Hey ChatGPT, you give me the definitive commands that will install a good version of python, create a venv, install the right packages and command line tools, setup the configuration just right and let me just code?
Cause for the code, codepilot is already fantastic.
Install a good version of Python:
# Install the latest version of Python 3
sudo apt update
sudo apt install python3
# Create a virtual environment for your project
python3 -m venv my_project
# Activate the virtual environment
source my_project/bin/activate
# Install the necessary packages and command line tools
pip install --upgrade pip
pip install setuptools wheel
pip install flake8 black isort
# Set up the configuration for your project
flake8 --install-hook git
black --install-hook git
isort --multi-line=3 --trailing-comma --force-grid-wrap=0 --combine-as --line-width 88 --recursive --thirdparty my_project
# Start coding in your project
cd my_project
nano my_project.py
To install Python, you can use the following command:
$ sudo apt-get install python3
$ python3 -m venv /path/to/venv
$ source /path/to/venv/bin/activate
$ pip install <package-name>
$ sudo apt-get install <tool-name>
Once you have set up your environment and installed the necessary packages and tools, you can start coding!
First line, you don't chose which Python you install, only the 3, so you don't know if the libs you need will be available, which syntax you can use and which stdlib is at your disposable.
Then, you use venv which may not be available because the first line you used doesn't install all python packages and in some debian/ubuntu versions they don't include venv, setuptools or pip in the python3 package.
Then you pip install packages, which may fail if this has a c extensions which is not a wheel (E.G: you do machine learning), because you didn't get the python headers in the previous commands.
Then you install command lines with apt-get, meaning your tool version may not be upgraded to match the python you used. For pylint and mypy, this means it at beast false positives, at worst crashes
Of course, none of that explains how to setup your IDE to use said virtual, setup things like databases, init the git repo or put good default in config files.
Don't get me wrong, it's a good start, and not worse than most tutorials I read online written by human.
I don't doubt that it will get better over time though, and I can't wait. Bootstrapping is a plague to teach people.
A lot of the things you complain about were underspecified in your question, it has to make some assumptions. And I'll bet if you ran into any of the errors that you describe you could get solutions by simply pasting the errors into the chat box. The code produced by these models is not going to be perfect any more than a human's would be. You'll still need an iterative process and some common sense. But it's easier and faster than doing everything yourself.
Most can't though, and they won't ask specific questions.
But with time, GPT will likely be able to ask questions to get the context, and that will change everything.
How is any being or AI meant to read your mind to know you don't want instructions for system X but assume the instructions would be good on system Y etc. I feel sorry for the humans you expect to query you to get what you really want out of you.
I gave up trying to install things locally on Mac. Sometimes I'm lucky with docker but that may be slow for trying new things out.
You're better off starting from an aws AMI and following whatever crazy setup the project you're using recommends.
Give me an environment where the test suite passes for this new repo I've cloned, based on all the environment config inside.
- Finds versions, sets them up
- Installs packages
- Asks about conflicts or problems
- Notes why the tests fail first time (no DB!)
- Creates the DB, runs migrations
- Notes why the tests fail next time (missing env variable)
- Finds a similar env variable in the example file, creates and local env file
- etc etc
That's all possible now if you paste the messages into GPT (with some errors, which repeated attempts can resolve), it needs connecting up into a self-aware environment.