minor (unrelated) fix to worstcaserisk, tests pass and full coverage …

…of new code

Makefile

@@ -37,7 +37,7 @@ test:  ## run tests w/ cov report
 	# $(BINDIR)/bandit $(PROJECT)/*.py $(TESTS)/*.py
 
 lint:  ## run linter
-	$(BINDIR)/pylint $(PROJECT) $(EXTRA_SCRIPTS) $(EXAMPLES)
+	$(BINDIR)/pylint $(PROJECT) # $(EXTRA_SCRIPTS) $(EXAMPLES)
 	$(BINDIR)/diff-quality --violations=pylint --config-file pyproject.toml
 
 docs:  ## build docs

cvxportfolio/costs.py

@@ -128,6 +128,8 @@ def __init__(self, costs, multipliers):
                     "You can only sum cost instances to other cost instances.")
         self.costs = costs
         self.multipliers = multipliers
+        # this is changed by WorstCaseRisk before compiling to Cvxpy
+        self.DO_CONVEXITY_CHECK = True
 
     def __add__(self, other):
         """Add other (combined) cost to self."""
@@ -194,8 +196,10 @@ def compile_to_cvxpy(self, w_plus, z, w_plus_minus_w_bm):
                     cost.compile_to_cvxpy(w_plus, z, w_plus_minus_w_bm)
             if not add.is_dcp():
                 raise ConvexSpecificationError(cost * multiplier)
-            if not add.is_concave():
+            if self.DO_CONVEXITY_CHECK and (not add.is_concave()):
                 raise ConvexityError(cost * multiplier)
+            if (not self.DO_CONVEXITY_CHECK) and add.is_concave():
+                raise ConvexityError(-cost * multiplier)
             expression += add
         return expression
 

cvxportfolio/estimator.py

@@ -85,6 +85,8 @@ def finalize_estimator(self, **kwargs):
         We aren't currently using in the rest of the library but we plan to
         move the caching logic in it.
 
+        .. versionadded:: 1.1.0
+
         :param kwargs: Reserved for future expansion.
         :type kwargs: dict
         """
@@ -95,6 +97,8 @@ def finalize_estimator(self, **kwargs):
     def finalize_estimator_recursive(self, **kwargs):
         """Recursively finalize all estimators in a policy.
 
+        .. versionadded:: 1.1.0
+
         :param kwargs: Parameters sent down an estimator tree to finalize it.
         :type kwargs: dict
         """

cvxportfolio/risks.py

@@ -399,13 +399,17 @@ def compile_to_cvxpy(self, w_plus, z, w_plus_minus_w_bm):
 class WorstCaseRisk(Cost):
     """Select the most restrictive risk model for each value of the allocation.
 
-    vector.
-
     Given a list of risk models, penalize the portfolio allocation by
     the one with highest risk value at the solution point. If uncertain
     about which risk model to use this procedure can be an easy
     solution.
 
+    :Example:
+
+        >>> risk_model = cvx.WorstCaseRisk(
+                [cvx.FullCovariance(),
+                cvx.DiagonalCovariance() + 0.25 * cvx.RiskForecastError()])
+
     :param riskmodels: risk model instances on which to compute the
         worst-case risk.
     :type riskmodels: list
@@ -446,8 +450,16 @@ def compile_to_cvxpy(self, w_plus, z, w_plus_minus_w_bm):
         :returns: Cvxpy expression representing the risk model.
         :rtype: cvxpy.expression
         """
-        risks = [risk.compile_to_cvxpy(w_plus, z, w_plus_minus_w_bm)
-                 for risk in self.riskmodels]
+        risks = []
+        for risk in self.riskmodels:
+            # this is needed if user provides individual risk terms
+            # that are composed objects (CombinedCost)
+            # it will check concavity instead of convexity
+            risk.DO_CONVEXITY_CHECK = False
+            risks.append(risk.compile_to_cvxpy(w_plus, z, w_plus_minus_w_bm))
+            # we also change it back in case the user is sharing the instance
+            risk.DO_CONVEXITY_CHECK = True
+
         return cp.max(cp.hstack(risks))
 
     def finalize_estimator_recursive(self, **kwargs):

cvxportfolio/tests/test_simulator.py

@@ -433,16 +433,34 @@ def test_backtest(self):
         """Test simple back-test."""
         pol = cvx.SinglePeriodOptimization(
             cvx.ReturnsForecast() - cvx.ReturnsForecastError()
-            - .5 * cvx.FullCovariance(),
+            - .5 * cvx.WorstCaseRisk(
+                [cvx.FullCovariance(),
+                cvx.DiagonalCovariance() + .25 * cvx.DiagonalCovariance()]),
             [cvx.LeverageLimit(1)], verbose=True,
-            solver=self.default_qp_solver)
+            solver=self.default_socp_solver)
         sim = cvx.MarketSimulator(
             market_data=self.market_data_4, base_location=self.datadir)
         result = sim.backtest(pol, pd.Timestamp(
             '2023-01-01'), pd.Timestamp('2023-04-20'))
 
         print(result)
 
+    def test_wrong_worstcase(self):
+        """Test wrong worst-case convexity."""
+        pol = cvx.SinglePeriodOptimization(
+            cvx.ReturnsForecast() - cvx.ReturnsForecastError()
+            - .5 * cvx.WorstCaseRisk(
+                [-cvx.FullCovariance(),
+                cvx.DiagonalCovariance() + .25 * cvx.DiagonalCovariance()]),
+            [cvx.LeverageLimit(1)], verbose=True,
+            solver=self.default_socp_solver)
+        sim = cvx.MarketSimulator(
+            market_data=self.market_data_4, base_location=self.datadir)
+
+        with self.assertRaises(ConvexityError):
+            sim.backtest(pol, pd.Timestamp(
+                '2023-01-01'), pd.Timestamp('2023-04-20'))
+
     def test_backtest_changing_universe(self):
         """Test back-test with changing universe."""
         sim = cvx.MarketSimulator(