A Sandox for Flask and Heroku

I am preparing for a new project, and am testing out the Flask framework. Additionally, I am becoming more comfortable with hosting via Heroku.

First, I followed this tutorial: https://www.geeksforgeeks.org/deploy-python-flask-app-on-heroku/.

As I go forward I am experimenting with Flask and Heroku some more.

What I've Learned so Far

Deploying the App

• To run locally, run app/main.py. In VSCode this can be done with F5.

• To push to the Heroku servers, simply push to the personal repository. On Heroku, I set it up to immediately run the version that is pushed to the main branch (if it passes CI, which is not yet set up).

• The site is https://eflask-app-067.herokuapp.com/, and it runs indefinitely.

Files

• The file .gitignore instructs git to ignore certain files, such as __pycache__ folders which are generated by python to store some information.

• The files Pipfile and Pipfile.lock were generated through the pipenv command, and appear to give information on dependencies and the environment.

• The file Procfile appears to tell the server what to run on startup; specifically, it runs gunicorn and targets the app variable in app/main.py.

• The file runtime.txt is used by Heroku. It contains info regarding which version of Python to use. Only some are available, and the one reccomended by the tutorial does not work.

• The file wsgi.py (alternatively, app.py) is the starting point for code execution.

• The folder app will contain code for the server to use, and resources like HTML, CSS, and images that are sent.

• The .env file (which is sensitive and not sent to github) contains environment variables for running the site locally. I believe I needed to install python-dotenv for the file to actually be loaded. Instead of a .env file, Heroku uses config variables that can be configured through the CLI or through the site in Settings. If a variable is not present, os.getenv() will return None, so it is import to handle this if necessary.

Flask 101

• Proper execution starts via a call to app.run() (where app is the Flask instance created in main.py). However, we need not call this function itself anywhere. For production, the Procfile tells gunicorn to run this variable, and VSCode (using flask) knows to run it as well.

• The app variable is set up in main.py. Setting it up entails defining functions that are to be called at different times to generate the site's response. These functions are linked to app by the @app.route() decorator, which takes a string representing which url(s) should trigger that function.

• For debuging purposes, it is possible that an endpoint function can send raw HTML in the form of a string. The \deprecated endpoint does this, for example.

• Alternatively, you can write HTML in seperate files (in fact, you can write templates, but normal HTML also works). The render_template() function generates the full response object to be sent. See \sample or the homepage.

• You can also send a redirect message via redirect(). Use url_for() to create the right url (that function takes the name (string) of another endpoint function). See \redirect

Request and Session

• Flask exposes two important variables to functions running in the app.route() decorator: request and session. This resource is very useful for understanding these: https://flask.palletsprojects.com/en/2.0.x/quickstart/.

• request contains info about the request. For example, if the user sends a form, the info for that form can be obtained via request.form["form_item's_name"]. I used this to get input for the /login route.

• request also contains url arguments with request.args["param"], which I used to make the login page return you to your original target.

• session is a python dictionary, except it is aware of its own modification and sends / updates encrypted cookies when it is changed. You need to set app.secret_key to use it, however. I used this to store whether or not a user has logged in.

Python

• Flask makes use of decorators, which are a very useful construct. In essence, a decorating function wraps an old function, so that calls to the old function call the wrapper, which (depending on context) may call the inner function. See https://realpython.com/primer-on-python-decorators/ for more info.
  • See the require_password decorator in login.py for an example of how these can be useful. This decorator can be added to any endpoint function to force the user to login before accessing the endpoint.

Some more links to consider

• https://flask.palletsprojects.com/en/2.0.x/quickstart/#sessions

• https://elements.heroku.com/addons/heroku-postgresql