[IMP] stock: improve performances of _product_reserve by reducing the number of locations that are checked (opw#590466)

bzr revid: mat@openerp.com-20130920112437-giswgxm7o6d17iaf
This commit is contained in:
Martin Trigaux
2013-09-20 13:24:37 +02:00
+10 -1
View File
@@ -399,7 +399,16 @@ class stock_location(osv.osv):
uom_rounding = self.pool.get('product.product').browse(cr, uid, product_id, context=context).uom_id.rounding
if context.get('uom'):
uom_rounding = uom_obj.browse(cr, uid, context.get('uom'), context=context).rounding
for id in self.search(cr, uid, [('location_id', 'child_of', ids)]):
locations_ids = self.search(cr, uid, [('location_id', 'child_of', ids)])
# Fetch only the locations in which this product has ever been processed (in or out)
cr.execute("""SELECT l.id FROM stock_location l WHERE l.id in %s AND
EXISTS (SELECT 1 FROM stock_move m WHERE m.product_id = %s
AND ((state = 'done' AND m.location_dest_id = l.id)
OR (state in ('done','assigned') AND m.location_id = l.id)))
""", (tuple(locations_ids), product_id,))
for id in [i for (i,) in cr.fetchall()]:
if lock:
try:
# Must lock with a separate select query because FOR UPDATE can't be used with