Added tests for skutils

stemu/emu.py

@@ -68,8 +68,8 @@ def fit(self, X, t, y):
         self.t = t
 
         X = self.X_pipeline.fit_transform(X)
-        t = self.t_pipeline.fit_transform(t, y)
         y = self.y_pipeline.fit_transform(y)
+        t = self.t_pipeline.fit_transform(t, y)
 
         ty = self.ty_pipeline.fit_transform(np.block([[t], [y]]))
         t, y = ty[0], ty[1:]
@@ -112,6 +112,6 @@ def predict(self, X, t=None):
         y = self.model.predict(X)
         _, _, y = unstack(X, y)
         ty = self.ty_pipeline.inverse_transform(np.block([[t], [y]]))
-        t, y = ty[0], ty[1:]
+        _, y = ty[0], ty[1:]
         y = self.y_pipeline.inverse_transform(y)
         return y

tests/test_skutils.py

@@ -0,0 +1,41 @@
+import numpy as np
+from numpy.testing import assert_allclose
+
+from stemu.skutils import CDFTransformer, FunctionScaler, IdentityTransformer
+
+
+def test_CDFTransformer():
+    t = np.linspace(0, 1, 100)
+    y = t**3 - t + 1
+    y = y * np.random.rand(20, len(t))
+    cdf = CDFTransformer()
+    assert isinstance(cdf.fit(t, y), CDFTransformer)
+    t_ = cdf.transform(t)
+    assert not (t == t_).all()
+    t_ = cdf.inverse_transform(t_)
+    assert (t == t_).all()
+
+
+def test_FunctionScaler():
+    t = np.linspace(0, 1, 100)
+    y = t**3 - t + 1
+    y = y * np.random.rand(20, len(t))
+    X = np.block([[t], [y]])
+
+    fs = FunctionScaler()
+    assert isinstance(fs.fit(X), FunctionScaler)
+    X_ = fs.transform(X)
+    assert_allclose(X_[1:].mean(axis=0), 0, atol=1e-15)
+    assert_allclose(X_[1:].std(axis=0), 1, atol=1e-15)
+    assert_allclose(fs.inverse_transform(X_), X)
+
+
+def test_IdentityTransformer():
+    y = np.random.rand(10, 20)
+
+    identity = IdentityTransformer()
+    assert isinstance(identity.fit(y), IdentityTransformer)
+    t_ = identity.transform(t)
+    assert (t == t_).all()
+    t_ = identity.inverse_transform(t_)
+    assert (t == t_).all()