backend: Deduplicate hosts returned by get_backends() (#572)

If a host appears multiple times in host specs (which can typically
happen if a host is a member of multiple Ansible groups for instance),
testinfra would run tests on that host multiple times.

If a host finds itself listed multiple times with the same backend and
connection parameters, testinfra will run the tests on that host only
once.

Signed-off-by: Benoît Knecht <[email protected]>
Co-Authored-By: Philippe Pepiot <[email protected]>

Author: BenoitKnecht

test/test_backends.py

@@ -519,3 +519,16 @@ def test_ssh_hostspec(hostspec, expected):
     cmd, cmd_args = backend._build_ssh_command('true')
     command = backend.quote(' '.join(cmd), *cmd_args)
     assert command == expected
+
+
+def test_get_hosts():
+    # Hosts returned by get_host must be deduplicated (by name & kwargs) and in
+    # same order as asked
+    hosts = testinfra.backend.get_backends([
+        'ssh://foo', 'ssh://a', 'ssh://a', 'ssh://a?timeout=1',
+    ])
+    assert [(h.host.name, h.timeout) for h in hosts] == [
+        ('foo', 10),
+        ('a', 10),
+        ('a', 1),
+    ]

testinfra/backend/__init__.py

@@ -77,7 +77,7 @@ def get_backend(hostspec, **kwargs):
 
 
 def get_backends(hosts, **kwargs):
-    backends = []
+    backends = {}
     for hostspec in hosts:
         host, kw = parse_hostspec(hostspec)
         for k, v in kwargs.items():
@@ -89,8 +89,12 @@ def get_backends(hosts, **kwargs):
             connection = "paramiko"
         klass = get_backend_class(connection)
         for name in klass.get_hosts(host, **kw):
+            key = (name, frozenset(kw.items()))
+            if key in backends:
+                continue
             if connection == "local":
-                backends.append(klass(**kw))
+                backend = klass(**kw)
             else:
-                backends.append(klass(name, **kw))
-    return backends
+                backend = klass(name, **kw)
+            backends[key] = backend
+    return list(backends.values())