Atone for PyCharm's automatic refactoring

desc/optimize/bound_utils.py

@@ -2,7 +2,6 @@
 
 import functools
 
-import desc.utils
 from desc.backend import cond, jit, jnp
 
 from .utils import evaluate_quadratic_form_hess, evaluate_quadratic_form_jac
@@ -362,7 +361,7 @@ def build_quadratic_1d_hess(H, g, s, diag=None, s0=None):
     c : float
         Free term. Returned only if `s0` is provided.
     """
-    a = desc.utils.dot(s)
+    a = H.dot(s)
     if diag is not None:
         a += jnp.dot(s * diag, s)
     a *= 0.5

desc/optimize/tr_subproblems.py

@@ -2,7 +2,6 @@
 
 import numpy as np
 
-import desc.utils
 from desc.backend import (
     cho_factor,
     cho_solve,
@@ -123,7 +122,7 @@ def solve_trust_region_2d_subspace(g, H, trust_radius, initial_alpha=None, **kwa
 
     R, lower = cho_factor(B)
     q1 = -cho_solve((R, lower), g)
-    p1 = desc.utils.dot(q1)
+    p1 = S.dot(q1)
 
     a = B[0, 0] * trust_radius**2
     b = B[0, 1] * trust_radius**2
@@ -140,7 +139,7 @@ def solve_trust_region_2d_subspace(g, H, trust_radius, initial_alpha=None, **kwa
     value = 0.5 * jnp.sum(q2 * B.dot(q2), axis=0) + jnp.dot(g, q2)
     i = jnp.argmin(jnp.where(jnp.isnan(value), jnp.inf, value))
     q2 = q2[:, i]
-    p2 = desc.utils.dot(q2)
+    p2 = S.dot(q2)
 
     out = cond(
         jnp.dot(q1, q1) <= trust_radius**2,

tests/test_derivatives.py

@@ -4,7 +4,6 @@
 import pytest
 from numpy.random import default_rng
 
-import desc.utils
 from desc.backend import jnp
 from desc.derivatives import AutoDiffDerivative, FiniteDiffDerivative
 
@@ -97,7 +96,7 @@ def test_fd_hessian(self):
         g = rando.random(n)
 
         def f(x):
-            return 5 + g.dot(x) + desc.utils.dot(1 / 2 * A.dot(x))
+            return 5 + g.dot(x) + x.dot(1 / 2 * A.dot(x))
 
         hess = FiniteDiffDerivative(f, argnum=0, mode="hess")