This solves the Numba issue; the Awkward one will need an Awkward fix…

… (update ak.nan_to_num).

src/vector/_backends/numba_object.py

@@ -52,16 +52,31 @@
 def nan_to_num(x, copy=True, nan=0.0, posinf=None, neginf=None):
     if isinstance(x, numba.types.Array):
 
-        def nan_to_num_impl(x, copy=True, nan=0.0, posinf=None, neginf=None):
-            out = numpy.copy(x).reshape(-1) if copy else x.reshape(-1)
-            for i in range(len(out)):
-                if numpy.isnan(out[i]):
-                    out[i] = nan
-                if posinf is not None and numpy.isinf(out[i]) and out[i] > 0:
-                    out[i] = posinf
-                if neginf is not None and numpy.isinf(out[i]) and out[i] < 0:
-                    out[i] = neginf
-            return out.reshape(x.shape)
+        if isinstance(nan, numba.types.Array):
+
+            def nan_to_num_impl(x, copy=True, nan=0.0, posinf=None, neginf=None):
+                out = numpy.copy(x).reshape(-1) if copy else x.reshape(-1)
+                for i in range(len(out)):
+                    if numpy.isnan(out[i]):
+                        out[i] = nan[i]
+                    if posinf is not None and numpy.isinf(out[i]) and out[i] > 0:
+                        out[i] = posinf
+                    if neginf is not None and numpy.isinf(out[i]) and out[i] < 0:
+                        out[i] = neginf
+                return out.reshape(x.shape)
+
+        else:
+
+            def nan_to_num_impl(x, copy=True, nan=0.0, posinf=None, neginf=None):
+                out = numpy.copy(x).reshape(-1) if copy else x.reshape(-1)
+                for i in range(len(out)):
+                    if numpy.isnan(out[i]):
+                        out[i] = nan
+                    if posinf is not None and numpy.isinf(out[i]) and out[i] > 0:
+                        out[i] = posinf
+                    if neginf is not None and numpy.isinf(out[i]) and out[i] < 0:
+                        out[i] = neginf
+                return out.reshape(x.shape)
 
     else: