Missing metadata (#170)

* TST: Fixing metadata

* STY: removing references to bokeh charts

* STY: fixing spacing

* Removing red ray, because bokeh has super weird scaling going on

gneiss/plot/_plot.py

@@ -289,9 +289,6 @@ def ols_summary(output_dir: str, model: OLSModel, ndim=10) -> None:
     mse_p.quad(top=mse_hist, bottom=0, left=edges[:-1], right=edges[1:],
                fill_color="#FFFF00", line_color="#033649", fill_alpha=0.5,
                legend='CV Mean Squared Error')
-    mse_p.ray(x=model.mse, y=0, length=h*max(mse_hist),
-              angle=1.57079633, color='red',
-              legend='Model Error', line_width=0.5)
 
     # Histogram of prediction error from cross validation
     pred_p = figure(title="Prediction Error",
@@ -300,9 +297,6 @@ def ols_summary(output_dir: str, model: OLSModel, ndim=10) -> None:
     pred_p.quad(top=pred_hist, bottom=0, left=edges[:-1], right=edges[1:],
                 fill_color="#00FFFF", line_color="#033649", fill_alpha=0.5,
                 legend='Prediction Error')
-    pred_p.ray(x=model.mse, y=0, length=h,
-               angle=1.57079633, color='red',
-               legend='Model Error', line_width=0.5)
 
     # Explained sum of squares
     ess = pd.Series({r.model.endog_names: r.ess for r in model.results})

gneiss/regression/_ols.py

@@ -164,11 +164,11 @@ def ols(formula, table, metadata, tree, **kwargs):
                                                               tree)
     ilr_table, basis = _to_balances(table, tree)
 
-    ilr_table, metadata = ilr_table.align(metadata, join='inner', axis=0)
     # one-time creation of exogenous data matrix allows for faster run-time
     metadata = _type_cast_to_float(metadata)
     x = dmatrix(formula, metadata, return_type='dataframe')
 
+    ilr_table, x = ilr_table.align(x, join='inner', axis=0)
     submodels = _fit_ols(ilr_table, x)
 
     basis = pd.DataFrame(basis, index=ilr_table.columns,

gneiss/regression/tests/test_ols.py

@@ -135,6 +135,37 @@ def test_ols_immutable(self):
         self.assertEqual(str(table), str(exp_table))
         self.assertEqual(str(exp_tree), str(res.tree))
 
+    def test_ols_missing_metadata(self):
+        np.random.seed(0)
+        A = np.array  # aliasing for the sake of pep8
+        table = pd.DataFrame({
+            's1': ilr_inv(A([1., 1.])),
+            's2': ilr_inv(A([1., 2.])),
+            's3': ilr_inv(A([1., 3.])),
+            's4': ilr_inv(A([1., 4.])),
+            's5': ilr_inv(A([1., 5.])),
+            's6': ilr_inv(A([1., 5.])),
+            's7': ilr_inv(A([1., 5.]))},
+            index=['a', 'b', 'c']).T
+
+        tree = TreeNode.read(['((c,d),(b,a));'])
+        exp_tree = TreeNode.read(['((b,a)y1,c)y0;\n'])
+        metadata = pd.DataFrame({
+            'lame': [1, 1, 1, 1, 1, 0],
+            'real': [1, 2, 3, 4, 5, np.nan]
+        }, index=['s1', 's2', 's3', 's4', 's5', 's7'])
+
+        res = ols('real + lame', table, metadata, tree)
+        res.fit()
+        self.assertEqual(str(exp_tree), str(res.tree))
+
+        exp_coefs = pd.DataFrame(
+            [[-7.494005e-16, -7.494005e-16, -1.000000e+00],
+             [5.000000e-01, 5.000000e-01, -1.294503e-16]],
+            columns=['Intercept', 'lame', 'real'], index=['y0', 'y1'])
+        pdt.assert_frame_equal(exp_coefs, res.coefficients(),
+                               check_less_precise=True)
+
     def test_ols_empty_table_error(self):
         A = np.array  # aliasing for the sake of pep8
         table = pd.DataFrame({