feat: python store bindings (wip)

bot/Cargo.toml

@@ -1,5 +1,5 @@
 [package]
-name = "bot"
+name = "trictrac-bot"
 version = "0.1.0"
 edition = "2021"
 
@@ -13,10 +13,10 @@ path = "src/burnrl/main.rs"
 pretty_assertions = "1.4.0"
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"
-store = { path = "../store" }
+trictrac-store = { path = "../store" }
-rand = "0.8"
+rand = "0.9"
 env_logger = "0.10"
-burn = { version = "0.18", features = ["ndarray", "autodiff"] }
+burn = { version = "0.20", features = ["ndarray", "autodiff"] }
 burn-rl = { git = "https://github.com/yunjhongwu/burn-rl-examples.git", package = "burn-rl" }
 log = "0.4.20"
 confy = "1.0.0"

bot/python/test.py

@@ -0,0 +1,5 @@
+import trictrac_store
+
+game = trictrac_store.TricTrac()
+print(game.current_player_idx())
+print(game.get_legal_actions(game.current_player_idx()))

bot/src/burnrl/environment.rs

@@ -1,10 +1,10 @@
 use std::io::Write;
 
-use crate::training_common;
 use burn::{prelude::Backend, tensor::Tensor};
 use burn_rl::base::{Action, Environment, Snapshot, State};
-use rand::{thread_rng, Rng};
+use rand::{rng, Rng};
-use store::{GameEvent, GameState, PlayerId, PointsRules, Stage, TurnStage};
+use trictrac_store::training_common;
+use trictrac_store::{GameEvent, GameState, PlayerId, PointsRules, Stage, TurnStage};
 
 const ERROR_REWARD: f32 = -1.0012121;
 const REWARD_VALID_MOVE: f32 = 1.0012121;
@@ -52,10 +52,10 @@ pub struct TrictracAction {
 
 impl Action for TrictracAction {
     fn random() -> Self {
-        use rand::{thread_rng, Rng};
+        use rand::{rng, Rng};
-        let mut rng = thread_rng();
+        let mut rng = rng();
         TrictracAction {
-            index: rng.gen_range(0..Self::size() as u32),
+            index: rng.random_range(0..Self::size() as u32),
         }
     }
 
@@ -288,11 +288,11 @@ impl TrictracEnvironment {
                 // reward += REWARD_VALID_MOVE;
                 // Simuler le résultat des dés après un Roll
                 if matches!(action, TrictracAction::Roll) {
-                    let mut rng = thread_rng();
+                    let mut rng = rng();
-                    let dice_values = (rng.gen_range(1..=6), rng.gen_range(1..=6));
+                    let dice_values = (rng.random_range(1..=6), rng.random_range(1..=6));
                     let dice_event = GameEvent::RollResult {
                         player_id: self.active_player_id,
-                        dice: store::Dice {
+                        dice: trictrac_store::Dice {
                             values: dice_values,
                         },
                     };
@@ -340,18 +340,18 @@ impl TrictracEnvironment {
 
             // Exécuter l'action selon le turn_stage
             let mut calculate_points = false;
-            let opponent_color = store::Color::Black;
+            let opponent_color = trictrac_store::Color::Black;
             let event = match self.game.turn_stage {
                 TurnStage::RollDice => GameEvent::Roll {
                     player_id: self.opponent_id,
                 },
                 TurnStage::RollWaiting => {
-                    let mut rng = thread_rng();
+                    let mut rng = rng();
-                    let dice_values = (rng.gen_range(1..=6), rng.gen_range(1..=6));
+                    let dice_values = (rng.random_range(1..=6), rng.random_range(1..=6));
                     calculate_points = true;
                     GameEvent::RollResult {
                         player_id: self.opponent_id,
-                        dice: store::Dice {
+                        dice: trictrac_store::Dice {
                             values: dice_values,
                         },
                     }
@@ -371,7 +371,7 @@ impl TrictracEnvironment {
                     }
                 }
                 TurnStage::MarkAdvPoints => {
-                    let opponent_color = store::Color::Black;
+                    let opponent_color = trictrac_store::Color::Black;
                     let dice_roll_count = self
                         .game
                         .players

bot/src/burnrl/environment_valid.rs

@@ -1,8 +1,8 @@
-use crate::training_common;
 use burn::{prelude::Backend, tensor::Tensor};
 use burn_rl::base::{Action, Environment, Snapshot, State};
-use rand::{thread_rng, Rng};
+use rand::{rng, Rng};
-use store::{GameEvent, GameState, PlayerId, PointsRules, Stage, TurnStage};
+use trictrac_store::training_common;
+use trictrac_store::{GameEvent, GameState, PlayerId, PointsRules, Stage, TurnStage};
 
 const ERROR_REWARD: f32 = -1.0012121;
 const REWARD_RATIO: f32 = 0.1;
@@ -48,10 +48,10 @@ pub struct TrictracAction {
 
 impl Action for TrictracAction {
     fn random() -> Self {
-        use rand::{thread_rng, Rng};
+        use rand::{rng, Rng};
-        let mut rng = thread_rng();
+        let mut rng = rng();
         TrictracAction {
-            index: rng.gen_range(0..Self::size() as u32),
+            index: rng.random_range(0..Self::size() as u32),
         }
     }
 
