perf: GLE reposting with progress and chunking (#31343)

If stock voucher count goes >1000 then fetching all gles and reposting
them all at once requires much more memory and can cause crash.

- This PR ensures that GLE reposting is done in chunks of 100 vouchers.
- This PR also starts keeping track of how many such chunks were
  processed so in future progress is resumed in event of timeout.

erpnext/accounts/test/test_utils.py

@@ -62,8 +62,8 @@ class TestUtils(unittest.TestCase):
 		stock_entry = {"item": item, "to_warehouse": "_Test Warehouse - _TC", "qty": 1, "rate": 10}
 
 		se1 = make_stock_entry(posting_date="2022-01-01", **stock_entry)
-		se2 = make_stock_entry(posting_date="2022-02-01", **stock_entry)
 		se3 = make_stock_entry(posting_date="2022-03-01", **stock_entry)
+		se2 = make_stock_entry(posting_date="2022-02-01", **stock_entry)
 
 		for doc in (se1, se2, se3):
 			vouchers.append((doc.doctype, doc.name))

erpnext/accounts/utils.py

@@ -2,8 +2,9 @@
 # License: GNU General Public License v3. See license.txt
 
 
+import itertools
 from json import loads
-from typing import List, Tuple
+from typing import TYPE_CHECKING, List, Optional, Tuple
 
 import frappe
 import frappe.defaults
@@ -22,6 +23,9 @@ from erpnext.accounts.doctype.accounting_dimension.accounting_dimension import g
 from erpnext.stock import get_warehouse_account_map
 from erpnext.stock.utils import get_stock_value_on
 
+if TYPE_CHECKING:
+	from erpnext.stock.doctype.repost_item_valuation.repost_item_valuation import RepostItemValuation
+
 
 class FiscalYearError(frappe.ValidationError):
 	pass
@@ -31,6 +35,9 @@ class PaymentEntryUnlinkError(frappe.ValidationError):
 	pass
 
 
+GL_REPOSTING_CHUNK = 100
+
+
 @frappe.whitelist()
 def get_fiscal_year(
 	date=None, fiscal_year=None, label="Date", verbose=1, company=None, as_dict=False
@@ -1120,7 +1127,11 @@ def update_gl_entries_after(
 
 
 def repost_gle_for_stock_vouchers(
-	stock_vouchers, posting_date, company=None, warehouse_account=None
+	stock_vouchers: List[Tuple[str, str]],
+	posting_date: str,
+	company: Optional[str] = None,
+	warehouse_account=None,
+	repost_doc: Optional["RepostItemValuation"] = None,
 ):
 
 	from erpnext.accounts.general_ledger import toggle_debit_credit_if_negative
@@ -1128,22 +1139,22 @@ def repost_gle_for_stock_vouchers(
 	if not stock_vouchers:
 		return
 
-	def _delete_gl_entries(voucher_type, voucher_no):
-		frappe.db.sql(
-			"""delete from `tabGL Entry`
-			where voucher_type=%s and voucher_no=%s""",
-			(voucher_type, voucher_no),
-		)
-
-	stock_vouchers = sort_stock_vouchers_by_posting_date(stock_vouchers)
-
 	if not warehouse_account:
 		warehouse_account = get_warehouse_account_map(company)
 
+	stock_vouchers = sort_stock_vouchers_by_posting_date(stock_vouchers)
+	if repost_doc and repost_doc.gl_reposting_index:
+		# Restore progress
+		stock_vouchers = stock_vouchers[cint(repost_doc.gl_reposting_index) :]
+
 	precision = get_field_precision(frappe.get_meta("GL Entry").get_field("debit")) or 2
 
-	gle = get_voucherwise_gl_entries(stock_vouchers, posting_date)
-	for idx, (voucher_type, voucher_no) in enumerate(stock_vouchers):
+	stock_vouchers_iterator = iter(stock_vouchers)
+
+	while stock_vouchers_chunk := list(itertools.islice(stock_vouchers_iterator, GL_REPOSTING_CHUNK)):
+		gle = get_voucherwise_gl_entries(stock_vouchers_chunk, posting_date)
+
+		for voucher_type, voucher_no in stock_vouchers_chunk:
 			existing_gle = gle.get((voucher_type, voucher_no), [])
 			voucher_obj = frappe.get_doc(voucher_type, voucher_no)
 			# Some transactions post credit as negative debit, this is handled while posting GLE
@@ -1157,12 +1168,22 @@ def repost_gle_for_stock_vouchers(
 					voucher_obj.make_gl_entries(gl_entries=expected_gle, from_repost=True)
 			else:
 				_delete_gl_entries(voucher_type, voucher_no)
-
-		if idx % 20 == 0:
-			# Commit every 20 documents to avoid losing progress
-			# and reducing memory usage
 		frappe.db.commit()
 
+		if repost_doc:
+			repost_doc.db_set(
+				"gl_reposting_index",
+				cint(repost_doc.gl_reposting_index) + GL_REPOSTING_CHUNK,
+			)
+
+
+def _delete_gl_entries(voucher_type, voucher_no):
+	frappe.db.sql(
+		"""delete from `tabGL Entry`
+		where voucher_type=%s and voucher_no=%s""",
+		(voucher_type, voucher_no),
+	)
+
 
 def sort_stock_vouchers_by_posting_date(
 	stock_vouchers: List[Tuple[str, str]]
@@ -1175,6 +1196,9 @@ def sort_stock_vouchers_by_posting_date(
 		.select(sle.voucher_type, sle.voucher_no, sle.posting_date, sle.posting_time, sle.creation)
 		.where((sle.is_cancelled == 0) & (sle.voucher_no.isin(voucher_nos)))
 		.groupby(sle.voucher_type, sle.voucher_no)
+		.orderby(sle.posting_date)
+		.orderby(sle.posting_time)
+		.orderby(sle.creation)
 	).run(as_dict=True)
 	sorted_vouchers = [(sle.voucher_type, sle.voucher_no) for sle in sles]
 

erpnext/stock/doctype/repost_item_valuation/repost_item_valuation.json

@@ -25,7 +25,8 @@
   "items_to_be_repost",
   "affected_transactions",
   "distinct_item_and_warehouse",
-  "current_index"
+  "current_index",
+  "gl_reposting_index"
  ],
  "fields": [
   {
@@ -181,12 +182,20 @@
    "label": "Affected Transactions",
    "no_copy": 1,
    "read_only": 1
+  },
+  {
+   "default": "0",
+   "fieldname": "gl_reposting_index",
+   "fieldtype": "Int",
+   "hidden": 1,
+   "label": "GL reposting index",
+   "read_only": 1
   }
  ],
  "index_web_pages_for_search": 1,
  "is_submittable": 1,
  "links": [],
- "modified": "2022-04-18 14:08:08.821602",
+ "modified": "2022-06-13 12:20:22.182322",
  "modified_by": "Administrator",
  "module": "Stock",
  "name": "Repost Item Valuation",

erpnext/stock/doctype/repost_item_valuation/repost_item_valuation.py

@@ -192,6 +192,7 @@ def repost_gl_entries(doc):
 		directly_dependent_transactions + list(repost_affected_transaction),
 		doc.posting_date,
 		doc.company,
+		repost_doc=doc,
 	)
 
 

erpnext/stock/doctype/repost_item_valuation/test_repost_item_valuation.py

@@ -2,10 +2,14 @@
 # See license.txt
 
 
+from unittest.mock import MagicMock, call
+
 import frappe
 from frappe.tests.utils import FrappeTestCase
 from frappe.utils import nowdate
+from frappe.utils.data import today
 
+from erpnext.accounts.utils import repost_gle_for_stock_vouchers
 from erpnext.controllers.stock_controller import create_item_wise_repost_entries
 from erpnext.stock.doctype.item.test_item import make_item
 from erpnext.stock.doctype.purchase_receipt.test_purchase_receipt import make_purchase_receipt
@@ -193,3 +197,31 @@ class TestRepostItemValuation(FrappeTestCase):
 			[["a", "b"], ["c", "d"]],
 			sorted(frappe.parse_json(frappe.as_json(set([("a", "b"), ("c", "d")])))),
 		)
+
+	def test_gl_repost_progress(self):
+		from erpnext.accounts import utils
+
+		# lower numbers to simplify test
+		orig_chunk_size = utils.GL_REPOSTING_CHUNK
+		utils.GL_REPOSTING_CHUNK = 1
+		self.addCleanup(setattr, utils, "GL_REPOSTING_CHUNK", orig_chunk_size)
+
+		doc = frappe.new_doc("Repost Item Valuation")
+		doc.db_set = MagicMock()
+
+		vouchers = []
+		company = "_Test Company with perpetual inventory"
+		posting_date = today()
+
+		for _ in range(3):
+			se = make_stock_entry(company=company, qty=1, rate=2, target="Stores - TCP1")
+			vouchers.append((se.doctype, se.name))
+
+		repost_gle_for_stock_vouchers(stock_vouchers=vouchers, posting_date=posting_date, repost_doc=doc)
+		self.assertIn(call("gl_reposting_index", 1), doc.db_set.mock_calls)
+		doc.db_set.reset_mock()
+
+		doc.gl_reposting_index = 1
+		repost_gle_for_stock_vouchers(stock_vouchers=vouchers, posting_date=posting_date, repost_doc=doc)
+
+		self.assertNotIn(call("gl_reposting_index", 1), doc.db_set.mock_calls)