ref(explorer): change transactions ranking (#102966)

Instead of picking the top transactions for a project by volume, pick
them by total time spent. This mirrors what's done on the Insights page
and seems to select the more meaningful parts of the app (rather than
stuff like health checks and middleware for example).

Author: roaga

src/sentry/seer/explorer/index_data.py

@@ -43,7 +43,7 @@ def get_transactions_for_project(
     project_id: int, limit: int = 500, start_time_delta: dict[str, int] | None = None
 ) -> list[Transaction]:
     """
-    Get a list of transactions for a project using EAP, sorted by volume/traffic.
+    Get a list of transactions for a project using EAP, sorted by total time spent.
 
     Args:
         project_id: The ID of the project to fetch transactions for
@@ -75,15 +75,15 @@ def get_transactions_for_project(
         auto_fields=True,
     )
 
-    # Query EAP for transactions with volume metrics
+    # Query EAP for most important transactions (highest total time spent)
     result = Spans.run_table_query(
         params=snuba_params,
-        query_string=f"is_transaction:true project.id:{project_id}",
+        query_string="is_transaction:true",
         selected_columns=[
             "transaction",
-            "count()",
+            "sum(span.duration)",
         ],
-        orderby=["-count()"],  # Sort by count descending (highest volume first)
+        orderby=["-sum(span.duration)"],
         offset=0,
         limit=limit,
         referrer=Referrer.SEER_RPC,

tests/sentry/seer/explorer/test_index_data.py

@@ -23,22 +23,24 @@ def setUp(self) -> None:
 
     def test_get_transactions_for_project(self) -> None:
         """Test the full end-to-end happy path for get_transactions_for_project."""
-        # Create spans for different transactions with varying volumes
+        # Create spans for different transactions with varying total time spent
+        # Format: (transaction_name, count, avg_duration_ms)
         transactions_data = [
-            ("api/users/profile", 5),  # High volume
-            ("api/posts/create", 3),  # Medium volume
-            ("api/health", 1),  # Low volume
+            ("api/users/profile", 5, 100.0),  # 5 * 100 = 500ms total (highest)
+            ("api/posts/create", 3, 150.0),  # 3 * 150 = 450ms total (middle)
+            ("api/health", 10, 10.0),  # 10 * 10 = 100ms total (lowest, despite high count)
         ]
 
-        # Store transaction spans with different volumes
+        # Store transaction spans with different volumes and durations
         spans = []
-        for transaction_name, count in transactions_data:
+        for transaction_name, count, duration_ms in transactions_data:
             for i in range(count):
                 span = self.create_span(
                     {
                         "description": f"transaction-span-{i}",
                         "sentry_tags": {"transaction": transaction_name},
                         "is_segment": True,  # This marks it as a transaction span
+                        "duration_ms": duration_ms,
                     },
                     start_ts=self.ten_mins_ago + timedelta(minutes=i),
                 )
@@ -51,6 +53,7 @@ def test_get_transactions_for_project(self) -> None:
                             "description": f"regular-span-{i}",
                             "sentry_tags": {"transaction": transaction_name},
                             "is_segment": False,  # This marks it as a regular span
+                            "duration_ms": 50.0,
                         },
                         start_ts=self.ten_mins_ago + timedelta(minutes=i, seconds=30),
                     )
@@ -64,11 +67,11 @@ def test_get_transactions_for_project(self) -> None:
         # Verify basic structure and data
         assert len(result) == 3
 
-        # Should be sorted by volume (count) descending - only transaction spans count
+        # Should be sorted by total time spent (sum of duration) descending
         transaction_names = [t.name for t in result]
-        assert transaction_names[0] == "api/users/profile"  # Highest count (5 transaction spans)
-        assert transaction_names[1] == "api/posts/create"  # Medium count (3 transaction spans)
-        assert transaction_names[2] == "api/health"  # Lowest count (1 transaction span)
+        assert transaction_names[0] == "api/users/profile"  # 500ms total (highest)
+        assert transaction_names[1] == "api/posts/create"  # 450ms total (middle)
+        assert transaction_names[2] == "api/health"  # 100ms total (lowest despite high count)
 
         # Verify all transactions have correct project_id and structure
         for transaction in result: