Fix hashing (libAtoms#121)

* Test hashing

* Fix Hasher

* Split up hash tests

Author: ElliottKasoar

abcd/model.py

@@ -12,8 +12,8 @@
 
 
 class Hasher(object):
-    def __init__(self, method=md5()):
-        self.method = method
+    def __init__(self, method=md5):
+        self.method = method()
 
     def update(self, value):
 
@@ -273,26 +273,25 @@ def pre_save(self):
         self["username"] = getpass.getuser()
 
         if not self.get("uploaded"):
-            self["uploaded"] = datetime.datetime.utcnow()
+            self["uploaded"] = datetime.datetime.now(datetime.timezone.utc)
 
-        self["modified"] = datetime.datetime.utcnow()
+        self["modified"] = datetime.datetime.now(datetime.timezone.utc)
 
-        m = Hasher()
+        hasher = Hasher()
 
         for key in ("numbers", "positions", "cell", "pbc"):
-            m.update(self[key])
+            hasher.update(self[key])
 
         self.derived_keys.append("hash_structure")
-        self["hash_structure"] = m()
+        self["hash_structure"] = hasher()
 
-        m = Hasher()
         for key in self.arrays_keys:
-            m.update(self[key])
+            hasher.update(self[key])
         for key in self.info_keys:
-            m.update(self[key])
+            hasher.update(self[key])
 
         self.derived_keys.append("hash")
-        self["hash"] = m()
+        self["hash"] = hasher()
 
 
 if __name__ == "__main__":

tests/test_abstract_model.py

@@ -1,3 +1,4 @@
+import datetime
 import io
 
 import ase
@@ -8,7 +9,7 @@
 from ase.io import read, write
 import numpy as np
 
-from abcd.model import AbstractModel
+from abcd.model import AbstractModel, Hasher
 from ase.calculators.lj import LennardJones
 
 
@@ -238,7 +239,6 @@ def test_write_and_read(store_calc):
         "hash",
         "modified",
         "uploaded",
-        "hash_structure",  # see issue #118
     }:
         assert (
             abcd_data[key] == abcd_data_after_read[key]
@@ -247,10 +247,75 @@ def test_write_and_read(store_calc):
     # expected differences - n.b. order of calls above
     assert abcd_data_after_read["modified"] > abcd_data["modified"]
     assert abcd_data_after_read["uploaded"] > abcd_data["uploaded"]
-    assert abcd_data_after_read["hash"] != abcd_data["hash"]
 
     # expect results to match within fp precision
     for key in set(abcd_data.results_keys):
         assert abcd_data[key] == approx(
             np.array(abcd_data_after_read[key])
         ), f"{key}'s value does not match"
+
+
+def test_hash_update():
+    """Test hash can be updated after initialisation."""
+    hasher_1 = Hasher()
+
+    init_hash = hasher_1()
+    hasher_1.update("Test value")
+    assert hasher_1() != init_hash
+
+
+@pytest.mark.parametrize(
+    "data",
+    [
+        1296,
+        3.14,
+        [1, 2, 3],
+        (4, 5, 6),
+        {"a": "value"},
+        datetime.datetime.now(datetime.timezone.utc),
+        b"test",
+    ],
+)
+def test_hash_data_types(data):
+    """Test updating hash for different data types."""
+    hasher_1 = Hasher()
+    hasher_1.update("Test value")
+    updated_hash = hasher_1()
+
+    hasher_1.update(data)
+    assert updated_hash != hasher_1()
+
+
+def test_second_hash_init():
+    """Test second hash is initialised correctly."""
+    hasher_1 = Hasher()
+
+    init_hash = hasher_1()
+    hasher_1.update("Test value")
+
+    hasher_2 = Hasher()
+    assert hasher_2() == init_hash
+
+
+@pytest.mark.parametrize(
+    "data",
+    [
+        1296,
+        3.14,
+        [1, 2, 3],
+        (4, 5, 6),
+        {"a": "value"},
+        datetime.datetime.now(datetime.timezone.utc),
+        b"test",
+    ],
+)
+def test_consistent_hash(data):
+    """Test two hashers agree with same data."""
+    hasher_1 = Hasher()
+    hasher_1.update("Test value")
+    hasher_1.update(data)
+
+    hasher_2 = Hasher()
+    hasher_2.update("Test value")
+    hasher_2.update(data)
+    assert hasher_1() == hasher_2()