fix: align integer-allocation prices by ticker

.github/workflows/discrete-allocation.yml

@@ -0,0 +1,24 @@
+name: Discrete allocation regressions
+on:
+  pull_request:
+  push:
+    branches: [main]
+permissions:
+  contents: read
+jobs:
+  discrete-allocation:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    strategy:
+      matrix:
+        python-version: ['3.11', '3.12']
+    steps:
+      - uses: actions/checkout@v6
+        with:
+          persist-credentials: false
+      - uses: astral-sh/setup-uv@v7
+        with:
+          python-version: ${{ matrix.python-version }}
+      - run: uv venv
+      - run: uv pip install -e . pytest ecos
+      - run: uv run --no-sync pytest tests/test_discrete_allocation.py -q

pypfopt/discrete_allocation.py

@@ -356,7 +356,8 @@ def lp_portfolio(self, reinvest=False, verbose=False, solver=None):
             self.allocation = self._remove_zero_positions(self.allocation)
             return self.allocation, long_leftover + short_leftover
 
-        p = self.latest_prices.values
+        # Match the weight order by ticker; Series order and extra prices may differ.
+        p = self.latest_prices.loc[[ticker for ticker, _ in self.weights]].values
         n = len(p)
         w = np.fromiter([i[1] for i in self.weights], dtype=float)
 

pypfopt/hierarchical_portfolio.py

@@ -176,9 +176,6 @@ def optimize(self, linkage_method="single"):
         OrderedDict
             weights for the HRP portfolio
         """
-        if linkage_method not in sch._LINKAGE_METHODS:
-            raise ValueError("linkage_method must be one recognised by scipy")
-
         if self.returns is None:
             cov = self.cov_matrix
             corr = cov_to_corr(self.cov_matrix).round(6)
@@ -192,6 +189,7 @@ def optimize(self, linkage_method="single"):
         matrix = np.sqrt(np.clip((1.0 - corr) / 2.0, a_min=0.0, a_max=1.0))
         dist = ssd.squareform(matrix, checks=False)
 
+        # Let the public API validate methods; SciPy's private registry can change.
         self.clusters = sch.linkage(dist, linkage_method)
         sort_ix = HRPOpt._get_quasi_diag(self.clusters)
         ordered_tickers = corr.index[sort_ix].tolist()

tests/test_discrete_allocation.py

@@ -7,6 +7,21 @@
 from tests.utilities_for_tests import get_data, setup_efficient_frontier
 
 
+@pytest.mark.parametrize(
+    "prices",
+    [
+        pd.Series({"B": 100.0, "A": 10.0}),
+        pd.Series({"UNUSED": 5.0, "B": 100.0, "A": 10.0}),
+    ],
+)
+def test_lp_portfolio_aligns_prices_with_weight_tickers(prices):
+    da = DiscreteAllocation({"A": 0.5, "B": 0.5}, prices, total_portfolio_value=200)
+    allocation, leftover = da.lp_portfolio(solver="SCIPY")
+    assert allocation == {"A": 10, "B": 1}
+    assert leftover == pytest.approx(0)
+    assert sum(prices[ticker] * shares for ticker, shares in allocation.items()) == 200
+
+
 def test_get_latest_prices():
     df = get_data()
     latest_prices = get_latest_prices(df)

tests/test_hrp.py

@@ -6,6 +6,17 @@
 from tests.utilities_for_tests import get_data, resource
 
 
+@pytest.mark.parametrize(
+    "method", ["single", "complete", "average", "weighted", "centroid", "median", "ward"]
+)
+def test_documented_linkage_methods(method):
+    cov = pd.DataFrame(np.eye(3), index=list("ABC"), columns=list("ABC"))
+    weights = HRPOpt(cov_matrix=cov).optimize(linkage_method=method)
+    assert set(weights) == set("ABC")
+    assert sum(weights.values()) == pytest.approx(1)
+    assert all(weight > 0 for weight in weights.values())
+
+
 def test_hrp_errors():
     with pytest.raises(ValueError):
         hrp = HRPOpt()