Simplified more

wafo/transform/models.py

@@ -395,16 +395,12 @@ class TrLinear(TrCommon2):
         return y
 
     def _dat2gauss(self, x, *xi):
-        ysigma, xsigma = self.ysigma, self.sigma
+        return self._transform(self.ymean, self.ysigma,
-        ymean, xmean = self.ymean, self.mean
+                               self.mean, self.sigma, x, xi)
-
-        return self._transform(ymean, ysigma, xmean, xsigma, x, xi)
 
     def _gauss2dat(self, y, *yi):
-        ysigma, xsigma = self.ysigma, self.sigma
+        return self._transform(self.mean, self.sigma,
-        ymean, xmean = self.ymean, self.mean
+                               self.ymean, self.ysigma, y, yi)
-
-        return self._transform(xmean, xsigma, ymean, ysigma, y, yi)
 
 
 class TrOchi(TrCommon2):
@@ -512,7 +508,6 @@ class TrOchi(TrCommon2):
         mean2 = a * sig22 / my2  # % choose the smallest mean
 
         self._phat = [sigma1, mean1, gam_a, gam_b, sigma2, mean2]
-        return
 
     def _get_par(self):
         '''
@@ -529,10 +524,16 @@ class TrOchi(TrCommon2):
         mean2 = self._phat[5]
         return ga, gb, sigma2, mean2
 
-    def _forward(self, g, xn, gab, idx):
+    def _transform(self, fun, x1, x2, gab, idx):
         if gab != 0:
-            np.put(g, idx, (-expm1(-gab * xn[idx])) / gab)
+            np.put(x1, idx, -fun(-gab * x2[idx]) / gab)
-        return g
+        return x1
+
+    def _backward(self, xn, y2, gab, idx):
+        return self._transform(log1p, xn, y2, gab, idx)
+
+    def _forward(self, y2, xn, gab, idx):
+        return self._transform(expm1, y2, xn, gab, idx)
 
     def _dat2gauss(self, x, *xi):
         if len(xi) > 0:
@@ -546,17 +547,12 @@ class TrOchi(TrCommon2):
         igp, = where(0 <= xn)
         igm, = where(xn < 0)
 
-        g = xn.copy()
+        y2 = xn.copy()
-        g = self._forward(g, xn, ga, igp)
+        y2 = self._forward(y2, xn, ga, igp)
-        g = self._forward(g, xn, gb, igm)
+        y2 = self._forward(y2, xn, gb, igm)
 
-        g.shape = shape0
+        y2.shape = shape0
-        return (g - mean2) * self.ysigma / sigma2 + self.ymean
+        return (y2 - mean2) * self.ysigma / sigma2 + self.ymean
-
-    def _backward(self, xn, gab, idx):
-        if gab != 0:
-            np.put(xn, idx, -log1p(-gab * xn[idx]) / gab)
-        return xn
 
     def _gauss2dat(self, y, *yi):
         if len(yi) > 0:
@@ -567,13 +563,13 @@ class TrOchi(TrCommon2):
         sigma = self.sigma
 
         yn = (atleast_1d(y) - self.ymean) / self.ysigma
-        xn = sigma2 * yn.ravel() + mean2
+        y2 = sigma2 * yn.ravel() + mean2
-
-        igp, = where(0 <= xn)
-        igm, = where(xn < 0)
 
-        xn = self._backward(xn, ga, igp)
+        igp, = where(0 <= y2)
-        xn = self._backward(xn, gb, igm)
+        igm, = where(y2 < 0)
+        xn = y2.copy()
+        xn = self._backward(xn, y2, ga, igp)
+        xn = self._backward(xn, y2, gb, igm)
 
         xn.shape = yn.shape
         return sigma * xn + mean