@@ -237,7 +237,7 @@ impl TrictracEnvironment {
 
         // Mapper l'index d'action sur une action valide
         let action_index = (action.index as usize) % valid_actions.len();
-        Some(valid_actions[action_index].clone())
+        Some(valid_actions[action_index])
     }
 
     /// Exécute une action Trictrac dans le jeu
@@ -258,11 +258,11 @@ impl TrictracEnvironment {
                 // reward += REWARD_VALID_MOVE;
                 // Simuler le résultat des dés après un Roll
                 if matches!(action, TrictracAction::Roll) {
-                    let mut rng = thread_rng();
+                    let mut rng = rng();
-                    let dice_values = (rng.gen_range(1..=6), rng.gen_range(1..=6));
+                    let dice_values = (rng.random_range(1..=6), rng.random_range(1..=6));
                     let dice_event = GameEvent::RollResult {
                         player_id: self.active_player_id,
-                        dice: store::Dice {
+                        dice: trictrac_store::Dice {
                             values: dice_values,
                         },
                     };
@@ -310,18 +310,18 @@ impl TrictracEnvironment {
 
             // Exécuter l'action selon le turn_stage
             let mut calculate_points = false;
-            let opponent_color = store::Color::Black;
+            let opponent_color = trictrac_store::Color::Black;
             let event = match self.game.turn_stage {
                 TurnStage::RollDice => GameEvent::Roll {
                     player_id: self.opponent_id,
                 },
                 TurnStage::RollWaiting => {
-                    let mut rng = thread_rng();
+                    let mut rng = rng();
-                    let dice_values = (rng.gen_range(1..=6), rng.gen_range(1..=6));
+                    let dice_values = (rng.random_range(1..=6), rng.random_range(1..=6));
                     calculate_points = true;
                     GameEvent::RollResult {
                         player_id: self.opponent_id,
-                        dice: store::Dice {
+                        dice: trictrac_store::Dice {
                             values: dice_values,
                         },
                     }

bot/src/burnrl/main.rs

@@ -1,7 +1,7 @@
-use bot::burnrl::algos::{dqn, dqn_valid, ppo, ppo_valid, sac, sac_valid};
+use trictrac_bot::burnrl::algos::{dqn, dqn_valid, ppo, ppo_valid, sac, sac_valid};
-use bot::burnrl::environment::TrictracEnvironment;
+use trictrac_bot::burnrl::environment::TrictracEnvironment;
-use bot::burnrl::environment_valid::TrictracEnvironment as TrictracEnvironmentValid;
+use trictrac_bot::burnrl::environment_valid::TrictracEnvironment as TrictracEnvironmentValid;
-use bot::burnrl::utils::{demo_model, Config};
+use trictrac_bot::burnrl::utils::{demo_model, Config};
 use burn::backend::{Autodiff, NdArray};
 use burn_rl::base::ElemType;
 use std::env;

bot/src/lib.rs

@@ -1,15 +1,16 @@
 pub mod burnrl;
 pub mod strategy;
-pub mod training_common;
 pub mod trictrac_board;
 
 use log::debug;
-use store::{CheckerMove, Color, GameEvent, GameState, PlayerId, PointsRules, Stage, TurnStage};
 pub use strategy::default::DefaultStrategy;
 pub use strategy::dqnburn::DqnBurnStrategy;
 pub use strategy::erroneous_moves::ErroneousStrategy;
 pub use strategy::random::RandomStrategy;
 pub use strategy::stable_baselines3::StableBaselines3Strategy;
+use trictrac_store::{
+    CheckerMove, Color, GameEvent, GameState, PlayerId, PointsRules, Stage, TurnStage,
+};
 
 pub trait BotStrategy: std::fmt::Debug {
     fn get_game(&self) -> &GameState;
@@ -144,7 +145,7 @@ impl Bot {
 #[cfg(test)]
 mod tests {
     use super::*;
-    use store::{Dice, Stage};
+    use trictrac_store::{Dice, Stage};
 
     #[test]
     fn test_new() {

bot/src/strategy/default.rs

@@ -1,5 +1,5 @@
 use crate::{BotStrategy, CheckerMove, Color, GameState, PlayerId};
-use store::MoveRules;
+use trictrac_store::MoveRules;
 
 #[derive(Debug)]
 pub struct DefaultStrategy {

bot/src/strategy/dqnburn.rs

@@ -4,11 +4,13 @@ use burn_rl::base::{ElemType, Model, State};
 
 use crate::{BotStrategy, CheckerMove, Color, GameState, PlayerId};
 use log::info;
-use store::MoveRules;
+use trictrac_store::MoveRules;
 
 use crate::burnrl::algos::dqn;
 use crate::burnrl::environment;
-use crate::training_common::{get_valid_action_indices, sample_valid_action, TrictracAction};
+use trictrac_store::training_common::{
+    get_valid_action_indices, sample_valid_action, TrictracAction,
+};
 
 type DqnBurnNetwork = dqn::Net<NdArray<ElemType>>;
 
