- Add optional locale parameter to formatNumber function
- Update tests to explicitly pass 'en-US' locale for deterministic results
- Maintains backward compatibility: defaults to system locale when no locale specified
- No need for cross-env or environment variable manipulation
- Tests now pass consistently regardless of system locale
Related to #4417
* fix: handle UTF-8 BOM in lyrics and playlist files
Added UTF-8 BOM (Byte Order Mark) detection and stripping for external lyrics files and playlist files. This ensures that files with BOM markers are correctly parsed and recognized as synced lyrics or valid playlists.
The fix introduces a new ioutils package with UTF8Reader and UTF8ReadFile functions that automatically detect and remove UTF-8, UTF-16 LE, and UTF-16 BE BOMs. These utilities are now used when reading external lyrics and playlist files to ensure consistent parsing regardless of BOM presence.
Added comprehensive tests for BOM handling in both lyrics and playlists, including test fixtures with actual BOM markers to verify correct behavior.
* test: add test for UTF-16 LE encoded LRC files
Signed-off-by: Deluan <deluan@navidrome.org>
---------
Signed-off-by: Deluan <deluan@navidrome.org>
* feat: add basic tag extraction fallback mechanism
Added basic tag extraction from TagLib's generic Tag interface as a fallback
when PropertyMap doesn't contain standard metadata fields. This ensures that
essential tags like title, artist, album, comment, genre, year, and track
are always available even when they're not present in format-specific
property maps.
Changes include:
- Extract basic tags (__title, __artist, etc.) in C++ wrapper
- Add parseBasicTag function to process basic tags in Go extractor
- Refactor parseProp function to be reusable across property parsing
- Ensure basic tags are preferred over PropertyMap when available
* feat(taglib): update tag parsing to use double underscores for properties
Signed-off-by: Deluan <deluan@navidrome.org>
---------
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: prevent infinite loop in Type filter autocomplete
Fixed an infinite loop issue in the album Type filter caused by an inline
arrow function in the optionText prop. The inline function created a new
reference on every render, causing React-Admin's AutocompleteInput to
continuously re-fetch data from the /api/tag endpoint.
The solution extracts the formatting function outside the component scope
as formatReleaseType, ensuring a stable function reference across renders.
This prevents unnecessary re-renders and API calls while maintaining the
humanized display format for release type values.
* fix: enable multi-valued releasetype in smart playlists
Smart playlists can now match all values in multi-valued releasetype tags.
Previously, the albumtype field was mapped to the single-valued mbz_album_type
database field, which only stored the first value from tags like album; soundtrack.
This prevented smart playlists from matching albums with secondary release types
like soundtrack, live, or compilation when tagged by MusicBrainz Picard.
The fix removes the direct database field mapping and allows both albumtype and
releasetype to use the multi-valued tag system. The albumtype field is now an
alias that points to the releasetype tag field, ensuring both query the same
JSON path in the tags column. This maintains backward compatibility with the
documented albumtype field while enabling proper multi-value tag matching.
Added tests to verify both releasetype and albumtype correctly generate
multi-valued tag queries.
Fixes#4616
* fix: resolve albumtype alias for all operators and sorting
Codex correctly identified that the initial fix only worked for Contains/StartsWith/EndsWith operators. The alias resolution was happening too late in the code path.
Fixed by resolving the alias in two places:
1. tagCond.ToSql() - now uses the actual field name (releasetype) in the JSON path
2. Criteria.OrderBy() - now uses the actual field name when building sort expressions
Added tests for Is/IsNot operators and sorting to ensure complete coverage.
* chore: update to Go 1.25.3
Signed-off-by: Deluan <deluan@navidrome.org>
* chore: update to golangci-lint
Signed-off-by: Deluan <deluan@navidrome.org>
* chore: update go dependencies
Signed-off-by: Deluan <deluan@navidrome.org>
* chore: update vite dependencies in package.json and improve EventSource mock in tests
- Upgraded @vitejs/plugin-react to version 5.1.0 and @vitest/coverage-v8 to version 4.0.3.
- Updated vite to version 7.1.12 and vite-plugin-pwa to version 1.1.0.
- Enhanced the EventSource mock implementation in eventStream.test.js for better test isolation.
* ci: remove coverage flag from Go test command in pipeline
* chore: update Node.js version to v24 in devcontainer, pipeline, and .nvmrc
* chore: prettier
Signed-off-by: Deluan <deluan@navidrome.org>
* chore: update actions/checkout from v4 to v5 in pipeline and update-translations workflows
* chore: update JS dependencies remove unused jest-dom import in Linkify.test.jsx
* chore: update actions/download-artifact from v4 to v5 in pipeline
---------
Signed-off-by: Deluan <deluan@navidrome.org>
Added genre as a toggleable column in the playlist songs table. The Genre column
displays genre information for each song in playlists and is available through
the column toggle menu but disabled by default.
Implements feature request from GitHub discussion #4400.
Signed-off-by: Deluan <deluan@navidrome.org>
Added functionality to populate the Folder field in GetUser and GetUsers API responses
with the library IDs that the user has access to. This allows Subsonic API clients
to understand which music folders (libraries) a user can access for proper
content filtering and UI presentation.
Signed-off-by: Deluan <deluan@navidrome.org>
* feat(plugins): add PluginList method
Signed-off-by: Deluan <deluan@navidrome.org>
* feat: enhance insights collection with plugin awareness and expanded metrics
Enhanced the insights collection system to provide more comprehensive telemetry data about Navidrome installations. This update adds plugin awareness through dependency injection integration, expands configuration detection capabilities, and includes additional library metrics.
