fix: example captures and taxa count speed up

Author: mihow

ami/main/api/serializers.py

@@ -84,6 +84,19 @@ class Meta:
         ]
 
 
+class ExampleSourceImageNestedSerializer(DefaultSerializer):
+    class Meta:
+        model = SourceImage
+        fields = [
+            "id",
+            "details",
+            "url",
+            "width",
+            "height",
+            "timestamp",
+        ]
+
+
 class DeviceNestedSerializer(DefaultSerializer):
     class Meta:
         model = Device
@@ -252,8 +265,7 @@ class EventListSerializer(DefaultSerializer):
     deployment = DeploymentNestedSerializer(
         read_only=True,
     )
-    example_captures = SourceImageNestedSerializer(many=True, read_only=True)
-    # captures = serializers.StringRelatedField(many=True, read_only=True)
+    example_captures = ExampleSourceImageNestedSerializer(many=True, read_only=True)
     captures = serializers.SerializerMethodField()
 
     class Meta:
@@ -1126,8 +1138,6 @@ class EventSerializer(DefaultSerializer):
     start = serializers.DateTimeField(read_only=True)
     end = serializers.DateTimeField(read_only=True)
     capture_page_offset = serializers.SerializerMethodField()
-    occurrences_count = serializers.SerializerMethodField()
-    taxa_count = serializers.SerializerMethodField()
 
     class Meta:
         model = Event
@@ -1210,18 +1220,6 @@ def get_capture_page_offset(self, obj) -> int | None:
 
         return offset
 
-    def get_occurrences_count(self, obj):
-        # This will display a count of the occurrences that are greater than the current classification threshold
-        # But this value will be different from the occurrences_count field, which is used for sorting & filtering
-        # @TODO refactor how the classification_threshold is used everywhere! delete it!
-        return obj.get_occurrences_count(
-            classification_threshold=get_active_classification_threshold(self.context["request"])
-        )
-
-    def get_taxa_count(self, obj):
-        # @TODO refactor how the classification_threshold is used everywhere! delete it!
-        return obj.taxa_count(classification_threshold=get_active_classification_threshold(self.context["request"]))
-
 
 class EventTimelineSourceImageSerializer(DefaultSerializer):
     class Meta:

ami/main/api/views.py

@@ -242,6 +242,29 @@ def get_queryset(self) -> QuerySet:
             duration=models.F("end") - models.F("start"),
         ).select_related("deployment", "project")
 
+        if self.action == "list":
+            num_example_captures = 1
+            qs = qs.prefetch_related(
+                Prefetch(
+                    "captures",
+                    queryset=SourceImage.objects.order_by("-size").select_related(
+                        "deployment",
+                        "deployment__data_source",
+                    )[:num_example_captures],
+                    to_attr="example_captures",
+                )
+            )
+
+            qs = qs.annotate(
+                taxa_count=models.Count(
+                    "occurrences__determination",
+                    distinct=True,
+                    filter=models.Q(
+                        occurrences__determination_score__gte=get_active_classification_threshold(self.request),
+                    ),
+                ),
+            )
+
         return qs
 
     @action(detail=True, methods=["get"], name="timeline")

ami/main/models.py

@@ -727,17 +727,16 @@ def stats(self) -> dict[str, int | None]:
         )
 
     def taxa_count(self, classification_threshold: int | None = None) -> int:
-        return self.taxa(classification_threshold).count()
+        # Move this to a pre-calculated field or prefetch_related in the view
+        # return self.taxa(classification_threshold).count()
+        return 0
 
     def taxa(self, classification_threshold: int | None = None) -> models.QuerySet["Taxon"]:
         return Taxon.objects.filter(
             Q(occurrences__event=self),
             occurrences__determination_score__gte=classification_threshold or settings.DEFAULT_CONFIDENCE_THRESHOLD,
         ).distinct()
 
-    def example_captures(self, num=5):
-        return SourceImage.objects.filter(event=self).order_by("-size")[:num]
-
     def first_capture(self):
         return SourceImage.objects.filter(event=self).order_by("timestamp").first()
 
@@ -1183,6 +1182,7 @@ def public_url(self) -> str:
 
         @TODO add support for thumbnail URLs here?
         @TODO consider if we ever need to access the original image directly!
+        @TODO every source image request requires joins for the deployment and data source, is this necessary?
         """
         # Get presigned URL if access keys are configured
         data_source = self.deployment.data_source if self.deployment and self.deployment.data_source else None