@@ -152,7 +154,7 @@ impl BotStrategy for DqnBurnStrategy {
                 to1 = if fto1 < 0 { 0 } else { fto1 as usize };
             }
 
-            let checker_move1 = store::CheckerMove::new(from1, to1).unwrap_or_default();
+            let checker_move1 = trictrac_store::CheckerMove::new(from1, to1).unwrap_or_default();
 
             let mut tmp_board = self.game.board.clone();
             let move_res = tmp_board.move_checker(&self.color, checker_move1);

bot/src/strategy/random.rs

@@ -1,5 +1,6 @@
 use crate::{BotStrategy, CheckerMove, Color, GameState, PlayerId};
-use store::MoveRules;
+use rand::{prelude::IndexedRandom, rng};
+use trictrac_store::MoveRules;
 
 #[derive(Debug)]
 pub struct RandomStrategy {
@@ -51,8 +52,7 @@ impl BotStrategy for RandomStrategy {
         let rules = MoveRules::new(&self.color, &self.game.board, self.game.dice);
         let possible_moves = rules.get_possible_moves_sequences(true, vec![]);
 
-        use rand::{seq::SliceRandom, thread_rng};
+        let mut rng = rng();
-        let mut rng = thread_rng();
         let choosen_move = possible_moves
             .choose(&mut rng)
             .cloned()

bot/src/strategy/stable_baselines3.rs

@@ -5,7 +5,7 @@ use std::io::Read;
 use std::io::Write;
 use std::path::Path;
 use std::process::Command;
-use store::MoveRules;
+use trictrac_store::MoveRules;
 
 #[derive(Debug)]
 pub struct StableBaselines3Strategy {
@@ -79,12 +79,12 @@ impl StableBaselines3Strategy {
 
         // Convertir l'étape du tour en entier
         let turn_stage = match self.game.turn_stage {
-            store::TurnStage::RollDice => 0,
+            trictrac_store::TurnStage::RollDice => 0,
-            store::TurnStage::RollWaiting => 1,
+            trictrac_store::TurnStage::RollWaiting => 1,
-            store::TurnStage::MarkPoints => 2,
+            trictrac_store::TurnStage::MarkPoints => 2,
-            store::TurnStage::HoldOrGoChoice => 3,
+            trictrac_store::TurnStage::HoldOrGoChoice => 3,
-            store::TurnStage::Move => 4,
+            trictrac_store::TurnStage::Move => 4,
-            store::TurnStage::MarkAdvPoints => 5,
+            trictrac_store::TurnStage::MarkAdvPoints => 5,
         };
 
         // Récupérer les points et trous des joueurs

bot/src/trictrac_board.rs

@@ -1,5 +1,4 @@
 // https://docs.rs/board-game/ implementation
-use crate::training_common::{get_valid_actions, TrictracAction};
 use board_game::board::{
     Board as BoardGameBoard, BoardDone, BoardMoves, Outcome, PlayError, Player as BoardGamePlayer,
 };
@@ -8,7 +7,8 @@ use internal_iterator::InternalIterator;
 use std::fmt;
 use std::hash::Hash;
 use std::ops::ControlFlow;
-use store::Color;
+use trictrac_store::training_common::{get_valid_actions, TrictracAction};
+use trictrac_store::Color;
 
 #[derive(Clone, Debug, Eq, PartialEq, Hash)]
 pub struct TrictracBoard(crate::GameState);

client_cli/Cargo.toml

@@ -1,5 +1,5 @@
 [package]
-name = "client_cli"
+name = "trictrac-client_cli"
 version = "0.1.0"
 edition = "2021"
 
@@ -11,8 +11,8 @@ bincode = "1.3.3"
 pico-args = "0.5.0"
 pretty_assertions = "1.4.0"
 renet = "0.0.13"
-store = { path = "../store" }
+trictrac-store = { path = "../store" }
-bot = { path = "../bot" }
+trictrac-bot = { path = "../bot" }
 itertools = "0.13.0"
 env_logger = "0.11.6"
 log = "0.4.20"

client_cli/src/app.rs

@@ -1,11 +1,11 @@
-use bot::{
+use trictrac_bot::{
     BotStrategy, DefaultStrategy, DqnBurnStrategy, ErroneousStrategy, RandomStrategy,
     StableBaselines3Strategy,
 };
 use itertools::Itertools;
 
 use crate::game_runner::GameRunner;
-use store::{CheckerMove, GameEvent, GameState, Stage, TurnStage};
+use trictrac_store::{CheckerMove, GameEvent, GameState, Stage, TurnStage};
 
 #[derive(Debug, Default)]
 pub struct AppArgs {

client_cli/src/game_runner.rs

@@ -1,6 +1,6 @@
-use bot::{Bot, BotStrategy};
+use trictrac_bot::{Bot, BotStrategy};
 use log::{debug, error};
-use store::{CheckerMove, DiceRoller, GameEvent, GameState, PlayerId, TurnStage};
+use trictrac_store::{CheckerMove, DiceRoller, GameEvent, GameState, PlayerId, TurnStage};
 
