = 1
    position = 4 
    gpa = 3.2
    id = 1 

    dictP["1"].addRanking(semester, position, gpa)

    semester = 2
    position = 4 
    gpa = 3.2
    id = 1 

    dictP["1"].addRanking(semester, position, gpa)

chris · Accepted Answer · 2018-08-09 00:48:54Z

1

For a dictionary to be serializable, note that all the keys & values in that dictionary must be serializable as well. You did not show us what listP contains, but I'm guessing it's something like this:

>>> listP
[<__main__.person instance at 0x107b65290>, <__main__.person instance at 0x107b65368>]

Python instances are not serializable.

I think you want a list of dictionaries, which would look like this:

>>> listP
[{'ranking': {}, 'age': 10, 'name': 'fred'}, {'ranking': {}, 'age': 20, 'name': 'mary'}]

This would serialize as you expect:

>>> import json
>>> json.dumps(listP)
'[{"ranking": {}, "age": 10, "name": "fred"}, {"ranking": {}, "age": 20, "name": "mary"}]'

UPDATE

(Thanks for adding example code.)

>>> pretending_to_read_csv()
>>> dictP
{'1': <__main__.person instance at 0x107b65368>, '2': <__main__.person instance at 0x107b863b0>}

Recall that user-defined classes cannot be serialized automatically. It's possible to extend the JSONEncoder directly to handle these cases, but all you really need is a function that can turn your object into a dictionary comprised entirely of primitives.

def convert_ranking(ranking):
    return {
        "semester": ranking.semester,
        "position": ranking.position,
        "gpa": ranking.gpa}

def convert_person(person):
    return {
        "age": person.age,
        "name": person.name,
        "ranking": {semester: convert_ranking(ranking) for semester, ranking in person.ranking.iteritems()}}

One more dictionary comprehension to actually do the conversion and you're all set:

>>> new_dict = {person_id: convert_person(person) for person_id, person in dictP.iteritems()}
>>> from pprint import pprint
>>> pprint(new_dict)
{'1': {'age': 12,
       'name': 'Alice',
       'ranking': {1: {'gpa': 3.2, 'position': 4, 'semester': 1},
                   2: {'gpa': 3.2, 'position': 4, 'semester': 2}}},
 '2': {'age': 13, 'name': 'Alice', 'ranking': {}}}

Since no user-defined objects are stuffed in there, this will serialize as you hope:

>>> json.dumps(new_dict)
'{"1": {"ranking": {"1": {"position": 4, "semester": 1, "gpa": 3.2}, "2": {"position": 4, "semester": 2, "gpa": 3.2}}, "age": 12, "name": "Alice"}, "2": {"ranking": {}, "age": 13, "name": "Alice"}}'

edited Aug 9, 2018 at 0:48

answered Aug 8, 2018 at 18:11

chris

2,0832 gold badges15 silver badges22 bronze badges

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

11 Comments

LBes Over a year ago

Currently on the phone so it’s hard to look at my code or comment more (or even make sense of everything you said) but my ListP contains an ID and an instance of Perso.

chris Over a year ago

Can you add a copy of the listP you're working with to your question?

LBes Over a year ago

well I just build it on the fly while reading a bunch of CSVs but it's basically the key is an ID and the value is an instance of Person

chris Over a year ago

Ok, that is the root of your issue: an instance of your person class is not serializable, whether it's the value of a dictionary or an element in a list.

LBes Over a year ago

so I have my dictionary with instances of person and then each instance of person got instances of rankings... How should I proceed. This thing is seriously confusing me... Can you edit your answer ?

|

Andrew Delgadillo · Accepted Answer · 2018-08-08 18:16:50Z

1

You can try calling json.dump on the .__dict__ member of your instance. You say that you have a list of person instances so try doing something like this:

listJSON = []
for p in listP
    #append the value of the dictionary containing data about your person instance to a list
    listJSON.append(p.__dict__)
json.dump(listJSON, jsonfile)

If you are storing your person instances in a dictionary like so: dictP = {'person1': p1, 'person2': p2} this solution will loop through the keys and change their corresponding values to the __dict__ member of the instance:

for key in dictP:
    dictP[key] = dictP[key].__dict__
json.dump(dictP, jsonfile)

edited Aug 8, 2018 at 18:16

answered Aug 8, 2018 at 18:06

Andrew Delgadillo

1612 gold badges3 silver badges12 bronze badges

2 Comments

LBes Over a year ago

I actually have a dictionary of Person. So what would happen then ? Does t change anything to call .__dict__?

LBes Over a year ago

will try that in a couple of hours. Currently on the phone.

Collectives™ on Stack Overflow

Nested classes are not serializable in python when trying JSON dump

2 Answers 2

11 Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

11 Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related