[16.0][FIX] project_hr: Slowdown when entering the task tree view

[iaranburu]

Odoo 16

When we have many projects, and each project has many tasks, accessing the task tree view from a project slows down the loading process.

This request corrects the calculation of the allowed_assigned_user_ids field so that it only performs one database access and uses filtering during task iteration.

[OCA-git-bot]

Hi @pedrobaeza,
some modules you are maintaining are being modified, check this out!

[luisDIXMIT]

Pre-commit is failing, maybe because you didn’t run it previously. Run it locally to fix the formatting and push it again.

[alexey-pelykh]

Good optimization idea — batching the search instead of one per record is the right approach for computed fields.

Pre-commit is currently failing though, so that needs a pass before this can move forward. Also, the early return when all_category_ids is empty sets the field on the full recordset, which is fine, but you might also want to handle the case where a task has hr_category_ids populated but company_id is falsy (edge case).

[alexey-pelykh]

When all_category_ids is empty, you set self.allowed_assigned_user_ids = False and return early. But you should still iterate the tasks to handle the else branch individually — a task might have no categories but another in the batch might. Actually, looking again, if all_category_ids is empty it means NO task in self has categories, so the early return is correct. Disregard this, the logic is fine.

[alexey-pelykh]

Review: Performance optimization for _compute_allowed_assigned_user_ids

Good approach -- replacing N individual search() calls with a single batched search + in-memory filtering is the right pattern for Odoo computed fields. The performance improvement should be significant when computing across many tasks.

However, there is a behavioral regression and some issues to address before this can be merged.

Behavioral Regression (blocking)

In the original code, when a task has no hr_category_ids, the domain is [] and user_obj.search([]) returns all users -- meaning allowed_assigned_user_ids is populated with every user. In the new code, the else branch sets task.allowed_assigned_user_ids = False (empty recordset). This changes the behavior: tasks without HR categories would show no allowed users instead of all users, potentially breaking the UI for task assignment.

Please restore the original behavior for the no-categories case:

else:
    task.allowed_assigned_user_ids = self.env["res.users"].search([])

Or, better yet, cache that full user list outside the loop so it is only fetched once.

Pre-commit failure (blocking)

CI shows pre-commit is still failing. Please run pre-commit run --all-files locally and push the fixes.

Minor: string quoting style

The new code uses single quotes ('hr_category_ids', 'company_id') while the rest of the file uses double quotes. OCA convention and the existing code style in this file use double quotes -- please align for consistency.

Suggestion: consider performance of filtered() lambda

The filtered() lambda iterates u.employee_ids.category_ids and u.employee_ids for every user in all_users for every task. If the number of users is large, this could still be slow. Consider building a mapping dict (user -> set of (company_id, category_ids)) before the loop for O(1) lookups per task-user pair.

Overall, the optimization direction is correct. Please fix the behavioral regression and pre-commit issues.

Review posted via CorporateHub OCA review campaign

[alexey-pelykh]

Behavioral regression: When task.hr_category_ids is falsy, the original code ran user_obj.search([]) which returns all users. Setting False here changes the behavior -- tasks without HR categories would show no allowed users instead of all users.

Please restore:

else:
    task.allowed_assigned_user_ids = self.env["res.users"].search([])

Or cache the full user list outside the loop to avoid a per-task search.

[alexey-pelykh]

Minor: single quotes here ('hr_category_ids') while the rest of the file uses double quotes. Please use double quotes for consistency with OCA style conventions.

Suggested change

	all_category_ids = self.mapped('hr_category_ids').ids
	all_category_ids = self.mapped("hr_category_ids").ids
	all_company_ids = self.mapped("company_id").ids