Prodict

Prodict = Dictionary with IDE friendly(auto code completion) and dot-accessible attributes and more.

What it does

Ever wanted to use a dict like a class and access keys as attributes? Prodict does exactly this.

Although there are number of modules doing this, Prodict does a little bit more.

You can provide type hints and get auto code completion!

With type hints, you also get nested object instantiations, which will blow your mind.

You will never want to use dict again.

Why?

Because accessing dict keys like d['key'] is error prone and ugly.
Because it becomes uglier if it is nested, like d['key1]['key2']['key3']. Compare d['key1]['key2']['key3'] to d.key1.key2.key3, which one looks better?
Because since web technologies mostly talk with JSON, it should be much more easy to use JSON data(see sample use case below).
Because auto code completion makes developers' life easier.
Because serializing a Python class to dict and deserializing from dict in one line is awesome!

Features

A class with dynamic properties, without defining it beforehand.

j = Prodict()
j.hi = 'there'

Pass named arguments and all arguments will become properties.

p = Prodict(lang='Python', pros='Rocks!')
print(p.lang)  # Python
print(p.pros)  # Rocks!
print(p)  # {'lang': 'Python', 'pros': 'Rocks!'}

Instantiate from a dict, get dict keys as properties

p = Prodict.from_dict({'lang': 'Python', 'pros': 'Rocks!'})
print(p.lang)   # Python
p.another_property = 'this is dynamically added'

Pass a dict as argument, get a nested Prodict!

p = Prodict(package='Prodict', makes='Python', rock={'even': 'more!'})
print(p)  #  {'package': 'Prodict', 'makes': 'Python', 'rock': {'even': 'more!'}}
print(p.rock.even)  #  'more!'
print(type(p.rock))  # <class 'prodict.Prodict'>

Extend Prodict and use type annotations for auto type conversion and auto code completion

class User(Prodict):
    user_id: int
    name: str

user = User(user_id="1", name="Ramazan")
type(user.user_id) # <class 'int'>
# IDE will be able to auto complete 'user_id' and 'name' properties(see example 1 below)

Why type conversion? Because it will be useful if the incoming data doesn't have the desired type.

class User(Prodict):
    user_id: int
    name: str
    literal: Any
    
response = requests.get("https://some.restservice.com/user/1").json()
user: User = User.from_dict(response)
type(user.user_id) # <class 'int'>

Notes on automatic type conversion:

In the above example code, user.user_id will be an int, even if rest service responded with str.
Same goes for all built-in types(int, str, float, bool, list, tuple), except dict. Because by default, all dict types will be converted to Prodict.
If you don't want any type conversion but still want to have auto code completion, use Any as type annotation, like the literal attribute defined in User class.
If the annotated type of an attribute is sub-class of a Prodict, the provided dict will be instantiated as the instance of sub-class. Even if it is List of the sub-class(see sample usa case below).

Sample use case

Suppose that you are getting this JSON response from https://some.restservice.com/user/1:

{
  user_id: 1,
  user_name: "rambo",
  posts: [
    {
      title:"Hello World",
      text:"This is my first blog post...",
      date:"2018-01-02 03:04:05",
      comments: [
          {
            user_id:2,
            comment:"Good to see you blogging",
            date:"2018-01-02 03:04:06"
          },
          {
            user_id:3,
            comment:"Good for you",
            date:"2018-01-02 03:04:07"
          }
        ]
    },
    {
      title:"Leave the old behind",
      text:"Stop using Python 2.x...",
      date:"2018-02-03 04:05:06",
      comments: [
          {
            user_id:4,
            comment:"Python 2 is dead, long live Python",
            date:"2018-02-03 04:05:07"
          },
          {
            user_id:5,
            comment:"You are god damn right :wears Heissenberg glasses:",
            date:"2018-02-03 04:05:08"
          }
        ]
    }
  ]
}

Despite the fact that JSON being schemaless, most REST services will respond with a certain structure. In the above example, the structure is something like this:

User
 |--> user_id
 |--> user_name
 |--> posts [post]
       |--> title
       |--> text
       |--> date
       |--> comments [comment]
             |--> user_id
             |--> comment
             |--> date

And you want to convert this to appropriate Python classes.

Without Prodict:

class Comment:
	def __init__(self, user_id, comment, date):
    	self.user_id = user_id
    	self.comment = comment
        self.date = date
        
class Post:
    def __init__(self, title, text, date):
    	self.title = title
        self.text = text
        self.date = date
        self.comments = []
        
class User:
    def __init__(self, user_id, user_name):
        self.user_id = user_id
        self.user_name = user_name
        self.posts = []

user_json = requests.get("https://some.restservice.com/user/1").json()
posts = [Post(post['title'], post['text'], post['date']) for post in user_json['posts']]
for post in posts:
    post.comments = [[comment for comment in post['comments']] for post in user_json['posts']]

user = User(user_json['user_id'], user_json['user_name'])
user.posts = posts

for post in user.posts:
    print(post.title)

With Prodict you just need to define the classes and let the prodict do the rest like this:

class Comment(Prodict):
    user_id: int
    comment: str
    date: str

class Post(Prodict):
    title: str
    text: str
    date: str
    comments: List[Comment]
        
class User(Prodict):
    user_id: int
    user_name: str
    posts: List[Post]

user_json = requests.get("https://some.restservice.com/user/1").json()
user:User = User.from_dict(user_json)
# Don't forget to annotate the `user` with `User` type in order to get auto code completion.