 // Application Game
 #[derive(Debug, Default)]
@@ -117,8 +117,8 @@ impl GameRunner {
         }
 
         if let Some(winner) = self.state.determine_winner() {
-            next_event = Some(store::GameEvent::EndGame {
+            next_event = Some(trictrac_store::GameEvent::EndGame {
-                reason: store::EndGameReason::PlayerWon { winner },
+                reason: trictrac_store::EndGameReason::PlayerWon { winner },
             });
         }
 

devenv.lock

@@ -3,10 +3,10 @@
     "devenv": {
       "locked": {
         "dir": "src/modules",
-        "lastModified": 1753667201,
+        "lastModified": 1770390537,
         "owner": "cachix",
         "repo": "devenv",
-        "rev": "4d584d7686a50387f975879788043e55af9f0ad4",
+        "rev": "d6f45cc00829254a9a6f8807c8fbfaf3efa7e629",
         "type": "github"
       },
       "original": {
@@ -19,14 +19,14 @@
     "flake-compat": {
       "flake": false,
       "locked": {
-        "lastModified": 1747046372,
+        "lastModified": 1767039857,
-        "owner": "edolstra",
+        "owner": "NixOS",
         "repo": "flake-compat",
-        "rev": "9100a0f413b0c601e0533d1d94ffd501ce2e7885",
+        "rev": "5edf11c44bc78a0d334f6334cdaf7d60d732daab",
         "type": "github"
       },
       "original": {
-        "owner": "edolstra",
+        "owner": "NixOS",
         "repo": "flake-compat",
         "type": "github"
       }
@@ -40,10 +40,10 @@
         ]
       },
       "locked": {
-        "lastModified": 1750779888,
+        "lastModified": 1769939035,
         "owner": "cachix",
         "repo": "git-hooks.nix",
-        "rev": "16ec914f6fb6f599ce988427d9d94efddf25fe6d",
+        "rev": "a8ca480175326551d6c4121498316261cbb5b260",
         "type": "github"
       },
       "original": {
@@ -60,10 +60,10 @@
         ]
       },
       "locked": {
-        "lastModified": 1709087332,
+        "lastModified": 1762808025,
         "owner": "hercules-ci",
         "repo": "gitignore.nix",
-        "rev": "637db329424fd7e46cf4185293b9cc8c88c95394",
+        "rev": "cb5e3fdca1de58ccbc3ef53de65bd372b48f567c",
         "type": "github"
       },
       "original": {
@@ -74,24 +74,40 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1753432016,
+        "lastModified": 1770136044,
         "owner": "NixOS",
         "repo": "nixpkgs",
-        "rev": "6027c30c8e9810896b92429f0092f624f7b1aace",
+        "rev": "e576e3c9cf9bad747afcddd9e34f51d18c855b4e",
         "type": "github"
       },
       "original": {
         "owner": "NixOS",
-        "ref": "nixpkgs-unstable",
+        "ref": "nixos-25.11",
         "repo": "nixpkgs",
         "type": "github"
       }
     },
+    "nixpkgs-cmake3": {
+      "locked": {
+        "lastModified": 1758213207,
+        "owner": "NixOS",
+        "repo": "nixpkgs",
+        "rev": "f4b140d5b253f5e2a1ff4e5506edbf8267724bde",
+        "type": "github"
+      },
+      "original": {
+        "owner": "NixOS",
+        "repo": "nixpkgs",
+        "rev": "f4b140d5b253f5e2a1ff4e5506edbf8267724bde",
+        "type": "github"
+      }
+    },
     "root": {
       "inputs": {
         "devenv": "devenv",
         "git-hooks": "git-hooks",
         "nixpkgs": "nixpkgs",
+        "nixpkgs-cmake3": "nixpkgs-cmake3",
         "pre-commit-hooks": [
           "git-hooks"
         ]

devenv.nix

@@ -1,13 +1,16 @@
-{ pkgs, ... }:
+{ inputs, pkgs, ... }:
 
+let
+  pkgs-cmake3 = import inputs.nixpkgs-cmake3 { system = pkgs.stdenv.system; };
+in
 {
-
   packages = [
 
     # pour burn-rs
     pkgs.SDL2_gfx
     #  (compilation sdl2-sys)
-    pkgs.cmake
+    pkgs-cmake3.cmake
+    pkgs.libxcb
     pkgs.libffi
     pkgs.wayland-scanner
 
@@ -15,6 +18,12 @@
     pkgs.samply # code profiler
     pkgs.feedgnuplot # to visualize bots training results
 
+    # --- AI training with python ---
+    # generate python classes from rust code
+    pkgs.maturin
+    # required by python numpy
+    pkgs.libz
+
     # for bevy
     pkgs.alsa-lib
     pkgs.udev
@@ -47,6 +56,25 @@
   # https://devenv.sh/languages/
   languages.rust.enable = true;
 
