How To Properly Deal With Config Settings // Python Tips

#arjancodes #softwaredesign #python

[zolika]

For programmers like myself, who are self taught, little pieces of wisdom like this advice, mean a lot :) thank you

[DistortedV]

Not gonna lie, I like these shorts sometimes more than the videos, because I view your videos after work and like this concise format. Please do more of these if possible.

[laml]

That's a very simple yet very useful tip! Thanks for sharing! Loved this shorter format with tips

[KaJaKTrial]

Good idea! Using ini files is what I often use - there is a built in configparser library to handle them easily ;)

[jmann]

Also makes it a lot easier to refactor so that the configuration can be gathered from an external service or given via CL arguments! (which you inevitably will have to do if your model goes to production or you want to gather more extensive validation)

[bunchofmax]

Seriously good contents, I hope every developer should do this. I've been using yaml to take care of all the parameters for our ML pipelines.

[samjay]

This can be super useful. For example at my company we have python modules setup to be clients within a web service. This method allows us to build with our python files and set the mapping within one projects and have the config accessible in more of a utilities package great way to allow the team to do updates without us touching anything.

[ozoniuss]

In Python, do you also have to make sure about being able to read/write the configuration file at run time? In compiled languages like C++ Java or Go it's additionally bad practice to store configuration details in constants because you then have to recompile your code to change them

[mikec]

Great tip, short & helpful! Even better, even easier: YAML instead of JSON. YAML supports comments, wraps long strings, isn't fussy about commas and {}.

[benhbr]

Good advice, I would only not call the config parameters „data“

[coreman]

I think JSON lacks a lot compared to a proper config file. No comments and too many brackets. If your giving it to lots of users, theres a percentage that will be turned off by it. I think that another good way would be to user configparser which is in the python standard library. It allows for things like comments so it can be understood by not so technical users. I think that something like that would be a good video idea.

[horhay]

In a recent project I used a package called dacite to load a yaml file with dataclasses. Seems like something up your alley

[obed]

The module configparser is also pretty cool for this

[gigagaming]

EXACTELY what I needed. Thanks so much

[Manwithnoname]

I like the dotenv package for this, better for security also since keys and passwords should be outside of git repository

[przwar]

Hello Arjan. Maybe a little off-topic, but I think, it is a good practice to use contextmanager to open a file 🙂 Besides that, a good tip

[lawrencedoliveiro]

The next question is, _where_ do you put the config files? For personal programs, you could use the XDG convention of creating your own subdirectory in ~/.config/.

[sleao_]

what I'm using right now, specially because I deal mostly with enviroment variables, is pydantic "baseSettings", which let's me load configs directly from the env and gives back the dataclass we are used to. but yeah, json is nice too!

[dacid]

Another tip: sometimes you just want to be able to use the script with some sensible defaults without needing to deal with a config file. In that case, you can write your defaults at the top in a dict, and then call .update() on it with the JSON like shown here, if it exists. That also allows you to only need to put config in the file if it's different from the default.

[ayehavgunne]

What I like to do is define all my configs in a dataclass and use a library like dacite or apischema to hydrate the dataclass with the config values from a json file.