Odoo Changelog — 2026-09-03

Thursday, September 3, 2026

2 changes · saas-18.3

Improvements(2)

Enhancements to existing features

[PERF] l10n_co_dian: avoid stored field computation during installation

Enterprisesaas-18.3l10n_co_dianl10n_co_ediMedium

Installing the Colombian DIAN e-invoicing module is now faster on large databases. The change avoids a lengthy setup-time recalculation for existing accounting entries while keeping normal behavior for future records unchanged.

Original PR description

- Pre-create the stored computed columns `l10n_co_edi_type`, `l10n_co_dian_state`, and `l10n_co_edi_cufe_cude_ref` in `_auto_init()`.

- This prevents Odoo from computing and writing these fields for all existing `account.move` records when installing `l10n_co_dian`.

- This is particularly important for large databases with a high volume of Colombian accounting moves, where the initial computation can take too long and cause the module installation to hit the time limit.

- The compute methods are kept unchanged, so the fields continue to be computed normally for subsequent record creation or dependency changes.

**opw-6451331**

Forward-Port-Of: odoo/enterprise#129507

#performance#installation#colombia#e-invoicing

[PERF] account_reports: avoid redundant distinct count in domain engine

PR #129048

Enterprisesaas-18.3account_reportsMedium

Financial reports now avoid an unnecessary counting step in common queries, making large Balance Sheet reports run much faster. This improves responsiveness for companies with high accounting transaction volumes while keeping report results unchanged.

Original PR description

On standard AML-backed domain-engine queries, `id` is unique.
`COUNT(DISTINCT id)` is therefore redundant and prevents PostgreSQL
from using a partial hash aggregation.

Analytic-groupby and cash-basis modes retain `DISTINCT`, as their
temporary AML relations can contain duplicate AML IDs.

Benchmark on the Balance Sheet query, 5 runs per size:

| Matching AMLs | Before | After |
|---:|---:|---:|
| 2.3M | 3.19 s | 2.06 s |
| 5.8M | 6.51 s | 2.91 s |
| 11.5M | 9.78 s | 3.64 s |
| 17.2M | 15.63 s | 4.67 s |
| 23.0M | 16.01 s | 4.68 s |

At 23M AMLs, execution time drops from 16.01 s to 4.68 s
(3.42x). Result sets are identical before and after.

opw-6459372

Forward-Port-Of: odoo/enterprise#128404

#performance#accounting#reports

- Pre-create the stored computed columns `l10n_co_edi_type`, `l10n_co_dian_state`, and `l10n_co_edi_cufe_cude_ref` in `_auto_init()`. - This prevents Odoo from computing and writing these fields for all existing `account.move` records when installing `l10n_co_dian`. - This is particularly important for large databases with a high volume of Colombian accounting moves, where the initial computation can take too long and cause the module installation to hit the time limit. - The compute methods are kept unchanged, so the fields continue to be computed normally for subsequent record creation or dependency changes. **opw-6451331** Forward-Port-Of: odoo/enterprise#129507

On standard AML-backed domain-engine queries, `id` is unique. `COUNT(DISTINCT id)` is therefore redundant and prevents PostgreSQL from using a partial hash aggregation. Analytic-groupby and cash-basis modes retain `DISTINCT`, as their temporary AML relations can contain duplicate AML IDs. Benchmark on the Balance Sheet query, 5 runs per size: | Matching AMLs | Before | After | |---:|---:|---:| | 2.3M | 3.19 s | 2.06 s | | 5.8M | 6.51 s | 2.91 s | | 11.5M | 9.78 s | 3.64 s | | 17.2M | 15.63 s | 4.67 s | | 23.0M | 16.01 s | 4.68 s | At 23M AMLs, execution time drops from 16.01 s to 4.68 s (3.42x). Result sets are identical before and after. opw-6459372 Forward-Port-Of: odoo/enterprise#128404