fix(ocean/tmaze): migrate binding.c from env_binding.h to vecenv.h API

ocean/tmaze/binding.c

@@ -1,17 +1,21 @@
 #include "tmaze.h"
 
+#define OBS_SIZE 4
+#define NUM_ATNS 1
+#define ACT_SIZES {3}
+#define OBS_TENSOR_T ByteTensor
+
 #define Env TMaze
-#include "../env_binding.h"
+#include "vecenv.h"
 
-static int my_init(Env* env, PyObject* args, PyObject* kwargs) {
-    env->size = unpack(kwargs, "size");
-    return 0;
+void my_init(Env* env, Dict* kwargs) {
+    env->size = (int)dict_get(kwargs, "size")->value;
+    env->num_agents = 1;
 }
 
-static int my_log(PyObject* dict, Log* log) {
-    assign_to_dict(dict, "perf", log->perf);
-    assign_to_dict(dict, "score", log->score);
-    assign_to_dict(dict, "episode_return", log->episode_return);
-    assign_to_dict(dict, "episode_length", log->episode_length);
-    return 0;
+void my_log(Log* log, Dict* out) {
+    dict_set(out, "perf", log->perf);
+    dict_set(out, "score", log->score);
+    dict_set(out, "episode_return", log->episode_return);
+    dict_set(out, "episode_length", log->episode_length);
 }

ocean/tmaze/tmaze.c

@@ -8,12 +8,12 @@ int main() {
     c_render(&env);
     while (!WindowShouldClose()) {
         if (IsKeyDown(KEY_LEFT_SHIFT)) {
-            env.actions[0] = FORWARD;
-            if (IsKeyDown(KEY_LEFT)  || IsKeyDown(KEY_A)) env.actions[0] = LEFT;
-            if (IsKeyDown(KEY_RIGHT) || IsKeyDown(KEY_D)) env.actions[0] = RIGHT;
+            env.actions[0] = (float)FORWARD;
+            if (IsKeyDown(KEY_LEFT)  || IsKeyDown(KEY_A)) env.actions[0] = (float)LEFT;
+            if (IsKeyDown(KEY_RIGHT) || IsKeyDown(KEY_D)) env.actions[0] = (float)RIGHT;
 
         } else {
-            env.actions[0] = rand() % 3;
+            env.actions[0] = (float)(rand() % 3);
         }
         c_step(&env);
         c_render(&env);

ocean/tmaze/tmaze.h

@@ -27,11 +27,13 @@ typedef struct {
 typedef struct {
     Log log; // Required field. Env binding code uses this to aggregate logs
     unsigned char* observations; // Required. You can use any obs type, but make sure it matches in Python!
-    int* actions; // Required. int* for discrete/multidiscrete, float* for box
+    float* actions; // Required. float* to match vecenv.h allocation
     float* rewards; // Required
-    unsigned char* terminals; // Required. We don't yet have truncations as standard yet
+    float* terminals; // Required. float* to match vecenv.h allocation
     int size; // length of the corridor
     int tick;
+    int num_agents; // Required by vecenv.h default my_vec_init
+    unsigned int rng; // Required by vecenv.h default my_vec_init
 
     unsigned char state; // Internal current position in the maze
     unsigned char starting_state; // Starting state (2 or 3)
@@ -42,9 +44,9 @@ typedef struct {
 
 TMaze* allocate_TMaze(TMaze *env) {
     env->observations = calloc(4, sizeof(unsigned char));
-    env->actions = calloc(1, sizeof(int));
+    env->actions = calloc(1, sizeof(float));
     env->rewards = calloc(1, sizeof(float));
-    env->terminals = calloc(1, sizeof(unsigned char));
+    env->terminals = calloc(1, sizeof(float));
     return env;
 }
 
@@ -99,7 +101,7 @@ void c_step(TMaze* env) {
     env->terminals[0] = 0;
     env->rewards[0] = 0;
 
-    int action = env->actions[0];
+    int action = (int)env->actions[0];
 
     if (env->state == env->size -1) {
         const int left_reward = (env->starting_state == 2) ? 1 : -1;