From 2d95075ee55c97c963729ed036b43a250ee871df Mon Sep 17 00:00:00 2001
From: Till Hoffmann <tillahoffmann@gmail.com>
Date: Fri, 23 Feb 2024 09:24:09 -0500
Subject: [PATCH] Promote `isclose` arguments to inexact dtype unless extended
 (fixes #19935).

---
 jax/_src/numpy/lax_numpy.py | 64 +++++++++++++++++++------------------
 tests/lax_numpy_test.py     |  4 +++
 2 files changed, 37 insertions(+), 31 deletions(-)

diff --git a/jax/_src/numpy/lax_numpy.py b/jax/_src/numpy/lax_numpy.py
index 1bbce5b7273c..8069f896b8d5 100644
--- a/jax/_src/numpy/lax_numpy.py
+++ b/jax/_src/numpy/lax_numpy.py
@@ -940,39 +940,41 @@ def isclose(a: ArrayLike, b: ArrayLike, rtol: ArrayLike = 1e-05, atol: ArrayLike
             equal_nan: bool = False) -> Array:
   a, b = util.promote_args("isclose", a, b)
   dtype = _dtype(a)
-  if issubdtype(dtype, inexact):
-    if issubdtype(dtype, complexfloating):
-      dtype = util._complex_elem_type(dtype)
-    rtol = lax.convert_element_type(rtol, dtype)
-    atol = lax.convert_element_type(atol, dtype)
-    out = lax.le(
-      lax.abs(lax.sub(a, b)),
-      lax.add(atol, lax.mul(rtol, lax.abs(b))))
-    # This corrects the comparisons for infinite and nan values
-    a_inf = ufuncs.isinf(a)
-    b_inf = ufuncs.isinf(b)
-    any_inf = ufuncs.logical_or(a_inf, b_inf)
-    both_inf = ufuncs.logical_and(a_inf, b_inf)
-    # Make all elements where either a or b are infinite to False
-    out = ufuncs.logical_and(out, ufuncs.logical_not(any_inf))
-    # Make all elements where both a or b are the same inf to True
-    same_value = lax.eq(a, b)
-    same_inf = ufuncs.logical_and(both_inf, same_value)
-    out = ufuncs.logical_or(out, same_inf)
-
-    # Make all elements where either a or b is NaN to False
-    a_nan = ufuncs.isnan(a)
-    b_nan = ufuncs.isnan(b)
-    any_nan = ufuncs.logical_or(a_nan, b_nan)
-    out = ufuncs.logical_and(out, ufuncs.logical_not(any_nan))
-    if equal_nan:
-      # Make all elements where both a and b is NaN to True
-      both_nan = ufuncs.logical_and(a_nan, b_nan)
-      out = ufuncs.logical_or(out, both_nan)
-    return out
-  else:
+  if dtypes.issubdtype(dtype, dtypes.extended):
     return lax.eq(a, b)
 
+  a, b = util.promote_args_inexact("isclose", a, b)
+  dtype = _dtype(a)
+  if issubdtype(dtype, complexfloating):
+    dtype = util._complex_elem_type(dtype)
+  rtol = lax.convert_element_type(rtol, dtype)
+  atol = lax.convert_element_type(atol, dtype)
+  out = lax.le(
+    lax.abs(lax.sub(a, b)),
+    lax.add(atol, lax.mul(rtol, lax.abs(b))))
+  # This corrects the comparisons for infinite and nan values
+  a_inf = ufuncs.isinf(a)
+  b_inf = ufuncs.isinf(b)
+  any_inf = ufuncs.logical_or(a_inf, b_inf)
+  both_inf = ufuncs.logical_and(a_inf, b_inf)
+  # Make all elements where either a or b are infinite to False
+  out = ufuncs.logical_and(out, ufuncs.logical_not(any_inf))
+  # Make all elements where both a or b are the same inf to True
+  same_value = lax.eq(a, b)
+  same_inf = ufuncs.logical_and(both_inf, same_value)
+  out = ufuncs.logical_or(out, same_inf)
+
+  # Make all elements where either a or b is NaN to False
+  a_nan = ufuncs.isnan(a)
+  b_nan = ufuncs.isnan(b)
+  any_nan = ufuncs.logical_or(a_nan, b_nan)
+  out = ufuncs.logical_and(out, ufuncs.logical_not(any_nan))
+  if equal_nan:
+    # Make all elements where both a and b is NaN to True
+    both_nan = ufuncs.logical_and(a_nan, b_nan)
+    out = ufuncs.logical_or(out, both_nan)
+  return out
+
 
 def _interp(x: ArrayLike, xp: ArrayLike, fp: ArrayLike,
            left: ArrayLike | str | None = None,
diff --git a/tests/lax_numpy_test.py b/tests/lax_numpy_test.py
index 586467db8502..1fd10ddc9749 100644
--- a/tests/lax_numpy_test.py
+++ b/tests/lax_numpy_test.py
@@ -3532,6 +3532,10 @@ def testIsClose(self):
         self.assertTrue(jnp.all(jnp.equal(result_np, result_jax)))
         self.assertTrue(jnp.all(jnp.equal(result_np, result_jit)))
 
+    self.assertEqual(np.isclose(6, 10, rtol=0.5), jnp.isclose(6, 10, rtol=0.5))
+    key = jax.random.key(0)
+    self.assertTrue(jnp.isclose(key, key))
+
   @jtu.sample_product(
     x=[1, [1], [1, 1 + 1E-4], [1, np.nan]],
     y=[1, [1], [1, 1 + 1E-4], [1, np.nan]],