cerealbox

CerealBox for Python

CerealBox is a blazingly fast Zero Dependency generic Serializer / Deserializer for python dictionaries. It has an extendable architecture that allows custom serializers to be built through config. The module also includes built in implementations of serializing common data types to a JSON compatible dictionary or DynamoDB JSON.

Getting started

Install the module

Using poetry

poetry add cerealbox

or using pip

pip install cerealbox

Using the built-in serializers

jsonable

The jsonable serializer converts any input dict or value into JSON serializable output value.

from cerealbox.jsonable import as_jsonable
from decimal import Decimal
from enum import Enum
from datetime import datetime


class Country(Enum):
    ZA = 'South Africa'
    AU = 'Australia'
    US = 'United States'


sample_input = {
    "name": "Jane",
    "age": 23,
    "balance": Decimal('250.10'),
    "country": Country.ZA,
    "updated_at": datetime(2020, 1, 1)
}

print(as_jsonable(sample_input))
# {'name': 'Jane', 'age': 23, 'balance': '250.10', 'country': 'South Africa', 'updated_at': '2020-01-01T00:00:00'}

# You can also use as_jsonable as the default function to json.dumps 
import json

print(json.dumps(sample_input, default=as_jsonable, indent=4))
# {
#     "name": "Jane",
#     "age": 23,
#     "balance": "250.10",
#     "country": "South Africa",
#     "updated_at": "2020-01-01T00:00:00"
# }

The default encoders for as_jsonable are as follows:

     | Python            | JSONABLE        |
     +===================+=================+
     | dict              | dict            |
     | list, tuple       | list            |
     | set               | list            |
     | string            | string          |
     | int, float        | int, float      |
     | bool              | bool            |
     | None              | None            |
     | Decimal           | string          |
     | datetime          | string (iso)    |
     | enum              | string (value)  |
     | uuid              | string          |

DynamoDB

The DynamoDB Serializer/Deserializer is capable of transforming python values into DynamoDB JSON and back. It supports most common data types. Some transformations are not reversible (eg converting a datetime to a string). This limitation is due to cerealbox being schemaless, and can be overcome by using a module such as typed-models or pydantic

from cerealbox.dynamo import from_dynamodb_json, as_dynamodb_json
from decimal import Decimal
from enum import Enum
from datetime import datetime
from pprint import pprint


class Country(Enum):
    ZA = 'South Africa'
    AU = 'Australia'
    US = 'United States'


sample_input = {
    "name": "Jane",
    "age": 23,
    "balance": Decimal('250.10'),
    "country": Country.ZA,
    "updated_at": datetime(2020, 1, 1)
}

ddb_json = as_dynamodb_json(sample_input)
pprint(ddb_json)
# {'M': {'age': {'N': '23'},
#        'balance': {'N': '250.10'},
#        'country': {'S': 'South Africa'},
#        'name': {'S': 'Jane'},
#        'updated_at': {'S': '2020-01-01T00:00:00'}}}

# Reversing the operation

pprint(from_dynamodb_json(ddb_json))

# {'age': Decimal('23'),
#  'balance': Decimal('250.10'),
#  'country': 'South Africa',
#  'name': 'Jane',
#  'updated_at': '2020-01-01T00:00:00'}

When serializing from a dictionary to DynamoDB JSON, the following mapping is used:

   Python                                  DynamoDB
   ------                                  --------
   None                                    {'NULL': True}
   True/False                              {'BOOL': True/False}
   int/Decimal                             {'N': str(value)}
   string                                  {'S': string}
   Binary/bytearray/bytes (py3 only)       {'B': bytes}
   set([int/Decimal])                      {'NS': [str(value)]}
   set([string])                           {'SS': [string])
   set([Binary/bytearray/bytes])           {'BS': [bytes]}
   list                                    {'L': list}
   dict                                    {'M': dict}
   float                                   {'S': str(value)}
   datetime/date/time                      {'S': str(value.isoformat())}
   Enum                                    {'S': str(value.value)}
   UUID                                    {'S': str(value)}

When serializing from DynamoDB JSON to a Python dict, the following mapping is used:

   DynamoDB                                Python
   --------                                ------
   {'NULL': True}                          None
   {'BOOL': True/False}                    True/False
   {'N': str(value)}                       Decimal(str(value))
   {'S': string}                           string
   {'B': bytes}                            Binary(bytes)
   {'NS': [str(value)]}                    set([Decimal(str(value))])
   {'SS': [string]}                        set([string])
   {'BS': [bytes]}                         set([bytes])
   {'L': list}                             list
   {'M': dict}                             dict

Writing your own serializer

A serializer is made up of a dict that maps each datatype to a function that produces its serialized version. There are 3 special cases for these functions:

1. If the type and the mapped function are the same (eg {str: str}), the value is not modified during serialization. This is a performance optimization.
2. Dealing with the value None is a special case, since type(None) is NoneType. If you would like to handle None , import NoneType from cerealbox and use it as the type
3. If a type maps to a function that accepts a parameter named serialize, an instance of the serializer is passed along with the function. This allows recursive calls to deal with items inside of dictionaries, lists etc.

Example use case:

When serializing a dict, convert all Decimal types to a String with the prefix $ . Redact any string that contains the word "classified". Handle nested items inside of a list in a similar manner

from cerealbox import Cereal
from decimal import Decimal
from pprint import pprint


def redact_strings(value):
    if 'classified' in value.lower():
        return "***classified***"

    return value


def serialize_list(value, serialize):
    return [serialize(item) for item in value]


ENCODERS = {
    str: redact_strings,
    Decimal: lambda num: f"$ {num}",
    list: serialize_list,
    dict: lambda v, serialize: {k_: serialize(v_) for k_, v_ in v.items()}
}

custom_serializer = Cereal(encoders=ENCODERS)

sample_input = {
    "name": "Jane",
    "assignment": "Eat Cereal. Mission is Classified",
    "funds": Decimal('1024.50'),
    "keywords": [Decimal('1.5'), "Hello, World", "I am classified."]
}

pprint(custom_serializer(sample_input))
# {'assignment': '***classified***',
#  'funds': '$ 1024.50',
#  'keywords': ['$ 1.5', 'Hello, World', '***classified***'],
#  'name': 'Jane'}

Extending an existing serializer

Extend jsonable to redact strings containing the word Classified

from cerealbox.jsonable import as_jsonable


def redact_strings(value):
    if 'classified' in value.lower():
        return "***classified***"

    return value


as_jsonable.extend_encoders({str: redact_strings})

sample_input = {
    "name": "Jane",
    "age": 23,
    "mission": "[Classified] Divide by zero and see what happens.",
}

print(as_jsonable(sample_input))
# {'name': 'Jane', 'age': 23, 'mission': '***classified***'}

DynamoDB JSON Serializer/Deserializer Benchmarks

cerealbox has crude benchmarks against boto3 (using TypeSerializer and TypeDeserializer) and dynamodb-json. The benchmark calculates the roundtrip conversion from a python dict to DynamoDB JSON and back to a python dict. See ./benchmarks

package	version	relative performance	mean time
cerealbox	0.1.2	-	102.4 uS
boto3	1.18.30	2.69x slower	275.2 uS
dynamodb-json	1.3	7.36x slower	754.0 uS

Contributing

To work on the cerealbox codebase, you'll want to clone the project locally and install the required dependencies via poetry.

git clone [e-mail]:a2d24/cerealbox.git