[PERF] account: optimize migration performance by pre-creating column and bulk transferring invoice_date #5465
Conversation
royle-vietnam
force-pushed
the
v17_imp_account_oca
branch
from
b24a98c to
909c789
Compare
and bulk transferring invoice_date Pre-create the target column to avoid triggering compute() during update, and perform a direct bulk data transfer from account_move to account_move_line. This avoids ORM overhead, reduces heap rewrites, and prevents unnecessary recomputation on 15M move lines. Performance result: - Previous approach: ~2 hours - Optimized approach: ~15 minutes (~8x faster)
royle-vietnam
force-pushed
the
v17_imp_account_oca
branch
from
909c789 to
4a07782
Compare
pedrobaeza
left a comment
pedrobaeza
left a comment
There was a problem hiding this comment.
Are you sure this improves the performance? Odoo already does a SQL for related fields with only one level of depth.
Yes, I'm aware that Odoo ORM already optimizes SQL for related fields with one level of depth. In my custom SQL, I explicitly exclude moves where invoice_date is NULL, which reduces the dataset before aggregation. This difference helps avoid unnecessary rows being processed and therefore improves performance in our specific case. So the improvement does not come from bypassing the related field resolution, but from pruning irrelevant records earlier in the query. |
pedrobaeza
left a comment
pedrobaeza
left a comment
There was a problem hiding this comment.
I have proposed odoo/odoo#238812, but meanwhile, let's merge it.
4 participants
Pre-create the target column to avoid triggering compute() during update,
and perform a direct bulk data transfer from account_move to account_move_line.
This avoids ORM overhead, reduces heap rewrites, and prevents unnecessary
recomputation on 15M move lines.
Performance result: