Resolves performance issue where prefix deletion with 2000+ children
took 5-10 minutes due to sequential scans in hierarchy depth/children
calculations. Adding PostgreSQL GiST index with inet_ops enables efficient
network containment operators (>>, <<, <<=) in annotate_hierarchy() queries.
Performance impact:
- 30-60x speedup: 5-10 minutes → 10 seconds for large prefix deletions
- Real-world validation: 4s migration time on 1.24M prefix dataset
- Storage cost: 47MB index (11% of table storage, 38 bytes per prefix)
Works in conjunction with existing B-tree indexes on vrf_id for optimal
query performance. Benefits all network containment operations including
hierarchy navigation, aggregate views, and available IP/prefix calculations.
* Closes#20003: Introduce mechanism to register callbacks for webhook context
* Swap ContentType with ObjectType
* Add plugin dev documentation for webhook callbacks
* Fix tests
* Add note about namespacing webhook data
* Fixes#19986: Fix plugin list view button URLs
Plugin list view action buttons (Add, Import, Export) were generating 404
errors because ObjectAction.get_url() was manually constructing viewnames
without the required "plugins:" namespace prefix for plugin models.
Replace manual viewname construction with NetBox's get_viewname() utility
function, which properly handles plugin detection and namespace prefixing
for both core and plugin models.
* Ensure expected URL patterns are registered, ensures tests pass
- Extract script list content into reusable partial template
- Add object-list CSS class for proper embedded table styling
- Hide module headers and management actions in widget context
- Use compact buttons with icon-only labels for widgets
- Add test coverage for embedded parameter handling
The embedded version now renders cleanly in dashboard widgets while
preserving full functionality in the main script list page.
* Fix MAC address pagination duplicates by adding 'pk' to model ordering
Add 'pk' to MACAddress model ordering to ensure deterministic results
when multiple MAC addresses have the same value. This prevents the same
MAC address from appearing on multiple pages during pagination.
The issue occurred because Django's default ordering by 'mac_address'
alone is non-deterministic when multiple records share the same MAC
address value, causing inconsistent pagination results when the same
MAC address is assigned to multiple interfaces on a device.
Added regression test that verifies MAC addresses with identical values
are properly ordered by their primary key, ensuring consistent pagination
behavior across the application.
Fixes netbox-community#19917
* Remove test
* Resolve migration conflict
---------
Co-authored-by: Jad Seifeddine <jseifeddine@macquarietelecom.com>
Co-authored-by: Jeremy Stretch <jstretch@netboxlabs.com>
Replace direct string concatenation with URLSearchParams to properly
encode user input in export link URLs, preventing injection of malicious
parameters or scripts through the search functionality.
Resolves CodeQL Alert #63 (js/xss-through-dom)
* Closes#19977: Denormalize site, location, and rack for device components
* Set blank=True on denormalized ForeignKeys
* Populate denormalized field in test data
* Ignore private fields when constructing test GraphQL requests
* Closes#19968: Use multiple selection lists for the assignment of object types when editing a permission
* Remove errant logging statements
* Defer compilation of choices for object_types
* Fix test data
* Convert ObjectType to a concrete child model of ContentType
* Add public flag to ObjectType
* Catch post_migrate signal to update ObjectTypes
* Reference ObjectType records instead of registry for feature support
* Automatically create ObjectTypes
* Introduce has_feature() utility function
* ObjectTypeManager should not inherit from ContentTypeManager
* Misc cleanup
* Don't populate ObjectTypes during migration
* Don't automatically create ObjectTypes when a ContentType is created
* Fix test
* Extend has_feature() to accept a model or OT/CT
* Misc cleanup
* Deprecate get_for_id() on ObjectTypeManager
* Rename contenttypes.py to object_types.py
* Add index to features ArrayField
* Keep FK & M2M fields pointing to ContentType
* Add get_for_models() to ObjectTypeManager
* Add tests for manager methods & utility functions
* Fix migrations for M2M relations to ObjectType
* model_is_public() should return False for non-core & non-plugin models
* Order ObjectType by app_label & model name
* Resolve migrations conflict
* Fixes#18900: introduce/raise QuerySetNotOrdered exception
Defines a new exception, `QuerySetNotOrdered`, and raises it in
`OptionalLimitOffsetPagination.paginate_queryset` in the right
conditions:
- the iterable to be paginated is a QuerySet isinstance
- the `queryset.ordered` flag is not truthy
* Don't try to reapply ordering if ordering is already present
* Add ordering for failing tagged-objects list API endpoint
I chose to implement this here for TaggedItemViewSet, rather than on the
model, because any meaningful ordering is going to be done on the
related Tag instance and I didn't want to introduce potential, not well
understood side-effects by applying a model-wide ordering via a related
model field.
