Monday, August 31, 2026
3 changes · 18.0
Enhancements to existing features
The mail module now formats warranty information in a standard JSON structure before sending it to the warranty server. This reduces the risk of rejected or misread warranty data and helps keep server communication reliable.
Original PR description
Format the publisher warranty message using json.dumps() instead of Python's native str() representation. This ensures serializeid JSON serialization for the payload sent to the warranty server. Forward-Port-Of: odoo/odoo#283139
Improves the speed of automatic field prediction on vendor bills, especially for databases with many accounting entries or heavy Peppol usage. This can reduce waiting time when matching invoice lines to products, accounts, or taxes, with the largest bills seeing substantial performance gains.
Original PR description
### Description: Predicting fields can be slow when the database has a large amount of move lines (AML). This is caused by the fact that `_predicted_field` will search on all these AMLs to find the…
### Description: Predicting fields can be slow when the database has a large amount of move lines (AML). This is caused by the fact that `_predicted_field` will search on all these AMLs to find the ones related to the searched fields. It can be an issue when using Peppol since it is used a lot in most localizations to match each line to its product/account/tax. To speed up the queries, the move IDs have been inlined in `_build_predictive_query` to avoid suboptimal execution plans caused by LIMIT and ORDER BY clauses. Additionally, materialization of the `account_move_line` CTE has been removed so the planner can inline filters and stream rows directly, which improves performance in most use cases. ### Benchmark: **For a `history.limit`[^1] of 100:** | Nb of invoice line | Before | After | AMLs scanned | |--------------------|-----------|---------|--------------| | 33 | 7s | 4s | 232 | | 172 | 11.39min | 5min | 99859 | | 1934 | 3h+ [^2] | 1h40min | 102503 | **For a `history.limit`[^1] of 50:** | Nb of invoice line | Before | After | AMLs scanned | |--------------------|---------|---------|--------------| | 33 | 6s | 4s | 187 | | 172 | 2.52min | 1.27min | 21002 | | 1934 | 40min | 20min | 21002 | [^1]: System parameter `account.bill.predict.history.limit` [^2]: Stopped manually after 3h; full runtime not measured ### Reference: opw-5937519
Importing large Italian electronic invoices is now much faster because invoice lines are prepared in memory and saved together at the end. This reduces waiting time for businesses processing high-volume supplier bills, with the biggest gains on very large invoices.
Original PR description
### Description: When importing a large invoice, the process can take a lot of time. This is caused by how `l10n_it_edi` handles the creation and writing of each line of the bill. To improve performance, most of the process is now performed in memory and record creation is deferred to a single batch at the end. Additionally, the check related to `account_accountant` is cached to avoid superfluous calls. ### Benchmark: **For `history.limit`[^1] of 50 (= 21002 `account.move.line`):** | Invoice lines | Before | After | Speedup | |---------------|----------|---------|---------| | 33 | 4.75s | 2.23s | 2.1× | | 172 | 2.3min | 45.1s | 3.1× | | 1,934 | 44.4min | 6.8min | 6.6× | [^1]: System parameter `account.bill.predict.history.limit` ### Reference: opw-5937519