Sqlalchemy Hack - How to convert a table to dict on the fly


In on of my recent projects, I came across the need to develop a JSON based REST API to fetch data from the sqlalchemy objects. Now, the Query object is a great way to access data using the powerful sqlalchemy orm, but it doesn’t give any built-in way to convert the result-set into a python dict.

For instance, I have a Professor table in my models.py. Here is what I get when I query the Professor model using sqlalchemy ORM:

>>> session.query(models.Professor).all()
2015-07-05 11:07:57,282 INFO sqlalchemy.engine.base.Engine SELECT professors.id AS professors_id, professors.name AS professors_name,
partment, professors.email AS professors_email, professors.password AS professors_password, professors.phone AS professors_phone
FROM professors
2015-07-05 11:07:57,282 INFO sqlalchemy.engine.base.Engine ()
[<Professor(name=`Albus Dumbledore`)>, <Professor(name=`Severus Snape`)>]

At the end, I got the two Professors’ name in a collection, but they are still sqlalchemy objects. How can I convert it to a python dict, so that it can be deserialized to JSON or be used for any other purpose. To solve this, just add the below code to your models.py module, and just call the models.to_dict() method to convert an sqlalchemy models collection to a dict (or alternatively, call models.from_dict() to build the model object collection from an existing dict!):

def to_dict(model_instance, query_instance=None):
    if hasattr(model_instance, '__table__'):
        return {c.name: str(getattr(model_instance, c.name)) for c in model_instance.__table__.columns}
    else:
        cols = query_instance.column_descriptions
        return { cols[i]['name'] : model_instance[i]  for i in range(len(cols)) }

def from_dict(dict, model_instance):
    for c in model_instance.__table__.columns:
        setattr(model_instance, c.name, dict[c.name])

The usage is as follows:

q = dbsession.query(models.Application).filter(models.Application.id==id)
professors = q.all()
di = models.to_dict(professors) #for converting a single table resultset

q = dbsession.query(models.Application)
application = q.first()
dd = models.to_dict(app)
dd['student'] = models.to_dict(application.student, q) #for converting a relationship object which refers to another table.

The reason we have to use a slightly different approach for the relationship objects (like application.student which refers to the student model) is that the instance doesn’t have an __table__ object which is required to build the dict. Hence, the query instance is passed. This is just one approach I’ve used to convert sqlalchemy objects to dict. If you have any other approach, please let me know.


Comments !