+
+  # AI training with python
+  enterShell = ''
+    PYTHONPATH=$PYTHONPATH:$PWD/.devenv/state/venv/lib/python3/site-packages
+  '';
+
+  languages.python = {
+    enable = true;
+    uv.enable = true;
+    venv.enable = true;
+    venv.requirements = "
+      pip
+      gymnasium
+      numpy
+      stable-baselines3
+      shimmy
+    ";
+  };
+
   # https://devenv.sh/scripts/
   # scripts.hello.exec = "echo hello from $GREET";
 

devenv.yaml

@@ -1,3 +1,5 @@
 inputs:
   nixpkgs:
-    url: github:NixOS/nixpkgs/nixpkgs-unstable
+    url: github:NixOS/nixpkgs/nixos-25.11
+  nixpkgs-cmake3:
+    url: github:NixOS/nixpkgs/f4b140d5b253f5e2a1ff4e5506edbf8267724bde

doc/backlog.md

@@ -53,6 +53,10 @@ Client
 
 ### Epic : Bot
 
+- PGX
+  - https://joe-antognini.github.io/ml/jax-tic-tac-toe
+  - https://www.sotets.uk/pgx/api_usage/
+
 - OpenAi gym
   - doc gymnasium <https://gymnasium.farama.org/introduction/basic_usage/>
   - Rust implementation for OpenAi gym <https://github.com/MathisWellmann/gym-rs>

doc/python.md

@@ -0,0 +1,31 @@
+# Python bindings
+
+## Génération bindings
+
+```sh
+# Generate trictrac python lib as a wheel
+maturin build -m store/Cargo.toml --release
+# Install wheel in local python env
+pip install --no-deps --force-reinstall --prefix .devenv/state/venv target/wheels/*.whl
+```
+
+## Usage
+
+Pour vérifier l'accès à la lib : lancer le shell interactif `python`
+
+```python
+Python 3.13.11 (main, Dec  5 2025, 16:06:33) [GCC 15.2.0] on linux
+Type "help", "copyright", "credits" or "license" for more information.
+>>> import trictrac_store
+>>> game = trictrac_store.TricTrac()
+>>> game.get_active_player_id()
+1
+```
+
+### Appels depuis python
+
+`python bot/python/test.py`
+
+## Interfaces
+
+## Entraînement

justfile

@@ -20,6 +20,7 @@ profile:
   cargo build --profile profiling
   samply record ./target/profiling/client_cli --bot dummy,dummy
 pythonlib:
+  rm -rf target/wheels
   maturin build -m store/Cargo.toml --release
   pip install --no-deps --force-reinstall --prefix .devenv/state/venv target/wheels/*.whl
 trainbot algo:

store/Cargo.toml

@@ -1,20 +1,23 @@
 [package]
-name = "store"
+name = "trictrac-store"
 version = "0.1.0"
 edition = "2021"
 
 # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
 
 [lib]
-name = "store"
+name = "trictrac_store"
+# "cdylib" is necessary to produce a shared library for Python to import from.
 # Only "rlib" is needed for other Rust crates to use this library
-crate-type = ["rlib"]
+crate-type = ["cdylib", "rlib"]
 
 [dependencies]
 base64 = "0.21.7"
 # provides macros for creating log messages to be used by a logger (for example env_logger)
 log = "0.4.20"
 merge = "0.1.0"
-rand = "0.8.5"
+# generate python lib (with maturin) to be used in AI training
+pyo3 = { version = "0.23", features = ["extension-module", "abi3-py38"] }
+rand = "0.9"
 serde = { version = "1.0", features = ["derive"] }
 transpose = "0.2.2"

store/pyproject.toml

@@ -0,0 +1,8 @@
+[build-system]
+requires = ["maturin>=1.0,<2.0"]
+build-backend = "maturin"
+
+[tool.maturin]
+# "extension-module" tells pyo3 we want to build an extension module (skips linking against libpython.so)
+features = ["pyo3/extension-module"]
+# python-source = "python"

store/src/dice.rs

@@ -1,4 +1,4 @@
-use rand::distributions::{Distribution, Uniform};
+use rand::distr::{Distribution, Uniform};
 use rand::{rngs::StdRng, SeedableRng};
 use serde::{Deserialize, Serialize};
 
@@ -17,7 +17,7 @@ impl DiceRoller {
     pub fn new(opt_seed: Option<u64>) -> Self {
         Self {
             rng: match opt_seed {
-                None => StdRng::from_rng(rand::thread_rng()).unwrap(),
+                None => StdRng::from_rng(&mut rand::rng()),
                 Some(seed) => SeedableRng::seed_from_u64(seed),
             },
         }
@@ -26,7 +26,7 @@ impl DiceRoller {
     /// Roll the dices which generates two random numbers between 1 and 6, replicating a perfect
     /// dice. We use the operating system's random number generator.
     pub fn roll(&mut self) -> Dice {
-        let between = Uniform::new_inclusive(1, 6);
+        let between = Uniform::new_inclusive(1, 6).expect("1 > 6 !?");
 
         let v = (between.sample(&mut self.rng), between.sample(&mut self.rng));
 

store/src/lib.rs

@@ -16,3 +16,8 @@ pub use board::CheckerMove;
 
 mod dice;
 pub use dice::{Dice, DiceRoller};
+
+pub mod training_common;
+
+// python interface "trictrac_engine" (for AI training..)
+mod pyengine;

store/src/player.rs

@@ -1,9 +1,11 @@
+use pyo3::prelude::*;
 use serde::{Deserialize, Serialize};
 use std::fmt;
 
