Fix: make __getitem__ delegate to states array

circadian/models.py

@@ -85,15 +85,9 @@ def __call__(self, timepoint: float) -> np.ndarray: # state of the system
 
 # %% ../nbs/api/00_models.ipynb 11
 @patch_to(DynamicalTrajectory)
-def __getitem__(self, time_idx: int) -> Tuple[float, np.ndarray]:
-    "Return the time and state at index idx"
-    # idx input checking
-    if not isinstance(time_idx, int):
-        raise TypeError("idx must be int")
-    if time_idx < -1 or time_idx >= len(self.time):
-        raise ValueError(f"idx must be within 0 and {len(self.time)-1}, got {time_idx}")
-
-    return self.time[time_idx], self.states[time_idx, ...]
+def __getitem__(self, idx):
+    "Index directly into the states array (time_idx, state_idx, batch_idx)"
+    return self.states[idx]
 
 # %% ../nbs/api/00_models.ipynb 12
 @patch_to(DynamicalTrajectory)

nbs/api/00_models.ipynb

@@ -194,15 +194,9 @@
     "#| export\n",
     "#| hide\n",
     "@patch_to(DynamicalTrajectory)\n",
-    "def __getitem__(self, time_idx: int) -> Tuple[float, np.ndarray]:\n",
-    "    \"Return the time and state at index idx\"\n",
-    "    # idx input checking\n",
-    "    if not isinstance(time_idx, int):\n",
-    "        raise TypeError(\"idx must be int\")\n",
-    "    if time_idx < -1 or time_idx >= len(self.time):\n",
-    "        raise ValueError(f\"idx must be within 0 and {len(self.time)-1}, got {time_idx}\")\n",
-    "    \n",
-    "    return self.time[time_idx], self.states[time_idx, ...]"
+    "def __getitem__(self, idx):\n",
+    "    \"Index directly into the states array (time_idx, state_idx, batch_idx)\"\n",
+    "    return self.states[idx]\n"
    ]
   },
   {
@@ -4818,4 +4812,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 4
-}
+}

nbs/test/test_models.ipynb

@@ -124,23 +124,26 @@
     "states[:, 0] = np.sin(time)\n",
     "states[:, 1] = np.cos(time)\n",
     "traj = DynamicalTrajectory(time, states)\n",
-    "test_eq(traj[0], (0.0, np.array([0.0, 1.0])))\n",
-    "states = traj[-1][1]\n",
-    "difference = np.abs(np.sum(states - (np.sin(np.pi), np.cos(np.pi))))\n",
-    "test_eq(difference < 1e-4, True)\n",
-    "# handle batch\n",
+    "# index by time point\n",
+    "test_eq(traj[0], np.array([0.0, 1.0]))\n",
+    "test_eq(np.abs(traj[-1][0] - np.sin(np.pi)) < 1e-4, True)\n",
+    "# index by time point and state\n",
+    "test_eq(traj[0, 0], 0.0)\n",
+    "test_eq(traj[0, 1], 1.0)\n",
+    "# slice\n",
+    "test_eq(traj[:2].shape, (2, variables))\n",
+    "# batch trajectory\n",
     "batches = 5\n",
     "batch_states = np.zeros((total_timepoints, variables, batches))\n",
     "batch_states[:, 0, :] = np.sin(time)[:, None]\n",
     "batch_states[:, 1, :] = np.cos(time)[:, None]\n",
     "batch_traj = DynamicalTrajectory(time, batch_states)\n",
-    "test_eq(batch_traj[0][0], 0.0)\n",
-    "test_eq(np.all(batch_traj[0][1][0]==np.zeros(batches)), True)\n",
-    "test_eq(np.all(batch_traj[0][1][1]==np.ones(batches)), True)\n",
-    "# test error handling\n",
-    "test_fail(lambda: traj[\"1\"], contains=\"idx must be int\")\n",
-    "test_fail(lambda: traj[-2], contains=\"idx must be within 0 and\")\n",
-    "test_fail(lambda: traj[1000], contains=\"idx must be within 0 and\")"
+    "test_eq(batch_traj[0].shape, (variables, batches))\n",
+    "test_eq(np.all(batch_traj[0, 0] == np.zeros(batches)), True)\n",
+    "test_eq(np.all(batch_traj[0, 1] == np.ones(batches)), True)\n",
+    "# index by time point, state, and batch\n",
+    "test_eq(batch_traj[0, 0, 0], 0.0)\n",
+    "test_eq(batch_traj[0, 1, 0], 1.0)"
    ]
   },
   {
@@ -2161,4 +2164,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 2
-}
+}

setup.py

@@ -1,4 +1,4 @@
-from pkg_resources import parse_version
+from packaging.version import Version as parse_version
 from configparser import ConfigParser
 import setuptools 
 from setuptools import Extension