Previously, generating the budget.report table was necessary to trigger _compute_all for budget.line fields. This table was built using three separate queries with a UNION operator. Because of the UNION, any filtering (like on specific budget_line_ids) happened after the full, unfiltered table was generated.

This post-filtering caused slowness, especially in nested loop joins with large tables like account.analytic.line. This commit optimizes performance by pushing the filter condition (using specific budget_line_ids) directly down into the three underlying queries. This reduces the number of budget.line records processed, speeding up joins and overall computation.

The benchmark below is done on a database that has **66396** `budget.line` records and **928567** `account.analytic.line` records. Opening a budget report for a specific year, only applied the filter with **40** `budget.line` records.

| Scenario | Execution Time |
| :--- | :--- |
| **Before this Commit** | **60.00 seconds**
| **After this Commit** | **1.84 seconds**

opw-5150569

Behavior before:
Exporting the TPAR report could fail with an error indicating that the GST amount exceeded the
gross paid for certain vendors. Some payment lines were excluded from the SQL query,
leading to discrepancies in the report.

Behavior after:
The SQL query has been updated to include all relevant payment lines, including those in
asset_cash accounts. This ensures that the gross paid is calculated correctly, allowing the
TPAR report to be exported without errors.

Root Cause:
The previous query only considered asset_current and liability_current accounts, ignoring
asset_cash accounts. This omission caused the gross paid amount to be understated relative to
the total GST paid.

Steps to Reproduce:

1. Create a vendor bill.

2. Navigate to the Bank Reconciliation page (Accounting → Dashboard → Bank Journal).

3. Create a bank statement line for the bill amount.

4. Validate the statement line with the corresponding invoice line.

5. Generate the TPAR report — the payment amount should now be correctly included.

OPW: 5148697

Forward-Port-Of: odoo/enterprise#99874

The \_retrieve \_product() function relies on a query with multiple domains OR'ed together. As the search will require a LEFT JOIN with the product_template table, the use of OR in the query prevents Postgres from utilizing indexes. This becomes a problem in databases with a large number of products since a seq scan would be very slow.

This commit changes the way this is done by performing separate queries instead of a single query with multiple conditions within an OR statement. Although this might seem a performance degradation, it actually allows these separate queries to utilize indexes and run much faster compared to the original approach. It also simplifies the priority logic and allows for faster early exits compared to the original one.

This function is mainly used with EDI crons (such as PEPPOL where this problem was noticed), which could require hundreds of product searches as it does a search per invoice line.

Benchmarks:

Importing a peppol document of 173 invoice lines.

| Num products | Num invoice lines | Before   | After   |
| ------------ | ----------------- | -------- | ------- |
| 864873       | 173               | 868.18 s | 19.43 s |
| 397005       | 173               | 468.91 s | 20.68 s |
| 8064         | 173               | 125.08 s | 20.1 s  |
| 564          | 173               | 119.9 s  | 20.21 s |

opw-5245007


---
I confirm I have signed the CLA and read the PR guidelines at www.odoo.com/submit-pr

Forward-Port-Of: odoo/odoo#240557
Forward-Port-Of: odoo/odoo#238853