Add unaccent to SearchFilter

[mgaligniana]

Description

Based on discussion #7759 I've created this PR to continue the work made in #8775 and #7733

There are no tests with PostgreSQL but I'm creating this to test with the CI

[mgaligniana]

Hi @auvipy!

In case I would like to add a postgres db creation step, to be able to add new tests: should I add a step here .github/workflows/main.yml?

[mgaligniana]

I move this PR to Ready for review as the work is done but I need confirmation on how to add (or if it worth it) a step using PostgreSQL to be able to test it.

Thank you!

[jensjeflensje]

Tried this locally and the unaccent works great. It got me thinking though: anyone using unaccent is probably not interested in a full match either? The current code limits icontains from being used too alongside unaccent. I personally set my lookup prefix to "&": "unaccent__icontains" as that better fits my (and probably also other people's) usecases.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[Copilot]

Pull Request Overview

This PR adds unaccent support to the SearchFilter, enabling accent‐insensitive searches on PostgreSQL by mapping the "&" prefix to the "unaccent" lookup.

• Updates the lookup prefixes in the SearchFilter backend
• Introduces a PostgreSQL‑specific test for unaccent search
• Updates the API documentation to include the new lookup operator

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
tests/test_filters.py	Added a new test class that verifies unaccent search functionality
rest_framework/filters.py	Updated lookup prefixes to include the unaccent operator
docs/api-guide/filtering.md	Added new documentation row for accent‑insensitive unaccent search lookup

[auvipy]

test is failing with following error

raise FieldError(
E django.core.exceptions.FieldError: Unsupported lookup 'unaccent' for CharField or join on the field not permitted.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[mgaligniana]

test is failing with following error

raise FieldError( E django.core.exceptions.FieldError: Unsupported lookup 'unaccent' for CharField or join on the field not permitted.

Yes!

But is because the test is not being run over a PostgreSQL database, right?

I've tried adding this @pytest.mark.skipif(not postgres_fields, reason='PostgreSQL support (postgres_fields) is not available') to see if it is skiped by the CI, but nope! Any suggesiton?

[mgaligniana]

I've noticed that @ search never had a test with @ filter.

How bad it would be add this without testing? (as it is the same case)

[mgaligniana]

Hi @browniebroke! Just wondering if my last comment makes sense, or if you have any suggestions. Thank you!

[browniebroke]

I've noticed that @ search never had a test with @ filter.

Yes precisely. We've got some blind spots in our test coverage and we're trying to cover these kind of missing cases

[mgaligniana]

Hi @browniebroke ! How would I add tests that run on Postgres?

[mgaligniana]

Thank you @auvipy for the update! Let me know if there's anything I can help!

[browniebroke]

I've added a PR to configure CI to run with Postgres and add missing test coverage for full text search filter. If you cherry pick that commit here that should unblock you to write tests.

[mgaligniana]

Thank you @browniebroke !! I'll do it!!

Spotted an issue

[Natgho]

Integrating this feature as-is seems a bit risky. It was added with the note that it is specific to PostgreSQL, but there is no explicit flag for this.

If this addition must be made, it should be done alongside a few changes:

1. To ensure this feature does not break DRF’s database-agnostic behavior, it should be made functional by adding a check for “apps.is_installed(‘django.contrib.postgres’)” or evaluated within a try-except block.

2. One of the biggest risks of using unaccent is that it can trigger a “full table scan” if there is no GIN index at the database level. Since this feature will be used in the infrastructure of large projects like DRF, this risk must be explicitly mentioned in the documentation.

3. We need to see how it affects the lookup structure. How does unaccent interact with lookup_prefixes (e.g., istartswith for ^) during searches? Has the field__unaccent__icontains structure been tested for all prefix variations?

4. Documentation should be updated to remind users that they must run the CREATE EXTENSION IF NOT EXISTS unaccent; command on the PostgreSQL side to use this feature.

Rather than directly adding this feature to the core-drf code as it stands now, it would be a much safer and more professional approach to implement it via a plugin or as an optional feature that can be enabled.