Skip to content


A model is an object representation of a database table. We will need a model
corresponding to each table in the ER diagram:

ER diagram Figure 13: ER diagram

When adding a model to a Flask application, the workflow consists of defining the model in code and then calling a convenience script to create the corresponding database table. This process is called migration and helps to eliminate human error. The term migration is also used in other contexts such as when moving code changes from one environment to another, but with Flask it is applied only to database changes.

A first model

We will start by creating subject_group which is one of the simplest models in our context. It has two properties:

  1. id: a synthetic primary key
  2. name: a string value representing the name of the subject group

Create a file called models/ and paste in the following code:

from app import db

class SubjectGroup(db.Model):
    __tablename__ = 'subject_group'

    id = db.Column(db.Integer, primary_key=True)
    name = db.Column(db.String(60), unique=True, nullable=False)

    def __repr__(self):
        return '{}'.format(


Line 1: Import the db object that was created in the app/ file

Line 4: Create a Python class to represent a subject group. Note that it is a subclass of the generic db.Model.

Line 5: Explicitly set the name to use for the corresponding database table

Line 7: Define the id property. By default, this will create an interger column in the database table with the AUTO_INCREMENT property set. This means that key values will be generated automatically for new rows in the table.

Line 8: Define the name property. The parameters define its maximum length to be 60 characters and require it to be unique and not nullable.

Lines 10 - 11: The __repr__() function defines the string representaton of an object of this class when it is rendered using the Python print() function.

We have set up the directory structure of the project so that each model is defined in a separate file. To ensure that it can be found by import statements, we need to turn the models directory into a Python package and provide an explicit link to the model. This may seem tedious, but the alternative would be to put all models into the same file. For simple application, that might be manageable but forlarger applications it makes maintenance more difficult. To make the modification, create the file models/ and add the following code:

from .subject_group import *

Notice that the reference to the file starts with a dot to show that it is located in the same directory as the current file, models/

Bootstrapping migrations

Migrations allow us to manage changes we make to the models, and propagate these changes to the database. For example, if we make a change to a property in one of the models, all we will need to do is create and apply a migration, and the database will reflect the change.

First, we need to install the flask-migrate package, which will handle the database migrations using Alembic, a lightweight database migration tool. Alembic emits ALTER statements to a database thus implementing changes made to the models. It also auto-generates minimalistic migration scripts which can be used to reverse database changes or repeat the same changes on a server.

Use the same procedure as before to install flask-migrate through PyCharm. Once that is done, we also need to modify the app/ file as shown below.

# third-party imports
from flask import Flask
from flask_sqlalchemy import SQLAlchemy
from flask_migrate import Migrate

# local imports
from config import app_config

db = SQLAlchemy()

def create_app(config_name):
    app = Flask(__name__, instance_relative_config=True)

    migrate = Migrate(app, db)
    from app import models

    from .public import public as public_blueprint

    return app

The changes are

Line 4: Import the Migrate class

Line 18: Initialise a migrate object

Line 19: Import all models

The flask-migrate package gives us access to several commands including

flask db init which creates a migration repository in the current project directory tree. This command only needs to be run once.

flask db migrate which detects model changes and creates a migration script.

flask db upgrade which runs the migration script to propagate the changes to the database

These commands need to be executed at the command line with the appropriate virtual environment activated which is easy to do using the terminal panel in PyCharm. Open the terminal and run the first of the three commands:

flask db init

If you get a KeyError, check that you have set the terminal environment variables correctly as described at the end of the introduction.

If the initialisation is successful, a directory called migrations will be added below the project root. Next, execute the second command:

flask db migrate

The feedback messages should show that the new modelin has been detected and that a migration script has been generated. The script can ve viewed in the migrations/versions directory. To run the script, execute the third command:

flask db upgrade

The effect will be to create the subject_group table in the database which you can view using the database panel in PyCharm. Remember that you will need to refresh the panel contents to see the changes.


Once migrations are configured, they make it very easy to keep your database up to date as long as a) you stick carefully to the workfow, and b) nothing goes wrong.

The workflow is:

  1. Update the appropriate Python model file
  2. If it is a new model, add it to the models/ file
  3. Run flask db migrate
  4. Run flask db upgrade

Performing the equivalent migration on a server will be discussed in a later part of the tutorial.

There are several things that could go wrong. There might be some reason why a migration script does not complete, for example. If the whole script fails, it may be sufficient to try the flask db upgrade command a second time. However, in more serious cases, a script may be partially successful. Running the migration a second time may not work because with each migration, a version number is stored in the database in the table alembic_version. If that value is updated even though there is a problem with the script, it may be impossible to resolve automatically. You would need to sort the problem out manually taking care that the version numbers match when you have finished. As a last resort, you can remove the migrations folder and the database table and start again from scratch.

Note that the automatic migration feature will not detect

  • changes to table names
  • changes to column names
  • constraints without explicit names

Further reading


 Database Migrations with Alembic