8000 GitHub - jeff-hykin/json_fix: 💾 📦 ✅ A fix for the 1000 upvotes python-json question
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

💾 📦 ✅ A fix for the 1000 upvotes python-json question

Notifications You must be signed in to change notification settings

jeff-hykin/json_fix

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

What is this?

A pip module that lets you define a __json__ method, that works like toJSON from JavaScript.
(e.g. it magically gets called whenever someone does json.dumps(your_object))

From a technical perspective, this module is a safe, backwards-compatible, reversable patch to the built-in python json object that allows classes to specify how they should be serialized.

Why?

Because sometimes someone eles's code (e.g. a pip module) tries to serialize your object, like

import json
json.dumps(list_containing_your_object)

And you'd have to fork their code to make it not throw an error.

How do I use this tool?

pip install json-fix

import json_fix # import this any time (any where) before the JSON.dumps gets called

# same file, or different file
class YOUR_CLASS:
    def __json__(self):
        # YOUR CUSTOM CODE HERE
        #    you probably just want to do:
        #        return self.__dict__
        return "a built-in object that is natually json-able"

How do I change how imported classes (numpy array, dataframe, etc) are jsonified?

There's 2 ways; the aggressive override_table or the more collaboration-friendly fallback_table. Note: some really powerful stuff can be done safely with the fallback table!

Override Table

CAUTION!

  • The override table is so powerful it even lets you change how built in objects are serialized. You can screw yourself (infinite loop) if you're using (abusing) the override table to change giant swaths of objects instead of specific classes.
  • Even if a class defines a .__json__() method, the json.override_table will take priority.
  • The order of keys matters a lot. The last entry takes the highest priority (this lets us override pip modules even if they try using the override table).

The override table is a dictionary. It has "check functions" as keys, and jsonifiers as values.

import json_fix # import this before the JSON.dumps gets called
import json
import pandas as pd

SomeClassYouDidntDefine = pd.DataFrame

# create a boolean function for identifying the class
check_func = lambda obj: isinstance(obj, SomeClassYouDidntDefine)
# then assign it to a function that does the converting
json.override_table[check_func] = lambda obj_of_that_class: json.loads(obj_of_that_class.to_json())

json.dumps([ 1, 2, SomeClassYouDidntDefine() ], indent=2) # dumps as expected

Fallback Table

If you want all python classes to be jsonable by default, we can easily do that with the fallback table. The logic is if nothing in override table, and no .__json__ method, then check the fallback table.

import json_fix # import this before the JSON.dumps gets called
import json

# a checker for custom objects
checker = lambda obj: hasattr(obj, "__dict__")
# use the __dict__ when they don't specify a __json__ method 
json.fallback_table[checker] = lambda obj_with_dict: obj_with_dict.__dict__

class SomeClass:
    def __init__(self):
        self.thing = 10

json.dumps([ 1, 2, SomeClass() ], indent=2) # dumps as expected

Like the override table, the most recently-added checker will have the highest priority.

0