* Add default Token ordering behavior
* Adds basic tests for raising QuerySetNotOrdered
* Note why ordering is not applied in TaggedItem.Meta
* Add message field to ObjectChange model
* Set max length on changelog message
* Enable changelog messages for single object operations
* Fix tests
* Add changelog message support for bulk edit & bulk delete
* Cosmetic improvements to form fields
* Fix bulk operation templates
* Add message support for bulk import/update
* Add REST API support for changelog messages (WIP)
* Fix changelog_message assignment
* Enable changelog message support for bulk deletions
* Add documentation
* Fix changelog message support for VirtualChassis
* Add ChangeLoggingMixin to necesssary model forms
* Introduce get_random_string() utility function for tests
* Incorporate changelog messages for object view tests
* Incorporate changelog messages for object bulk view tests
* Add missing mixins for changelog message support
* Tweak test to generate expected number of change records
* Finish adding tests for changelog message functionality
* Misc cleanup
* Fixes#19956: Prevent duplicate deletion records from cascading deletions
* Tweak bulk deletion test to work around cascading deletions issue
* Correct API URL
* Show human-friendly values for file size
* Introduce optional dedicated columns for name & filename
* Add combined dimensions column
* Restore image preview on hover
* Remove object_type from default columns list
* Parent column is not orderable
* Filter/search image attachments by filename
* Correct table column name
* Closes: #18588: Relabel Service model to Application Service
Updates the `verbose_name` of the `Service` and `ServiceTemplate` models to "Application Service" and
"Application Service Template" respectively. This serves as the foundational change for relabeling
the model throughout the user interface to reduce ambiguity.
To preserve backward compatibility for the REST and GraphQL APIs, the test suites have been updated
to assert the stability of the original field and parameter names. This includes:
* Using `filter_name_map` in the filterset test case to ensure API query parameters remain
`service` and `service_id`.
* Employing the GraphQL test suite's aliasing mechanism to ensure the public schema remains
unchanged despite the underlying `verbose_name` modification.
Subsequent commits will address UI-specific labels in navigation, tables, forms, and templates.
* Rename to Application Services/Application Service Templates in nav menu
* Rename ~service to ~'Application Service' in templates
This was done for both the Service model and Service Template model
appearances in templates where the word was hardcoded.
* Change ~service to ~'application service' hardcoded strings in Python files
* Update ~service to ~'application service' in docs
* Add background_job toggle to BulkEditForm
* Account for bug fix in v4.3.4
* Enable background jobs for bulk edit & bulk delete
* Move background_job field to a mixin
* Cosmetic improvements
* Misc cleanup
* Fix BackgroundJobMixin
* Initial work on #19589
* Add tooling for handling background requests
* UI notification should link to enqueued job
* Use an informative name for the job
* Disable background jobs for file uploads
* Closes#19231: Add bulk renaming support for all models
* Introduce a template filter for getattr()
* Extend BulkRenameView to support arbitrary field names
* Address bulk renaming support for remaining models
* Bulk rename URL resolution should fail silently
* Update documentation
* Fix bulk button rendering for HTMX requests
* Initial work on #19735
* Work in progress
* Remove ClusterRemoveDevicesView (anti-pattern)
* Misc cleanup
* Fix has_bulk_actions
* Fix has_bulk_actions for ObjectChildrenView
* Restore clone button
* Misc cleanup
* Clean up custom bulk actions
* Rename individual object actions
* Collapse into a single template tag
* Fix support for legacy action dicts
* Rename bulk attr to multi
* clone_button tag should fail silently if view name is invalid
* Clean up action buttons
* Fix export button label
* Replace clone_button with an ObjectAction
* Create object actions for adding device/VM components
* Move core_sync.html to core app
* Remove extra_bulk_buttons from template doc
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.