perf(patch): Use INSERT with ON DUPLICATE KEY UPDATE

update_total_qty_field patch was using UPDATE statement with CASE

Now use INSERT INTO with ON DUPLICATE KEY UPDATE clause which
allows this query to use indexes and update multiple rows per query.

erpnext/patches/v11_0/update_total_qty_field.py

@@ -14,7 +14,7 @@ def execute():
 		"Purchase Order", "Purchase Invoice", "Purchase Receipt", "Quotation", "Supplier Quotation"]
 
 	for doctype in doctypes:
-		total_qty =	frappe.db.sql('''
+		total_qty = frappe.db.sql('''
 			SELECT
 				parent, SUM(qty) as qty
 			FROM
@@ -22,14 +22,25 @@ def execute():
 			GROUP BY parent
 		''' % (doctype), as_dict = True)
 
-		when_then = []
+		# Query to update total_qty might become too big, Update in batches
-		for d in total_qty:
+		# batch_size is chosen arbitrarily, Don't try too hard to reason about it
-			when_then.append("""
+		batch_size = 100000
-				when dt.name = '{0}' then {1}
+		for i in range(0, len(total_qty), batch_size):
-			""".format(frappe.db.escape(d.get("parent")), d.get("qty")))
+			batch_transactions = total_qty[i:i + batch_size]
 
-		if when_then:
+			# UPDATE with CASE for some reason cannot use PRIMARY INDEX,
-			frappe.db.sql('''
+			# causing all rows to be examined, leading to a very slow update
-				UPDATE
+
-					`tab%s` dt SET dt.total_qty = CASE %s ELSE 0.0 END
+			# UPDATE with WHERE clause uses PRIMARY INDEX, but will lead to too many queries
-			''' % (doctype, " ".join(when_then)))
+
+			# INSERT with ON DUPLICATE KEY UPDATE uses PRIMARY INDEX
+			# and can perform multiple updates per query
+			# This is probably never used anywhere else as of now, but should be
+			values = []
+			for d in batch_transactions:
+				values.append("('{}', {})".format(d.parent, d.qty))
+			conditions = ",".join(values)
+			frappe.db.sql("""
+				INSERT INTO `tab{}` (name, total_qty) VALUES {}
+				ON DUPLICATE KEY UPDATE name = VALUES(name), total_qty = VALUES(total_qty)
+			""".format(doctype, conditions))