 // This just makes it easier to dissern between a player id and any ol' u64
 pub type PlayerId = u64;
 
+#[pyclass(eq, eq_int)]
 #[derive(Copy, Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub enum Color {
     White,

store/src/pyengine.rs

@@ -0,0 +1,218 @@
+//! # Expose trictrac game state and rules in a python module
+use pyo3::prelude::*;
+use pyo3::types::PyDict;
+
+use crate::board::CheckerMove;
+use crate::dice::{Dice, DiceRoller};
+use crate::game::{GameEvent, GameState, Stage, TurnStage};
+use crate::game_rules_moves::MoveRules;
+use crate::game_rules_points::PointsRules;
+use crate::player::{Color, PlayerId};
+use crate::training_common::{get_valid_action_indices, TrictracAction};
+
+#[pyclass]
+struct TricTrac {
+    game_state: GameState,
+    dice_roll_sequence: Vec<(u8, u8)>,
+    current_dice_index: usize,
+}
+
+#[pymethods]
+impl TricTrac {
+    #[new]
+    fn new() -> Self {
+        let mut game_state = GameState::new(false); // schools_enabled = false
+
+        // Initialiser 2 joueurs
+        game_state.init_player("player1");
+        game_state.init_player("player2");
+
+        // Commencer la partie avec le joueur 1
+        game_state.consume(&GameEvent::BeginGame { goes_first: 1 });
+
+        TricTrac {
+            game_state,
+            dice_roll_sequence: Vec::new(),
+            current_dice_index: 0,
+        }
+    }
+
+    fn needs_roll(&self) -> bool {
+        self.game_state.turn_stage == TurnStage::RollWaiting
+    }
+
+    fn is_game_ended(&self) -> bool {
+        self.game_state.stage == Stage::Ended
+    }
+
+    // 0 or 1
+    fn current_player_idx(&self) -> u64 {
+        self.game_state.active_player_id - 1
+    }
+
+    fn get_legal_actions(&self, player_id: u64) -> Vec<usize> {
+        if player_id == self.current_player_idx() {
+            get_valid_action_indices(&self.game_state)
+        } else {
+            vec![]
+        }
+    }
+
+    fn action_to_string(&self, player_idx: u64, action_idx: usize) -> String {
+        TrictracAction::from_action_index(action_idx)
+            .map(|a| a.to_string())
+            .unwrap_or("unknown action".into())
+    }
+
+    fn apply_dice_roll(&mut self, dices: (u8, u8)) -> PyResult<()> {
+        let player_id = self.game_state.active_player_id;
+
+        if self.game_state.turn_stage != TurnStage::RollDice {
+            return Err(pyo3::exceptions::PyRuntimeError::new_err(
+                "Not in RollDice stage",
+            ));
+        }
+
+        self.game_state.consume(&GameEvent::Roll { player_id });
+        let dice = Dice { values: dices };
+        self.game_state
+            .consume(&GameEvent::RollResult { player_id, dice });
+        Ok(())
+    }
+
+    fn apply_action(&mut self, action_idx: usize) {
+        if let Some(event) =
+            TrictracAction::from_action_index(action_idx).and_then(|a| a.to_event(&self.game_state))
+        {
+            if self.game_state.validate(&event) {
+                self.game_state.consume(&event);
+                // return Ok(());
+            }
+        }
+        // Err(pyo3::exceptions::PyRuntimeError::new_err(
+        //     "Could not apply action",
+        // ))
+    }
+
+    /// Obtenir l'état du jeu sous forme de chaîne de caractères compacte
+    fn get_state_id(&self) -> String {
+        self.game_state.to_string_id()
+    }
+
+    /// Renvoie les positions des pièces pour un joueur spécifique
+    fn get_checker_positions(&self, color: Color) -> Vec<(usize, i8)> {
+        self.game_state.board.get_color_fields(color)
+    }
+
+    /// Obtenir la liste des mouvements légaux sous forme de paires (from, to)
+    fn get_available_moves(&self) -> Vec<((usize, usize), (usize, usize))> {
+        // L'agent joue toujours le joueur actif
+        let color = self
+            .game_state
+            .player_color_by_id(&self.game_state.active_player_id)
+            .unwrap_or(Color::White);
+
+        // Si ce n'est pas le moment de déplacer les pièces, retourner une liste vide
+        if self.game_state.turn_stage != TurnStage::Move
+            && self.game_state.turn_stage != TurnStage::HoldOrGoChoice
+        {
+            return vec![];
+        }
+
+        let rules = MoveRules::new(&color, &self.game_state.board, self.game_state.dice);
+        let possible_moves = rules.get_possible_moves_sequences(true, vec![]);
+
+        // Convertir les mouvements CheckerMove en tuples (from, to) pour Python
+        possible_moves
+            .into_iter()
+            .map(|(move1, move2)| {
+                (
+                    (move1.get_from(), move1.get_to()),
+                    (move2.get_from(), move2.get_to()),
+                )
+            })
+            .collect()
+    }
+
+    /// Calcule les points maximaux que le joueur actif peut obtenir avec les dés actuels
+    fn calculate_points(&self) -> u8 {
+        let active_player = self
+            .game_state
+            .players
+            .get(&self.game_state.active_player_id);
+
+        if let Some(player) = active_player {
+            let dice_roll_count = player.dice_roll_count;
+            let color = player.color;
+
+            let points_rules =
+                PointsRules::new(&color, &self.game_state.board, self.game_state.dice);
+            let (points, _) = points_rules.get_points(dice_roll_count);
+
+            points
+        } else {
+            0
+        }
+    }
+
+    /// Réinitialise la partie
+    fn reset(&mut self) {
+        self.game_state = GameState::new(false);
+
+        // Initialiser 2 joueurs
+        self.game_state.init_player("player1");
+        self.game_state.init_player("player2");
+
+        // Commencer la partie avec le joueur 1
+        self.game_state
+            .consume(&GameEvent::BeginGame { goes_first: 1 });
+
+        // Réinitialiser l'index de la séquence de dés
+        self.current_dice_index = 0;
+    }
+
+    /// Vérifie si la partie est terminée
+    fn is_done(&self) -> bool {
+        self.game_state.stage == Stage::Ended || self.game_state.determine_winner().is_some()
+    }
+
+    /// Obtenir le gagnant de la partie
+    fn get_winner(&self) -> Option<PlayerId> {
+        self.game_state.determine_winner()
+    }
+
+    /// Obtenir le score du joueur actif (nombre de trous)
+    fn get_score(&self, player_id: PlayerId) -> i32 {
+        if let Some(player) = self.game_state.players.get(&player_id) {
+            player.holes as i32
+        } else {
+            -1
+        }
+    }
+
+    /// Obtenir l'ID du joueur actif
+    fn get_active_player_id(&self) -> PlayerId {
+        self.game_state.active_player_id
+    }
+
+    /// Définir une séquence de dés à utiliser (pour la reproductibilité)
+    fn set_dice_sequence(&mut self, sequence: Vec<(u8, u8)>) {
+        self.dice_roll_sequence = sequence;
+        self.current_dice_index = 0;
+    }
+
+    /// Afficher l'état du jeu (pour le débogage)
+    fn __str__(&self) -> String {
+        format!("{}", self.game_state)
+    }
+}
+
+/// A Python module implemented in Rust. The name of this function must match
+/// the `lib.name` setting in the `Cargo.toml`, else Python will not be able to
+/// import the module.
+#[pymodule]
+fn trictrac_store(m: &Bound<'_, PyModule>) -> PyResult<()> {
+    m.add_class::<TricTrac>()?;
+
+    Ok(())
+}

