ga4-export-fixer 0.5.2-dev.2 → 0.5.2-dev.4

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ga4-export-fixer",
-  "version": "0.5.2-dev.2",
+  "version": "0.5.2-dev.4",
   "description": "",
   "main": "index.js",
   "files": [

package/tables/ga4EventsEnhanced/index.js

@@ -161,6 +161,9 @@ const _generateEnhancedEventsSQL = (mergedConfig) => {
 
     // item list attribution config
     const itemListAttribution = mergedConfig.itemListAttribution;
+    const ecommerceEventsFilter = itemListAttribution
+        ? helpers.ga4EcommerceEvents.filter(e => e !== 'refund').map(e => `'${e}'`).join(', ')
+        : null;
 
     // auto-adjust bufferDays for time-based item list attribution lookback
     const effectiveBufferDays = (itemListAttribution && itemListAttribution.lookbackType === 'TIME')
@@ -225,8 +228,12 @@ const _generateEnhancedEventsSQL = (mergedConfig) => {
             // ecommerce
             ecommerce: helpers.fixEcommerceStruct('ecommerce'),
             items: 'items',
-            // unique row id for item list attribution join
-            _event_row_id: itemListAttribution ? `farm_fingerprint(concat(user_pseudo_id, cast(event_timestamp as string), event_name, to_json_string(items)))` : undefined,
+            // unique row id for item list attribution join. Only computed for ecommerce events.
+            // row_number() breaks hash collisions for batched events with identical data.
+            // partition by event_name avoids a single-partition bottleneck in the window function.
+            // Non-determinism is safe: colliding rows have identical items (to_json_string(items) is in the hash),
+            // so swapping row numbers between them produces the same final result.
+            _item_list_attribution_row_id: itemListAttribution ? `if(event_name in (${ecommerceEventsFilter}), farm_fingerprint(concat(user_pseudo_id, cast(event_timestamp as string), event_name, to_json_string(items), cast(row_number() over(partition by event_name, user_pseudo_id) as string))), null)` : undefined,
             // flag if the data is "final" and is not expected to change anymore
             data_is_final: helpers.isFinalData(mergedConfig.dataIsFinal.detectionMethod, mergedConfig.dataIsFinal.dayThreshold),
             export_type: helpers.getGa4ExportType('_table_suffix'),
@@ -269,12 +276,11 @@ ${excludedEventsSQL}`,
             itemListAttribution.lookbackTimeMs
         );
         const passthroughEvents = `event_name in ('view_item_list', 'select_item', 'view_promotion', 'select_promotion')`;
-        const ecommerceFilter = helpers.ga4EcommerceEvents.filter(e => e !== 'refund').map(e => `'${e}'`).join(', ');
 
         return {
             name: 'item_list_data',
             columns: {
-                '_event_row_id': '_event_row_id',
+                '_item_list_attribution_row_id': '_item_list_attribution_row_id',
                 'items': `array_agg(
       (select as struct item.* replace(
         coalesce(if(${passthroughEvents}, item.item_list_name, _item_list_attr.item_list_name), '(not set)') as item_list_name,
@@ -283,19 +289,19 @@ ${excludedEventsSQL}`,
       ))
     )`,
             },
-            from: `(select _event_row_id, event_name, item, ${attrExpr} as _item_list_attr from event_data, unnest(items) as item where event_name in (${ecommerceFilter}))`,
-            groupBy: ['_event_row_id'],
+            from: `(select _item_list_attribution_row_id, event_name, item, ${attrExpr} as _item_list_attr from event_data, unnest(items) as item where event_name in (${ecommerceEventsFilter}))`,
+            groupBy: ['_item_list_attribution_row_id'],
         };
     })() : null;
 
     const finalColumnOrder = getFinalColumnOrder(eventDataStep, sessionDataStep);
 
-    // When item list attribution is enabled, override the items column and exclude _event_row_id
+    // When item list attribution is enabled, override the items column and exclude _item_list_attribution_row_id
     // COALESCE handles events without items (not in ecommerce filter) where the LEFT JOIN returns NULL
     const itemListOverrides = itemListDataStep ? {
         items: 'coalesce(item_list_data.items, event_data.items)',
     } : {};
-    const itemListExcludedColumns = itemListDataStep ? ['_event_row_id'] : [];
+    const itemListExcludedColumns = itemListDataStep ? ['_item_list_attribution_row_id'] : [];
 
     // Join event_data and session_data, include additional logic
     const finalStep = {
@@ -331,7 +337,7 @@ ${excludedEventsSQL}`,
         leftJoin: [
             ...(itemListDataStep ? [{
                 table: 'item_list_data',
-                condition: 'using(_event_row_id)'
+                condition: 'using(_item_list_attribution_row_id)'
             }] : []),
             {
                 table: 'session_data',