have to wrap delta with np.array to make sure it is ndarray and result stays CoordinateFrame

Reviewed-on: #5

src/mdevaluate/checksum.py

@@ -73,7 +73,9 @@ def checksum(*args, csum=None):
         elif isinstance(arg, FunctionType):
             csum.update(strip_comments(inspect.getsource(arg)).encode())
             c = inspect.getclosurevars(arg)
-            for v in {**c.nonlocals, **c.globals}.values():
+            merged = {**c.nonlocals, **c.globals}
+            for key in sorted(merged):  # deterministic ordering
+                v = merged[key]
                 if v is not arg:
                     checksum(v, csum=csum)
         elif isinstance(arg, functools.partial):

src/mdevaluate/distribution.py

@@ -182,10 +182,10 @@ def tetrahedral_order(
     )
 
     # Connection vectors
-    neighbors_1 -= atoms
+    neighbors_1 = pbc_diff(neighbors_1, atoms, box=atoms.box)
-    neighbors_2 -= atoms
+    neighbors_2 = pbc_diff(neighbors_2, atoms, box=atoms.box)
-    neighbors_3 -= atoms
+    neighbors_3 = pbc_diff(neighbors_3, atoms, box=atoms.box)
-    neighbors_4 -= atoms
+    neighbors_4 = pbc_diff(neighbors_4, atoms, box=atoms.box)
 
     # Normed Connection vectors
     neighbors_1 /= np.linalg.norm(neighbors_1, axis=-1).reshape(-1, 1)

src/mdevaluate/pbc.py

@@ -149,32 +149,21 @@ def nojump(frame: CoordinateFrame, usecache: bool = True) -> CoordinateFrame:
             i0 = 0
             delta = 0
 
-        delta = (
+        delta = (delta
-            delta
+            + np.vstack(
-            + np.array(
-                np.vstack(
                 [m[i0 : abstep + 1].sum(axis=0) for m in reader.nojump_matrices]
-                ).T
+            ).T)
-            )
-            @ frame.box
-        )
 
         reader._nojump_cache[abstep] = delta
         while len(reader._nojump_cache) > NOJUMP_CACHESIZE:
             reader._nojump_cache.popitem(last=False)
-        delta = delta[selection, :]
     else:
-        delta = (
+        delta = np.vstack(
-            np.array(
+                [m[: frame.step + 1, selection].sum(axis=0) for m in reader.nojump_matrices]
-                np.vstack(
-                    [
-                        m[: frame.step + 1, selection].sum(axis=0)
-                        for m in reader.nojump_matrices
-                    ]
                 ).T
-            )
+    
-            @ frame.box
+    delta = delta[selection, :]
-        )
+    delta = np.array(delta @ frame.box)
     return frame - delta