feat(product-analytics): Use sessions table for session duration

posthog/hogql/database/schema/events.py

@@ -11,15 +11,12 @@
     FieldTraverser,
     FieldOrTable,
 )
-from posthog.hogql.database.schema.event_sessions import (
-    EventsSessionSubTable,
-    join_with_events_table_session_duration,
-)
 from posthog.hogql.database.schema.groups import GroupsTable, join_with_group_n_table
 from posthog.hogql.database.schema.person_distinct_ids import (
     PersonDistinctIdsTable,
     join_with_person_distinct_ids_table,
 )
+from posthog.hogql.database.schema.sessions import join_events_table_to_sessions_table, SessionsTable
 
 
 class EventsPersonSubTable(VirtualTable):
@@ -116,8 +113,8 @@ class EventsTable(Table):
         ),
         "session": LazyJoin(
             from_field=["$session_id"],
-            join_table=EventsSessionSubTable(),
-            join_function=join_with_events_table_session_duration,
+            join_table=SessionsTable(),
+            join_function=join_events_table_to_sessions_table,
         ),
     }
 

posthog/hogql/database/schema/sessions.py

@@ -1,4 +1,4 @@
-from typing import Dict, List, cast
+from typing import Dict, List, cast, Any
 
 from posthog.hogql import ast
 from posthog.hogql.context import HogQLContext
@@ -14,9 +14,12 @@
 )
 from posthog.hogql.database.schema.channel_type import create_channel_type_expr
 from posthog.hogql.database.schema.util.session_where_clause_extractor import SessionMinTimestampWhereClauseExtractor
-
+from posthog.hogql.errors import HogQLException
 
 SESSIONS_COMMON_FIELDS: Dict[str, FieldOrTable] = {
+    "id": StringDatabaseField(
+        name="session_id"
+    ),  # TODO remove this, it's a duplicate of the correct session_id field below to get some trends working on a deadline
     "session_id": StringDatabaseField(name="session_id"),
     "team_id": IntegerDatabaseField(name="team_id"),
     "distinct_id": StringDatabaseField(name="distinct_id"),
@@ -71,6 +74,10 @@ def select_from_sessions_table(
 
     table_name = "raw_sessions"
 
+    # Always include "session_id", as it's the key we use to make further joins, and it'd be great if it's available
+    if "session_id" not in requested_fields:
+        requested_fields = {**requested_fields, "session_id": ["session_id"]}
+
     aggregate_fields = {
         "distinct_id": ast.Call(name="any", args=[ast.Field(chain=[table_name, "distinct_id"])]),
         "min_timestamp": ast.Call(name="min", args=[ast.Field(chain=[table_name, "min_timestamp"])]),
@@ -163,3 +170,24 @@ def to_printed_clickhouse(self, context):
 
     def to_printed_hogql(self):
         return "sessions"
+
+
+def join_events_table_to_sessions_table(
+    from_table: str, to_table: str, requested_fields: Dict[str, Any], context: HogQLContext, node: ast.SelectQuery
+) -> ast.JoinExpr:
+    from posthog.hogql import ast
+
+    if not requested_fields:
+        raise HogQLException("No fields requested from events")
+
+    join_expr = ast.JoinExpr(table=select_from_sessions_table(requested_fields, node, context))
+    join_expr.join_type = "LEFT JOIN"
+    join_expr.alias = to_table
+    join_expr.constraint = ast.JoinConstraint(
+        expr=ast.CompareOperation(
+            op=ast.CompareOperationOp.Eq,
+            left=ast.Field(chain=[from_table, "$session_id"]),
+            right=ast.Field(chain=[to_table, "session_id"]),
+        )
+    )
+    return join_expr

