BUG: sample list-valued stochastic attributes through the seeded generator

dict_generator drew list-valued attributes with the stdlib random.choice, which reads from an unseeded global Random instance, so random_seed did not make those attributes reproducible. Draw the index from this model's seeded numpy generator instead. Indexing (not numpy.random.choice) also avoids coercing a heterogeneous list -- Function objects, paths, arrays -- to a single dtype.

Adds a unit test that a list-valued attribute is reproducible under a fixed seed and that heterogeneous entries are returned unchanged.

Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>

Author: thc1006

rocketpy/stochastic/stochastic_model.py

@@ -3,8 +3,6 @@
 Stochastic classes.
 """
 
-from random import choice
-
 import numpy as np
 
 from rocketpy.mathutils.function import Function
@@ -532,7 +530,16 @@ def dict_generator(self):
                 dist_sampler = value[-1]
                 generated_dict[arg] = dist_sampler(value[0], value[1])
             elif isinstance(value, list):
-                generated_dict[arg] = choice(value) if value else value
+                # Draw the index from this model's seeded generator so a
+                # list-valued attribute is reproducible under random_seed. The
+                # stdlib random.choice draws from an unseeded global instance,
+                # and numpy's choice coerces a heterogeneous list (Function,
+                # paths, arrays) to a single dtype; indexing avoids both.
+                if value:
+                    index = int(self.__random_number_generator.integers(len(value)))
+                    generated_dict[arg] = value[index]
+                else:
+                    generated_dict[arg] = value
             elif isinstance(value, CustomSampler):
                 try:
                     generated_dict[arg] = value.sample(n_samples=1)[0]

tests/unit/stochastic/test_stochastic_model.py

@@ -1,5 +1,38 @@
+from types import SimpleNamespace
+
 import pytest
 
+from rocketpy.stochastic.stochastic_model import StochasticModel
+
+
+def _sampled_option(model):
+    """Return the value ``dict_generator`` picks for the ``options`` attribute."""
+    return next(model.dict_generator())["options"]
+
+
+def test_list_attribute_sampling_is_reproducible_under_seed():
+    """A list-valued stochastic attribute is drawn through the model's own seeded
+    numpy generator, so a fixed seed reproduces the choice. It used to be drawn
+    with the stdlib ``random.choice`` (an unseeded global instance), which
+    ``random_seed`` could not govern. Heterogeneous entries (paths, callables,
+    lists) are returned unchanged rather than coerced to a numpy dtype the way
+    ``numpy.random.choice`` would.
+    """
+    options = ["/motor/a.eng", "/motor/b.eng", (lambda t: t), [1, 2, 3]]
+    model = StochasticModel(obj=SimpleNamespace(), options=options)
+
+    model._set_stochastic(42)
+    first = _sampled_option(model)
+    model._set_stochastic(42)
+    assert _sampled_option(model) == first, "same seed must reproduce the choice"
+    assert any(first is option for option in options), "object returned unchanged"
+
+    chosen_ids = set()
+    for seed in range(16):
+        model._set_stochastic(seed)
+        chosen_ids.add(id(_sampled_option(model)))
+    assert len(chosen_ids) > 1, "different seeds must be able to pick differently"
+
 
 @pytest.mark.parametrize(
     "fixture_name",