store/src/training_common.rs

@@ -3,11 +3,11 @@
 use std::cmp::{max, min};
 use std::fmt::{Debug, Display, Formatter};
 
+use crate::{CheckerMove, GameEvent, GameState};
 use serde::{Deserialize, Serialize};
-use store::{CheckerMove, GameEvent, GameState};
 
-// 1 (Roll) + 1 (Go) + mouvements possibles
+// 1 (Roll) + 1 (Go) + 512 (mouvements possibles)
-// Pour les mouvements : 2*16*16 = 514 (choix du dé + choix de la dame 0-15 pour chaque from)
+// avec 512 = 2 (choix du dé) * 16 * 16 (choix de la dame 0-15 pour chaque from)
 pub const ACTION_SPACE_SIZE: usize = 514;
 
 /// Types d'actions possibles dans le jeu
@@ -15,7 +15,8 @@ pub const ACTION_SPACE_SIZE: usize = 514;
 pub enum TrictracAction {
     /// Lancer les dés
     Roll,
-    /// Continuer après avoir gagné un trou
+    /// Faire un nouveau 'relevé' (repositionnement des dames à l'état de départ) après avoir gagné un trou,
+    /// au lieu de continuer dans la position courante
     Go,
     /// Effectuer un mouvement de pions
     Move {
@@ -93,13 +94,13 @@ impl TrictracAction {
                     (state.dice.values.1, state.dice.values.0)
                 };
 
-                let color = &store::Color::White;
+                let color = &crate::Color::White;
                 let from1 = state
                     .board
                     .get_checker_field(color, *checker1 as u8)
                     .unwrap_or(0);
                 let mut to1 = from1 + dice1 as usize;
-                let checker_move1 = store::CheckerMove::new(from1, to1).unwrap_or_default();
+                let checker_move1 = CheckerMove::new(from1, to1).unwrap_or_default();
 
                 let mut tmp_board = state.board.clone();
                 let move_result = tmp_board.move_checker(color, checker_move1);
@@ -119,8 +120,8 @@ impl TrictracAction {
                         to2 -= 1;
                     }
 
-                    let checker_move1 = store::CheckerMove::new(from1, to1).unwrap_or_default();
+                    let checker_move1 = CheckerMove::new(from1, to1).unwrap_or_default();
-                    let checker_move2 = store::CheckerMove::new(from2, to2).unwrap_or_default();
+                    let checker_move2 = CheckerMove::new(from2, to2).unwrap_or_default();
 
                     Some(GameEvent::Move {
                         player_id: state.active_player_id,
@@ -166,33 +167,11 @@ impl TrictracAction {
     pub fn action_space_size() -> usize {
         ACTION_SPACE_SIZE
     }
-
-    // pub fn to_game_event(&self, player_id: PlayerId, dice: Dice) -> GameEvent {
-    //     match action {
-    //         TrictracAction::Roll => Some(GameEvent::Roll { player_id }),
-    //         TrictracAction::Mark => Some(GameEvent::Mark { player_id, points }),
-    //         TrictracAction::Go => Some(GameEvent::Go { player_id }),
-    //         TrictracAction::Move {
-    //             dice_order,
-    //             from1,
-    //             from2,
-    //         } => {
-    //             // Effectuer un mouvement
-    //             let checker_move1 = store::CheckerMove::new(move1.0, move1.1).unwrap_or_default();
-    //             let checker_move2 = store::CheckerMove::new(move2.0, move2.1).unwrap_or_default();
-    //
-    //             Some(GameEvent::Move {
-    //                 player_id: self.agent_player_id,
-    //                 moves: (checker_move1, checker_move2),
-    //             })
-    //         }
-    //     };
-    // }
 }
 
 /// Obtient les actions valides pour l'état de jeu actuel
-pub fn get_valid_actions(game_state: &crate::GameState) -> Vec<TrictracAction> {
+pub fn get_valid_actions(game_state: &GameState) -> Vec<TrictracAction> {
-    use store::TurnStage;
+    use crate::TurnStage;
 
     let mut valid_actions = Vec::new();
 
@@ -215,11 +194,11 @@ pub fn get_valid_actions(game_state: &crate::GameState) -> Vec<TrictracAction> {
                 valid_actions.push(TrictracAction::Go);
 
                 // Ajoute aussi les mouvements possibles
-                let rules = store::MoveRules::new(&color, &game_state.board, game_state.dice);
+                let rules = crate::MoveRules::new(&color, &game_state.board, game_state.dice);
                 let possible_moves = rules.get_possible_moves_sequences(true, vec![]);
 
                 // Modififier checker_moves_to_trictrac_action si on doit gérer Black
-                assert_eq!(color, store::Color::White);
+                assert_eq!(color, crate::Color::White);
                 for (move1, move2) in possible_moves {
                     valid_actions.push(checker_moves_to_trictrac_action(
                         &move1, &move2, &color, game_state,
@@ -227,7 +206,7 @@ pub fn get_valid_actions(game_state: &crate::GameState) -> Vec<TrictracAction> {
                 }
             }
             TurnStage::Move => {
-                let rules = store::MoveRules::new(&color, &game_state.board, game_state.dice);
+                let rules = crate::MoveRules::new(&color, &game_state.board, game_state.dice);
                 let mut possible_moves = rules.get_possible_moves_sequences(true, vec![]);
                 if possible_moves.is_empty() {
                     // Empty move
@@ -235,7 +214,7 @@ pub fn get_valid_actions(game_state: &crate::GameState) -> Vec<TrictracAction> {
                 }
 
                 // Modififier checker_moves_to_trictrac_action si on doit gérer Black
-                assert_eq!(color, store::Color::White);
+                assert_eq!(color, crate::Color::White);
                 for (move1, move2) in possible_moves {
                     valid_actions.push(checker_moves_to_trictrac_action(
                         &move1, &move2, &color, game_state,
@@ -255,8 +234,8 @@ pub fn get_valid_actions(game_state: &crate::GameState) -> Vec<TrictracAction> {
 fn checker_moves_to_trictrac_action(
     move1: &CheckerMove,
     move2: &CheckerMove,
-    color: &store::Color,
+    color: &crate::Color,
-    state: &crate::GameState,
+    state: &GameState,
 ) -> TrictracAction {
     let to1 = move1.get_to();
     let to2 = move2.get_to();
@@ -314,7 +293,7 @@ fn checker_moves_to_trictrac_action(
 }
 
 /// Retourne les indices des actions valides
-pub fn get_valid_action_indices(game_state: &crate::GameState) -> Vec<usize> {
+pub fn get_valid_action_indices(game_state: &GameState) -> Vec<usize> {
     get_valid_actions(game_state)
         .into_iter()
         .map(|action| action.to_action_index())
@@ -322,11 +301,11 @@ pub fn get_valid_action_indices(game_state: &crate::GameState) -> Vec<usize> {
 }
 
 /// Sélectionne une action valide aléatoire
-pub fn sample_valid_action(game_state: &crate::GameState) -> Option<TrictracAction> {
+pub fn sample_valid_action(game_state: &GameState) -> Option<TrictracAction> {
-    use rand::{seq::SliceRandom, thread_rng};
+    use rand::{prelude::IndexedRandom, rng};
 
     let valid_actions = get_valid_actions(game_state);
-    let mut rng = thread_rng();
+    let mut rng = rng();
     valid_actions.choose(&mut rng).cloned()
 }