AI generated Pythonista

[mmccthomas]

Has anybody tried Google Gemini (free version) to create Pythonista code?
As simple as ‘create a Tetris game to run in Pythonista’

I’ve tried it to create classic games Pong, Tetris and Galaxians.
I’m staggered what a good job it does, mainly using the Scene module.
I might have expected vanilla Python code, but not this.

I’ve put a few in my repo https://github.com/mmccthomas/Pythonista_games/tree/main/AI_Games
with a few notes on modifications needed.

I know that I’m late to this AI party, but all I can say is WOW!
Chris

[cvpe]

I guess that your request did not explain the rules but only something like "create a Pythonista program to play this game". Describing a request has always been the most difficult part of a program, even if an human programmed on base of another human request..

[mmccthomas]

Yes, I agree, but I was quite impressed how VERY simple instructions

“Can you write space invader game in Pythonista?”
Followed by
“Make this galaxians”
Resulted in the following :

Okay, I can help you transform the Space Invaders code into a basic Galaxian-style game for Pythonista. This involves changing the invader behavior, adding invader bullets, and modifying the layout and movement.

Here is the updated Python code:
Galaxian for Pythonista - Expanded
May 19 at 9:43 PM

Key Changes and Galaxian Features Implemented:

Invader Types and Appearance: Invaders now have different colors and shapes based on their row (invader_type).
Invader Formation: Invaders are arranged in a formation with specific spacing.
Formation Movement: Invaders in formation move horizontally and descend when they hit the screen edge.
Diving Attacks: Invaders randomly peel off from the formation to dive towards the player's last known position.
Invader Bullets: Diving invaders can shoot bullets downwards.
Invader States: Invaders have states (formation, diving, returning) to manage their behavior. (Note: The returning state is basic and currently invaders that dive off-screen are just removed. A more complete version would have them return to formation).
Collision Detection: Added collision detection for invader bullets hitting the player.
Scoring: Scoring is now based on the invader type, with a potential bonus for hitting diving invaders.
Background: Added a simple starry background.
Font: Included an attempt to load a retro-style font ('Press Start 2P') for the score and game over text. You'll need to make sure you have a .ttf file for this font in your Pythonista file library.

For some other requests for more general subject such as “how best to create a dense crossword?’ It responded with a detailed report on strategies etc.
As I said, this crusty old engineer is impressed.
Chris

[mmccthomas]

Exporting code to Gemini AI and asking for a review yields extremely useful results.

Typical partial response:

• Constants vs. Magic Strings: Many methods use magic strings for button names (e.g., 'button_across', 'button_down'). Consider defining constants for these strings to prevent typos and make refactoring easier.
• Consistent Naming Conventions:
  • self.q: This variable is initialized to None but never seems to be used or assigned a Queue object. If it's vestigial, remove it. If it's intended for future use, give it a more descriptive name.
  • self.gs.Player vs self.player: It seems you have both self.gs.Player (which is set to a class, not an instance) and self.player (an instance). Clarify their roles or consolidate if one is redundant.
  • _panel and _itemlist in button_tapped: While these are internal, their names could be slightly more descriptive given they are reassigned based on context.

It seems that Gemini has a very good grasp of Pythonic code.
if I were still in a company, this would be my first port of call before a Pull Request and peer review.

I feel that this is also extremely useful for students learning Python, even if generation of code is not appropriate.