See the difference? Plus you can add new attributes to User, Post and Comment objects dynamically and access them as dot-accessible attributes.

Examples

Example 0: Use it like regular dict, because it is a dict.

from prodict import Prodict

d = dict(lang='Python', pros='Rocks!')
p = Prodict(lang='Python', pros='Rocks!')

print(d)  # {'lang': 'Python', 'pros': 'Rocks!'}
print(p)  # {'lang': 'Python', 'pros': 'Rocks!'}
print(d == p)  # True

p2 = Prodict.from_dict({'Hello': 'world'})

print(p2)  # {'Hello': 'world'}
print(issubclass(Prodict, dict))  # True
print(isinstance(p, dict))  # True
print(set(dir(dict)).issubset(dir(Prodict)))  # True

Example 1: Accessing keys as attributes and auto completion.

from prodict import Prodict
class Country(Prodict):
    name: str
    population: int

turkey = Country()
turkey.name = 'Turkey'
turkey.population = 79814871

Example 2: Auto type conversion

germany = Country(name='Germany', population='82175700', flag_colors=['black', 'red', 'yellow'])

print(germany.population)  # 82175700
print(type(germany.population))  # <class 'int'> <-- The type is `int` !
# If you don't want type conversion and still want to have auto code completion, use `Any` as type.
print(germany.flag_colors)  # ['black', 'red', 'yellow']
print(type(germany.population))  # <class 'int'>

Example 3: Nested class instantiation

class Ram(Prodict):
    capacity: int
    unit: str
    type: str
    clock: int

class Computer(Prodict):
    name: str
    cpu_cores: int
    rams: List[Ram]

    def total_ram(self):
        return sum([ram.capacity for ram in self.rams])

comp1 = Computer.from_dict(
    {
        'name': 'My Computer',
        'cpu_cores': 4,
        'rams': [
            {'capacity': 4,
             'unit': 'GB',
             'type': 'DDR3',
             'clock': 2400}
        ]
    })
print(comp1.rams)  #  [{'capacity': 4, 'unit': 'GB', 'type': 'DDR3', 'clock': 2400}]

comp1.rams.append(Ram(capacity=8, type='DDR3'))
comp1.rams.append(Ram.from_dict({'capacity': 12, 'type': 'DDR3', 'clock': 2400}))

print(comp1.rams)
# [
#   {'capacity': 4, 'unit': 'GB', 'type': 'DDR3', 'clock': 2400}, 
#   {'capacity': 8, 'type': 'DDR3'}, 
#   {'capacity': 12, 'type': 'DDR3', 'clock': 2400}
# ]
print(type(comp1.rams))  # <class 'list'>
print(type(comp1.rams[0]))  # <class 'Ram'> <-- Mind the type !

Example 4: Provide default values

You can use init method to provide default values. Keep in mind that init is NOT __init__ but init method will be called in __init__ method.

Additionally, you can use init method instead of __init__ without referring to super.

class MyDataWithDefaults(Prodict):
    an_int: int
    a_str: str
    
    def init(self):
        self.an_int = 42
        self.a_str = 'string'
        

data = MyDataWithDefaults(dynamic=43)
print(data)
# {'an_int':42, 'a_str':'string', 'dynamic':43}

Class attributes vs Instance attributes

Prodict only works for instance attributes. Even if you try to set an inherited class attribute, a new instance attribute is created and set.

Consider this example:

from prodict import Prodict

class MyClass(Prodict):
    class_attr: int = 42  # class_attr is a class attribute, not instance attribute

my_class = MyClass()
print(f"my_class.class_attr: {my_class.class_attr}")  # 42
# There is no 'class_attr' defined as instance attribute, so class attribute will be returned (42).
print(f"MyClass.class_attr: {MyClass.class_attr}") # 42
# This is a class attribute, it will be returned as is.

# Now an instance attribute 'class_attr' is created and set to 77
my_class.class_attr = 77
print(f"my_class.class_attr: {my_class.class_attr}")  # 42
# For this matter, avoid setting class_attribute with dot notation, use class name instead

MyClass.class_attr = 88
print(f"MyClass.class_attr: {my_class.class_attr}")  # 88

# So where did 77 go? It is in instance attribute of the class and since it's name is colliding with
# the class attribute, you can't get it by dot notation. You can use .get tho.
print(f"my_class.get('class_attr'): {my_class.get('class_attr')}")  # 77

Installation

If your default Python is 3.7:

pip install prodict

If you have more than one Python versions installed:

python3.7 -m pip install prodict

Limitations

You cannot use dict method names as attribute names because of ambiguity.
You cannot use Prodict method names as attribute names(I will change Prodict method names with dunder names to reduce the limitation).
You must use valid variable names as Prodict attribute names(obviously). For example, while '1' cannot be an attribute for Prodict, it is perfectly valid for a dict to have '1' as a key. You can still use prodict.set_attribute('1',123) tho.
Requires Python 3.7+

Thanks

I would like to thank to JetBrains for creating PyCharm, the IDE that made my life better.

ramazanpolat/prodict

ramazanpolat

Reviews

Repository Details