posthog/hogql/database/schema/test/test_sessions.py

@@ -55,3 +55,51 @@ def test_channel_type(self):
             result[0],
             "Paid Search",
         )
+
+    def test_event_dot_session_dot_channel_type(self):
+        session_id = "event_dot_session_dot_channel_type"
+
+        _create_event(
+            event="$pageview",
+            team=self.team,
+            distinct_id="d1",
+            properties={"gad_source": "1", "$session_id": session_id},
+        )
+
+        response = execute_hogql_query(
+            parse_select(
+                "select events.session.channel_type from events where $session_id = {session_id}",
+                placeholders={"session_id": ast.Constant(value=session_id)},
+            ),
+            self.team,
+        )
+
+        result = (response.results or [])[0]
+        self.assertEqual(
+            result[0],
+            "Paid Search",
+        )
+
+    def test_events_session_dot_channel_type(self):
+        session_id = "event_dot_session_dot_channel_type"
+
+        _create_event(
+            event="$pageview",
+            team=self.team,
+            distinct_id="d1",
+            properties={"gad_source": "1", "$session_id": session_id},
+        )
+
+        response = execute_hogql_query(
+            parse_select(
+                "select session.channel_type from events where $session_id = {session_id}",
+                placeholders={"session_id": ast.Constant(value=session_id)},
+            ),
+            self.team,
+        )
+
+        result = (response.results or [])[0]
+        self.assertEqual(
+            result[0],
+            "Paid Search",
+        )

posthog/hogql/database/test/__snapshots__/test_database.ambr

@@ -214,10 +214,30 @@
           {
               "key": "session",
               "type": "lazy_table",
-              "table": "events",
+              "table": "sessions",
               "fields": [
                   "id",
-                  "duration"
+                  "session_id",
+                  "team_id",
+                  "distinct_id",
+                  "min_timestamp",
+                  "max_timestamp",
+                  "urls",
+                  "entry_url",
+                  "exit_url",
+                  "initial_utm_source",
+                  "initial_utm_campaign",
+                  "initial_utm_medium",
+                  "initial_utm_term",
+                  "initial_utm_content",
+                  "initial_referring_domain",
+                  "initial_gclid",
+                  "initial_gad_source",
+                  "event_count_map",
+                  "pageview_count",
+                  "autocapture_count",
+                  "duration",
+                  "channel_type"
               ]
           }
       ],
@@ -569,6 +589,10 @@
           }
       ],
       "sessions": [
+          {
+              "key": "id",
+              "type": "string"
+          },
           {
               "key": "session_id",
               "type": "string"
@@ -875,6 +899,10 @@
           }
       ],
       "raw_sessions": [
+          {
+              "key": "id",
+              "type": "string"
+          },
           {
               "key": "session_id",
               "type": "string"
@@ -1123,10 +1151,30 @@
           {
               "key": "session",
               "type": "lazy_table",
-              "table": "events",
+              "table": "sessions",
               "fields": [
                   "id",
-                  "duration"
+                  "session_id",
+                  "team_id",
+                  "distinct_id",
+                  "min_timestamp",
+                  "max_timestamp",
+                  "urls",
+                  "entry_url",
+                  "exit_url",
+                  "initial_utm_source",
+                  "initial_utm_campaign",
+                  "initial_utm_medium",
+                  "initial_utm_term",
+                  "initial_utm_content",
+                  "initial_referring_domain",
+                  "initial_gclid",
+                  "initial_gad_source",
+                  "event_count_map",
+                  "pageview_count",
+                  "autocapture_count",
+                  "duration",
+                  "channel_type"
               ]
           }
       ],
@@ -1478,6 +1526,10 @@
           }
       ],
       "sessions": [
+          {
+              "key": "id",
+              "type": "string"
+          },
           {
               "key": "session_id",
               "type": "string"
@@ -1784,6 +1836,10 @@
           }
       ],
       "raw_sessions": [
+          {
+              "key": "id",
+              "type": "string"
+          },
           {
               "key": "session_id",
               "type": "string"

posthog/hogql/test/test_query.py

@@ -1452,7 +1452,9 @@ def test_events_sessions_table(self):
                 properties={"$session_id": random_uuid},
             )
 
-        query = "SELECT session.id, session.duration from events WHERE distinct_id={distinct_id} order by timestamp"
+        query = (
+            "SELECT session.session_id, session.duration from events WHERE distinct_id={distinct_id} order by timestamp"
+        )
         response = execute_hogql_query(
             query, team=self.team, placeholders={"distinct_id": ast.Constant(value=random_uuid)}
         )