Speed up Cube.subset/Coord.intersect (#4955)

* Speed up Cube.subset/Coord.intersect

* Add a whatsnew entry

Author: bouweandela

docs/src/whatsnew/latest.rst

@@ -82,6 +82,10 @@ This document explains the changes made to Iris for this release
    :meth:`iris.cube.Cube.subset`, and :meth:`iris.coords.Coord.intersect`.
    (:pull:`4969`)
 
+#. `@bouweandela`_ improved the speed of :meth:`iris.cube.Cube.subset` /
+   :meth:`iris.coords.Coord.intersect`.
+   (:pull:`4955`)
+
 🔥 Deprecations
 ===============
 

lib/iris/coords.py

@@ -1344,7 +1344,14 @@ def __add__(self, mod):
         return Cell(point, bound)
 
     def __hash__(self):
-        return super().__hash__()
+        # See __eq__ for the definition of when two cells are equal.
+        if self.bound is None:
+            return hash(self.point)
+        bound = self.bound
+        rbound = bound[::-1]
+        if rbound < bound:
+            bound = rbound
+        return hash((self.point, bound))
 
     def __eq__(self, other):
         """
@@ -2374,18 +2381,16 @@ def intersect(self, other, return_indices=False):
             )
             raise ValueError(msg)
 
-        # Cache self.cells for speed. We can also use the index operation on a
-        # list conveniently.
-        self_cells = [cell for cell in self.cells()]
+        # Cache self.cells for speed. We can also use the dict for fast index
+        # lookup.
+        self_cells = {cell: idx for idx, cell in enumerate(self.cells())}
 
         # Maintain a list of indices on self for which cells exist in both self
         # and other.
         self_intersect_indices = []
         for cell in other.cells():
-            try:
-                self_intersect_indices.append(self_cells.index(cell))
-            except ValueError:
-                pass
+            if cell in self_cells:
+                self_intersect_indices.append(self_cells[cell])
 
         if return_indices is False and self_intersect_indices == []:
             raise ValueError(

lib/iris/tests/test_cell.py

@@ -169,9 +169,35 @@ def test_coord_bounds_cmp(self):
         self.assertTrue(self.e < 2)
 
     def test_cell_cell_cmp(self):
+        self.e = iris.coords.Cell(1)
+        self.f = iris.coords.Cell(1)
+
+        self.assertTrue(self.e == self.f)
+        self.assertEqual(hash(self.e), hash(self.f))
+
+        self.e = iris.coords.Cell(1)
+        self.f = iris.coords.Cell(1, [0, 2])
+
+        self.assertFalse(self.e == self.f)
+        self.assertNotEqual(hash(self.e), hash(self.f))
+
+        self.e = iris.coords.Cell(1, [0, 2])
+        self.f = iris.coords.Cell(1, [0, 2])
+
+        self.assertTrue(self.e == self.f)
+        self.assertEqual(hash(self.e), hash(self.f))
+
+        self.e = iris.coords.Cell(1, [0, 2])
+        self.f = iris.coords.Cell(1, [2, 0])
+
+        self.assertTrue(self.e == self.f)
+        self.assertEqual(hash(self.e), hash(self.f))
+
         self.e = iris.coords.Cell(0.7, [1.1, 1.9])
         self.f = iris.coords.Cell(0.8, [1.1, 1.9])
 
+        self.assertFalse(self.e == self.f)
+        self.assertNotEqual(hash(self.e), hash(self.f))
         self.assertFalse(self.e > self.f)
         self.assertTrue(self.e <= self.f)
         self.assertTrue(self.f >= self.e)