Caching feature and message fingerprints (#11625)

* Caching feature and message fingerprints (#11596)

changelog/11596.improvement.md

@@ -0,0 +1 @@
+Caching `Message` and `Features` fingerprints unless they are altered, saving up to 2/3 of fingerprinting time in our tests.

rasa/shared/nlu/training_data/features.py

@@ -32,6 +32,7 @@ def __init__(
         self.type = feature_type
         self.origin = origin
         self.attribute = attribute
+        self._cached_fingerprint: Optional[Text] = None
         if not self.is_dense() and not self.is_sparse():
             raise ValueError(
                 "Features must either be a numpy array for dense "
@@ -102,6 +103,7 @@ def _combine_dense_features(self, additional_features: Features) -> None:
         self.features = np.concatenate(
             (self.features, additional_features.features), axis=-1
         )
+        self._cached_fingerprint = None
 
     def _combine_sparse_features(self, additional_features: Features) -> None:
         from scipy.sparse import hstack
@@ -114,6 +116,7 @@ def _combine_sparse_features(self, additional_features: Features) -> None:
             )
 
         self.features = hstack([self.features, additional_features.features])
+        self._cached_fingerprint = None
 
     def __key__(
         self,
@@ -148,16 +151,17 @@ def __eq__(self, other: Any) -> bool:
 
     def fingerprint(self) -> Text:
         """Calculate a stable string fingerprint for the features."""
-        if self.is_dense():
-            f_as_text = self.features.tostring()
-        else:
-            f_as_text = rasa.shared.nlu.training_data.util.sparse_matrix_to_string(
-                self.features
+        if self._cached_fingerprint is None:
+            if self.is_dense():
+                f_as_text = self.features.tobytes()
+            else:
+                f_as_text = rasa.shared.nlu.training_data.util.sparse_matrix_to_string(
+                    self.features
+                )
+            self._cached_fingerprint = rasa.shared.utils.io.deep_container_fingerprint(
+                [self.type, self.origin, self.attribute, f_as_text]
             )
-
-        return rasa.shared.utils.io.deep_container_fingerprint(
-            [self.type, self.origin, self.attribute, f_as_text]
-        )
+        return self._cached_fingerprint
 
     @staticmethod
     def filter(

rasa/shared/nlu/training_data/message.py

@@ -54,6 +54,7 @@ def __init__(
         self.features = features if features else []
 
         self.data.update(**kwargs)
+        self._cached_fingerprint: Optional[Text] = None
 
         if output_properties:
             self.output_properties = output_properties
@@ -62,8 +63,10 @@ def __init__(
         self.output_properties.add(TEXT)
 
     def add_features(self, features: Optional["Features"]) -> None:
+        """Add more vectorized features to the message."""
         if features is not None:
             self.features.append(features)
+        self._cached_fingerprint = None
 
     def add_diagnostic_data(self, origin: Text, data: Dict[Text, Any]) -> None:
         """Adds diagnostic data from the `origin` component.
@@ -80,6 +83,7 @@ def add_diagnostic_data(self, origin: Text, data: Dict[Text, Any]) -> None:
             )
         self.data.setdefault(DIAGNOSTIC_DATA, {})
         self.data[DIAGNOSTIC_DATA][origin] = data
+        self._cached_fingerprint = None
 
     def set(self, prop: Text, info: Any, add_to_output: bool = False) -> None:
         """Sets the message's property to the given value.
@@ -92,8 +96,10 @@ def set(self, prop: Text, info: Any, add_to_output: bool = False) -> None:
         self.data[prop] = info
         if add_to_output:
             self.output_properties.add(prop)
+        self._cached_fingerprint = None
 
     def get(self, prop: Text, default: Optional[Any] = None) -> Any:
+        """Retrieve message property."""
         return self.data.get(prop, default)
 
     def as_dict_nlu(self) -> dict:
@@ -143,9 +149,11 @@ def fingerprint(self) -> Text:
         Returns:
             Fingerprint of the message.
         """
-        return rasa.shared.utils.io.deep_container_fingerprint(
-            [self.data, self.features]
-        )
+        if self._cached_fingerprint is None:
+            self._cached_fingerprint = rasa.shared.utils.io.deep_container_fingerprint(
+                [self.data, self.features]
+            )
+        return self._cached_fingerprint
 
     @classmethod
     def build(

tests/shared/nlu/training_data/test_features.py

@@ -37,6 +37,7 @@ def test_combine_with_existing_dense_features():
     existing_features = Features(
         np.array([[1, 0, 2, 3], [2, 0, 0, 1]]), FEATURE_TYPE_SEQUENCE, TEXT, "test"
     )
+    fingerprint = existing_features.fingerprint()
     new_features = Features(
         np.array([[1, 0], [0, 1]]), FEATURE_TYPE_SEQUENCE, TEXT, "origin"
     )
@@ -45,6 +46,8 @@ def test_combine_with_existing_dense_features():
     existing_features.combine_with_features(new_features)
 
     assert np.all(expected_features == existing_features.features)
+    # check that combining features changes fingerprint
+    assert fingerprint != existing_features.fingerprint()
 
 
 def test_combine_with_existing_dense_features_shape_mismatch():
@@ -64,6 +67,7 @@ def test_combine_with_existing_sparse_features():
         TEXT,
         "test",
     )
+    fingerprint = existing_features.fingerprint()
     new_features = Features(
         scipy.sparse.csr_matrix([[1, 0], [0, 1]]), FEATURE_TYPE_SEQUENCE, TEXT, "origin"
     )
@@ -73,6 +77,8 @@ def test_combine_with_existing_sparse_features():
     actual_features = existing_features.features.toarray()
 
     assert np.all(expected_features == actual_features)
+    # check that combining features changes fingerprint
+    assert fingerprint != existing_features.fingerprint()
 
 
 def test_combine_with_existing_sparse_features_shape_mismatch():

tests/shared/nlu/training_data/test_message.py

@@ -408,3 +408,19 @@ def test_message_fingerprint_includes_data_and_features(
     assert fp3 != fp4
 
     assert len({fp1, fp2, fp3, fp4}) == 4
+
+
+def test_message_fingerprint_is_recalculated_after_setting_data():
+    message = Message(data={TEXT: "This is a test sentence."})
+    fp1 = message.fingerprint()
+    message.set(INTENT, "test")
+    fp2 = message.fingerprint()
+    assert fp1 != fp2
+
+
+def test_message_fingerprint_is_recalculated_after_adding_diagnostics_data():
+    message = Message(data={TEXT: "This is a test sentence."})
+    fp1 = message.fingerprint()
+    message.add_diagnostic_data("origin", "test")
+    fp2 = message.fingerprint()
+    assert fp1 != fp2