bug fix: double counting water in Water bridge analysis (#3120)

Fixes #3119

* Fixes bug linked with double counting of waters in WaterBridgeAnalysis

package/CHANGELOG

@@ -23,6 +23,7 @@ The rules for this file:
   * 2.0.0
 
 Fixes
+  * WaterBridgeAnalysis double counts the water (Issue #3119, PR #3120)
   * NCDFReader now defaults to a dt value of 1.0 ps when it cannot estimate
     it from the first two frames of the file (Issue #3166)
   * Get chi1_selection to match canonical gamma atoms (Issue #2044)

package/MDAnalysis/analysis/hydrogenbonds/wbridge_analysis.py

@@ -1349,7 +1349,8 @@ def traverse_water_network(graph, node, end, route, maxdepth, result):
             else:
                 if node in end:
                     # check if any duplication happens
-                    if len(route) == len(set(route)):
+                    heavy_atom = [line[3] or line[2] for line in route]
+                    if len(heavy_atom) == len(set(heavy_atom)):
                         add_route(result, route)
                 else:
                     for new_node in graph[node]:

testsuite/MDAnalysisTests/analysis/test_wbridge.py

@@ -723,3 +723,26 @@ def test_timesteps_by_type(self, wb_multiframe):
 
         timesteps = sorted(wb_multiframe.timesteps_by_type())
         assert_array_equal(timesteps[3], [1, 12, 'ALA', 1, 'H', 'ALA', 6, 'O', 0, 2])
+
+    def test_duplicate_water(self):
+        '''A case #3119 where
+        Acceptor···H−O···H-Donor
+                     |
+                     H···O-H
+        will be recognised as 3rd order water bridge.
+        '''
+        grofile = '''Test gro file
+    7
+    1LEU      O    1   1.876   0.810   1.354
+  117SOL    HW1    2   1.853   0.831   1.162
+  117SOL     OW    3   1.877   0.890   1.081
+  117SOL    HW2    4   1.908   0.828   1.007
+  135SOL     OW    5   1.924   0.713   0.845
+    1LEU      H    6   1.997   0.991   1.194
+    1LEU      N    7   2.041   1.030   1.274
+   2.22092   2.22092   2.22092'''
+        u = MDAnalysis.Universe(StringIO(grofile), format='gro')
+        wb = WaterBridgeAnalysis(u, 'resname LEU and name O',
+                                 'resname LEU and name N H', order=4)
+        wb.run()
+        assert len(wb.timeseries[0]) == 2