cached the algorithms in pytest

For speed, I've cached the Pytest algorithms such that they are not re-initiated every run. Testing is faster now. Especially good for Matlab-based testing where the Matlab engine needs initiating (different PR)

Author: oliverchampion

tests/IVIMmodels/unit_tests/test_ivim_fit.py

@@ -29,6 +29,7 @@ def tolerances_helper(tolerances, data):
         tolerances["atol"] = tolerances.get("atol", {"f": 2e-1, "D": 5e-4, "Dp": 4e-2})
     return tolerances
 
+
 def data_ivim_fit_saved():
     # Find the algorithms from algorithms.json
     file = pathlib.Path(__file__)
@@ -45,8 +46,8 @@ def data_ivim_fit_saved():
     bvals = all_data.pop('config')
     bvals = bvals['bvalues']
     first=True
-    for name, data in all_data.items():
-        for algorithm in algorithms:
+    for algorithm in algorithms:
+        for name, data in all_data.items():
             algorithm_dict = algorithm_information.get(algorithm, {})
             xfail = {"xfail": name in algorithm_dict.get("xfail_names", {}),
                 "strict": algorithm_dict.get("xfail_names", {}).get(name, True)}
@@ -59,15 +60,38 @@ def data_ivim_fit_saved():
                     first = False
             yield name, bvals, data, algorithm, xfail, kwargs, tolerances, skiptime
 
+
+def make_hashable(obj):
+    if isinstance(obj, dict):
+        return tuple(sorted((k, make_hashable(v)) for k, v in obj.items()))
+    elif isinstance(obj, (list, tuple)):
+        return tuple(make_hashable(i) for i in obj)
+    else:
+        return obj
+
+
+@pytest.fixture(scope="module")
+def algorithm_cache():
+    cache = {}
+
+    def get_instance(algorithm, kwargs):
+        hashable_key = (algorithm, make_hashable(kwargs))
+        if hashable_key not in cache:
+            cache[hashable_key] = OsipiBase(algorithm=algorithm, **kwargs)
+        return cache[hashable_key]
+
+    return get_instance
+
+
 @pytest.mark.parametrize("name, bvals, data, algorithm, xfail, kwargs, tolerances, skiptime", data_ivim_fit_saved())
-def test_ivim_fit_saved(name, bvals, data, algorithm, xfail, kwargs, tolerances,skiptime, request, record_property):
+def test_ivim_fit_saved(name, bvals, data, algorithm, xfail, kwargs, tolerances, skiptime, request, record_property, algorithm_cache):
     if xfail["xfail"]:
         mark = pytest.mark.xfail(reason="xfail", strict=xfail["strict"])
         request.node.add_marker(mark)
     signal = signal_helper(data["data"])
     tolerances = tolerances_helper(tolerances, data)
+    fit = algorithm_cache(algorithm, kwargs)
     start_time = time.time()  # Record the start time
-    fit = OsipiBase(algorithm=algorithm,  **kwargs)
     fit_result = fit.osipi_fit(signal, bvals)
     elapsed_time = time.time() - start_time  # Calculate elapsed time
     def to_list_if_needed(value):
@@ -153,10 +177,14 @@ def bound_input():
 
 
 @pytest.mark.parametrize("name, bvals, data, algorithm, xfail, kwargs, tolerances", bound_input())
-def test_bounds(name, bvals, data, algorithm, xfail, kwargs, tolerances, request):
+def test_bounds(name, bvals, data, algorithm, xfail, kwargs, tolerances, request, algorithm_cache):
+    if xfail["xfail"]:
+        mark = pytest.mark.xfail(reason="xfail", strict=xfail["strict"])
+        request.node.add_marker(mark)
     bounds = ([0.0008, 0.2, 0.01, 1.1], [0.0012, 0.3, 0.02, 1.3])
     # deliberately have silly bounds to see whether they are used
-    fit = OsipiBase(algorithm=algorithm, bounds=bounds, initial_guess = [0.001, 0.25, 0.015, 1.2], **kwargs)
+    extended_kwargs = {**kwargs, "bounds": bounds, "initial_guess": [0.001, 0.25, 0.015, 1.2]}
+    fit = algorithm_cache(algorithm, extended_kwargs)
     if fit.use_bounds:
         signal = signal_helper(data["data"])
         fit_result = fit.osipi_fit(signal, bvals)