Improved Game State

[S1M0N38]

Related to #23

[Copilot]

Pull Request Overview

This PR significantly enhances the game state representation by improving data structure matching between Lua and Python, expanding game state information, and updating test files to align with the new schema. The changes support more comprehensive game state tracking as referenced in issue #23.

• Complete restructuring of game state models to match Lua types exactly
• Expanded game state extraction with detailed field documentation
• Updated test files to use new data structures and handle highlighted card differences

Reviewed Changes

Copilot reviewed 12 out of 13 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tests/runs/one_ante_no_shop_1.jsonl	Updated test data with expanded game state structure and detailed card information
tests/lua/test_runs.py	Added hack to handle highlighted card differences between logged and replayed states
tests/lua/test_api_functions.py	Updated to access cards through new nested structure (hand.cards vs hand)
tests/balatrobot/test_models.py	Changed import paths and model names (GameState → G)
tests/balatrobot/test_exceptions.py	Updated import paths for relocated modules
tests/balatrobot/test_client.py	Updated model references and import paths throughout
src/lua/utils.lua	Significantly expanded game state extraction with comprehensive field documentation
src/lua/types.lua	Complete rewrite of type definitions to match new game state structure
src/lua/api.lua	Added type annotations for better code clarity
src/balatrobot/models.py	Complete restructuring to match Lua types with hierarchical game state models
src/balatrobot/init.py	Updated exports (GameState → G)
docs/balatrobot-api.md	Updated documentation to reflect new model structure

[Copilot]

This hack indicates a design issue where logged state differs from replayed state. Consider logging the game state before card selection instead of after, or implement a proper normalization function rather than this inline hack.

Suggested change

	# HACK: Remove highlighted field from cards before comparison
	# The logger hook, log the game state after the card selection,
	# so in the replay we have the highlighted cards, while the game_state
	# before the action has the non-highlighted cards.
	if "hand" in actual_game_state and "cards" in actual_game_state["hand"]:
	for card in actual_game_state["hand"]["cards"]:
	card.pop("highlighted", None)
	if (
	"hand" in expected_game_state
	and "cards" in expected_game_state["hand"]
	):
	for card in expected_game_state["hand"]["cards"]:
	card.pop("highlighted", None)
	# Normalize game states before comparison
	actual_game_state = normalize_game_state(actual_game_state)
	expected_game_state = normalize_game_state(expected_game_state)

[Copilot]

Remove TODO comment as it appears to be leftover development notes and doesn't provide actionable information for the current implementation.

Suggested change

	-- TODO: continue from here (in the select blind state), maybe try to skip a blind
	--

[Copilot]

[nitpick] The logic for skipping capital letter keys is hardcoded and may be too broad. Consider using a more specific whitelist or blacklist approach to avoid accidentally filtering legitimate game state fields.

Suggested change

	-- Skip keys that start with capital letters (UI-related)
	-- Skip predefined fields to avoid circular references and large data
	if not (type(key) == "string" and string.sub(key, 1, 1):match("[A-Z]")) and not skip_fields[key] then
	-- Skip UI-related keys based on a specific blacklist
	-- Skip predefined fields to avoid circular references and large data
	local ui_skip_fields = {
	UIElement = true,
	UIButton = true,
	UILabel = true,
	UIPanel = true,
	}
	if not ui_skip_fields[key] and not skip_fields[key] then

[S1M0N38]

@stirby , so far I've implemented these new fields. G.jokers will be added later after other shop actions implementation so we can reach a game state with bought jolly programmatically.