SQLAlchemy DST is a set of must have functions and classes for the dictionary serialisation. It allows you to quickly convert a SQLAlchemy model to a python dictionary and vise versa.

Why use it? Well, for example, I use this tools to reduce amount of requests to a database. So, this is a simple kind of caching complex objects in a session.

This module successfully passed unit testing on Debian Jessie with Python 2.7 and 3.5 versions.

Depends on the following packages (see requirements.txt):

• sqlalchemy >= 0.9.4

This module is available as PyPi package, therefore you can install one as follows:

$ pip install sqlalchemy-dst

SQLAlchemy DST use only two functions (row2dict() and dict2row()) to serialize objects. Let's see how it works...

Imagine you have a User model with several relations (Role, Permissions), most likely, you will want to store instance of the model in a web server session for efficency purposes. Unfortunatelly, it's not possible to store instances of Python classes in the session. So, you need to serialize one into a dictionary first:

>>> from sqlalchemy_dst import dict2row, row2dict
>>> from tests.main import db
>>> from tests.schema import Permission, Role, User
>>> user = db.session.query(User).first() # create an instance of the User model

Serialize the instance of the User model into the one-dimensional dictionary:

>>> row2dict(user)
{'username': 'yarbshk', 'role': None, 'role_id': 1, '_secret': 'x', 'password': 'x', 'id': 1}

Serialize the instance of the User model into the three-dimensional dictionary and exclude a few attributes using different methods:

>>> row2dict(user, depth=3, exclude={'role_id', '_secret'})
{'username': 'yarbshk', 'password': 'x', 'id': 1, 'role': {'description': None, 'id': 1, 'permissions': [{'id': 1, 'name': 'posts:r'}, {'id': 2, 'name': 'posts:w'}], 'name': 'Moderator'}}
>>> row2dict(user, depth=3, exclude_pk=True, exclude_underscore=True)
{'username': 'yarbshk', 'password': 'x', 'id': 1, 'role': {'description': None, 'id': 1, 'permissions': [{'id': 1, 'name': 'posts:r'}, {'id': 2, 'name': 'posts:w'}], 'name': 'Moderator'}}

Serialize the instance of the User model and store one in a local variable to work with:

>>> user_dict = row2dict(user, depth=3)
>>> user_dict
{'username': 'yarbshk', 'role': {'description': None, 'id': 1, 'permissions': [{'id': 1, 'name': 'posts:r'}, {'id': 2, 'name': 'posts:w'}], 'name': 'Moderator'}, 'role_id': 1, '_secret': 'x', 'password': 'x', 'id': 1}

Suppose in other part of the application you need to deserialize that dictionary to be able to use all features of the User's model (e.g. calling methods, querying data):

>>> dict2row(user_dict, User)
<User (transient 139927902911456)>

The example above will create new instance of the User model without any relations (such as role or permissions). To create a user's instance with all relations it's necessary to explicitly specify the models (Role, Permission) for that relations:

>>> user_row = dict2row(user_dict, User, rel={'role': Role, 'permissions': Permission})
<User (transient 139927902904272)>

Well, now you are able to do any manipulations with this instance in the same way as before serialization:

>>> [p.name for p in user_row.role.permissions]
['posts:r', 'posts:w']

Awesome, isn't it? :)

If you want to exclude some attribute of model which has synonym (and vice versa) you MUST exclude both attribute and synonym (in other case SQLAlchemy automatically sets the same value for attribute and their synonym even if one of them is excluded):

>>> dict2row(user_dict, User, exclude={'_secret', 'password'})

Note that default values of the optional arguments in methods below are setting implicitly. This behavior is required by class DictionarySerializableModel (see detailed explanation below).

Recursively walk row attributes to serialize ones into a dict.

• row (required) – instance of the declarative base class (base SQLAlchemy model).
• depth (optional, default: 1) – number that represent the depth of related relationships.
• exclude (optional, default: set()) – set of attributes names to exclude.
• exclude_pk (optional, default: False) – are foreign keys (e.g. fk_name_id) excluded.
• exclude_underscore (optional, default: False) – are private and protected attributes excluded.
• only (optional, default: set()) – set of attributes names to include.
• fk_suffix (optional, default: _id) – str that represent a foreign key suffix.

Recursively walk dict attributes to serialize ones into a row.

• d (required) – dict which represent a serialized row.
• model (required) – class nested from the declarative base class.
• rel (optional, default: dict()) – dict of key (relationship name) -value (class) pairs.
• exclude (optional, default: set()) – set of attributes names to exclude.
• exclude_pk (optional, default: False) – are foreign keys (e.g. fk_name_id) excluded.
• exclude_underscore (optional, default: False) – are private and protected attributes excluded.
• only (optional, default: set()) – set of attributes names to include.
• fk_suffix (optional, default: _id) – str that represent a foreign key suffix.

Class that extends serialization functionality of your models.

• as_dict(**kwargs) – wrapper around the row2dict() method – where kwargs are mapped optional arguments for the row2dict() method.
• from_dict(d, **kwargs) – wrapper around the dict2row() method – where d is source dictionary, kwargs are mapped optional arguments for the dict2row() method.
• __iter__() – object that can be iterated upon (it uses dictionary serialized by the row2dict() method).

Use it as a base class for the sqlalchemy.ext.declarative_base() method (try to explore the cls argument in depth).

If you decide to use the DictionarySerializableModel class as a base model, you may keep frequently used arguments values of the serialization methods in your model(s). Just set necessary configuraion attributes in your model(s) as follows:

from tests.main import db
...
class User(db.Model):
    ...
    _sa_dst_exclude_pk = True
    _sa_dst_exclude_underscore = True
    ...

The class declaration above is equal to calling the following method:

>>> user.as_dict(exclude_pk=True, exclude_underscore=True)
{'username': 'yarbshk', 'id': 1, 'password': 'x', 'role': None}

After, we're simply calling the method below without any parameters, because configuration attributes already are set in the model above:

>>> user.as_dict()
{'username': 'yarbshk', 'id': 1, 'password': 'x', 'role': None}

The list of available configuration attributes:

• _sa_dst_depth
• _sa_dst_exclude
• _sa_dst_exclude_pk
• _sa_dst_exclude_underscore
• _sa_dst_only
• _sa_dst_rel
• _sa_dst_fk_suffix

You can see an example of instantiating Flask + SQLAlchemy with DictionarySerializableModel in the tests/main.py file.

Copyright (c) 2018 Yuriy Rabeshko. Code released under the MIT license.