Hypothesis tests

.coveragerc

@@ -0,0 +1,6 @@
+[run]
+source = symfit
+branch = True
+concurrency = multiprocessing
+omit = 
+    tests/*

.github/workflows/test.yaml

@@ -19,10 +19,10 @@ jobs:
       run: |
         python -m pip install --upgrade pip
         pip install wheel
-        pip install pytest
         pip install -r requirements.txt
+        pip install -r requirements_test.txt
         pip install matplotlib
         pip install -e .
     - name: Test with coverage
       run:
-        pytest
+        pytest --cov=symfit --hypothesis-show-statistics

.github/workflows/test_and_coveralls.yml

@@ -16,9 +16,8 @@ jobs:
       run: |
         python -m pip install --upgrade pip
         pip install wheel
-        pip install coveralls
-        pip install pytest
         pip install -r requirements.txt
+        pip install -r requirements_test.txt
         pip install matplotlib
         pip install -e .
     - name: Test with coverage
@@ -28,5 +27,4 @@ jobs:
       env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
       run: |
-          pip install --upgrade coveralls
           coveralls --service=github

.gitignore

@@ -6,3 +6,11 @@ build
 docs/_build
 docs/_doctrees
 dist
+
+.hypothesis/
+
+.eggs/
+
+symfit.egg-info/
+
+.coverage

requirements_test.txt

@@ -0,0 +1,6 @@
+pytest
+pytest-cov
+coverage
+coveralls
+hypothesis[numpy]
+

symfit/core/fit.py

@@ -26,6 +26,7 @@
 else:
     import funcsigs as inspect_sig
 
+
 class TakesData(object):
     """
     An base class for everything that takes data. Most importantly, it takes care

symfit/core/minimizers.py

@@ -347,6 +347,17 @@ def execute(self, bounds=None, jacobian=None, hessian=None, constraints=None, **
             :func:`scipy.optimize.minimize`. Note that your `method` will
             usually be filled by a specific subclass.
         """
+        if not self.params:
+            return FitResults(
+                model=DummyModel(params=[]),
+                popt=[],
+                covariance_matrix=np.array([]),
+                objective=self.objective,
+                minimizer=self,
+                message="No free parameters",
+                fun=self.objective(),
+                niter=0,
+            )
         ans = minimize(
             self.objective,
             self.initial_guesses,

symfit/core/models.py

@@ -328,8 +328,14 @@ def _init_from_dict(self, model_dict):
         # Everything at the bottom of the toposort is independent, at the top
         # dependent, and the rest interdependent.
         ordered = list(toposort(self.connectivity_mapping))
-        independent = sorted(ordered.pop(0), key=sort_func)
-        self.dependent_vars = sorted(ordered.pop(-1), key=sort_func)
+        if ordered:
+            independent = sorted(ordered.pop(0), key=sort_func)
+        else:
+            independent = []
+        if ordered:
+            self.dependent_vars = sorted(ordered.pop(-1), key=sort_func)
+        else:
+            self.dependent_vars = []
         self.interdependent_vars = sorted(
             [item for items in ordered for item in items],
             key=sort_func
@@ -907,7 +913,10 @@ def eval_jacobian(self, *args, **kwargs):
         # the parameter dimension. We do not include the component direction in
         # this, because the components can have independent shapes.
         for idx, comp in enumerate(jac):
-            jac[idx] = np.stack(np.broadcast_arrays(*comp))
+            if comp:
+                jac[idx] = np.stack(np.broadcast_arrays(*comp))
+            else:
+                jac[idx] = np.array([])
 
         return ModelOutput(self.keys(), jac)
 
@@ -951,7 +960,10 @@ def eval_hessian(self, *args, **kwargs):
         # the parameter dimension. We do not include the component direction in
         # this, because the components can have independent shapes.
         for idx, comp in enumerate(hess):
-            hess[idx] = np.stack(np.broadcast_arrays(*comp))
+            if comp:
+                hess[idx] = np.stack(np.broadcast_arrays(*comp))
+            else:
+                hess[idx] = np.array([])
 
         return ModelOutput(self.keys(), hess)