Key improvements include:
- Added PluginLoader interface integration to collect plugin information when enabled
- Enhanced configuration detection with proper credential validation for LastFM, Spotify, and Deezer
- Added new library metrics including Libraries count and smart playlist detection
- Expanded configuration insights with reverse proxy, custom PID, and custom tags detection
- Updated Wire dependency injection to support the new plugin loader requirement
- Added corresponding data structures for plugin information collection
This enhancement provides valuable insights into feature usage patterns and plugin adoption while maintaining privacy and following existing telemetry practices.
* fix: correct type assertion in plugin manager test
Fixed type mismatch in test where PluginManifestCapabilitiesElem was being
compared with string literal. The test now properly casts the string to the
correct enum type for comparison.
* refactor: move static config checks to staticData function
Moved HasCustomTags, ReverseProxyConfigured, and HasCustomPID configuration checks from the dynamic collect() function to the static staticData() function where they belong. This eliminates redundant computation on every insights collection cycle and implements the actual logic for HasCustomTags instead of the hardcoded false value.
The HasCustomTags field now properly detects if custom tags are configured by checking the length of conf.Server.Tags. This change improves performance by computing static configuration values only once rather than on every insights collection.
* feat: add granular control for insights collection
Added DevEnablePluginsInsights configuration option to allow fine-grained control over whether plugin information is collected as part of the insights data. This change enhances privacy controls by allowing users to opt-out of plugin reporting while still participating in general insights collection.
The implementation includes:
- New configuration option DevEnablePluginsInsights with default value true
- Gated plugin collection in insights.go based on both plugin enablement and permission flag
- Enhanced plugin information to include version data alongside name
- Improved code organization with clearer conditional logic for data collection
* refactor: rename PluginNames parameter from serviceName to capability
Signed-off-by: Deluan <deluan@navidrome.org>
---------
Signed-off-by: Deluan <deluan@navidrome.org>
* fix(auth): Do not try reverse proxy auth if internal auth succeeds
* cmp.Or will still require function results to be evaluated...
* move to a function
* fix: resolve playlist import issues with Unicode character paths
Fixes#3332 where songs with accented characters failed to import from Apple Music M3U playlists. The issue occurred because Apple Music exports use NFC Unicode normalization while macOS filesystem stores paths in NFD normalization.
Added normalizePathForComparison() function that normalizes both filesystem and M3U playlist paths to NFC form before comparison. This ensures consistent path matching regardless of the Unicode normalization form used.
Changes include comprehensive test coverage for Unicode normalization scenarios with both NFC and NFD character representations.
* address comments
Signed-off-by: Deluan <deluan@navidrome.org>
* fix(tests): add check for unequal original Unicode paths in playlist normalization tests
Signed-off-by: Deluan <deluan@navidrome.org>
---------
Signed-off-by: Deluan <deluan@navidrome.org>
* fix(server): remove includeMissing from search (always false)
Signed-off-by: Deluan <deluan@navidrome.org>
* fix(search): optimize search order by using natural order for improved performance
Signed-off-by: Deluan <deluan@navidrome.org>
---------
Signed-off-by: Deluan <deluan@navidrome.org>
* refactor: improve URL path handling in local storage system
Enhanced the local storage implementation to properly handle URL-decoded paths
and fix issues with file paths containing special characters. Added decodedPath
field to localStorage struct to separate URL parsing concerns from file system
operations.
Key changes:
- Added decodedPath field to localStorage struct for proper URL decoding
- Modified newLocalStorage to use decoded path instead of modifying original URL
- Fixed Windows path handling to work with decoded paths
- Improved URL escaping in storage.For() to handle special characters
- Added comprehensive test suite covering URL decoding, symlink resolution,
Windows paths, and edge cases
- Refactored test extractor to use mockTestExtractor for better isolation
This ensures that file paths with spaces, hash symbols, and other special
characters are handled correctly throughout the storage system.
Signed-off-by: Deluan <deluan@navidrome.org>
* fix(tests): fix test file permissions and add missing tests.Init call
* refactor(tests): remove redundant test
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: URL building for Windows and remove redundant variable
Signed-off-by: Deluan <deluan@navidrome.org>
* refactor: simplify URL path escaping in local storage
Signed-off-by: Deluan <deluan@navidrome.org>
---------
Signed-off-by: Deluan <deluan@navidrome.org>
* ui: reset activity icon after viewing error
* refactor: improve ActivityPanel error acknowledgment logic
Replaced boolean errorAcknowledged state with acknowledgedError string state to track which specific error was acknowledged. This prevents icon flickering when error messages change and simplifies the logic by removing the need for useEffect.
Key changes:
- Changed from errorAcknowledged boolean to acknowledgedError string state
- Added derived isErrorVisible computed value for cleaner logic
- Removed useEffect dependency on scanStatus.error changes
- Updated handleMenuOpen to store actual error string instead of boolean flag
- Fixed test mock to return proper error state matching test expectations
This change addresses code review feedback and follows React best practices by using derived state instead of imperative effects.
* fix: prevent foreign key constraint error in album participants
Prevent foreign key constraint errors when album participants contain
artist IDs that don't exist in the artist table. The updateParticipants
method now filters out non-existent artist IDs before attempting to
insert album_artists relationships.
- Add defensive filtering in updateParticipants() to query existing artist IDs
- Only insert relationships for artist IDs that exist in the artist table
- Add comprehensive regression test for both albums and media files
- Fixes scanner errors when JSON participant data contains stale artist references
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: optimize foreign key handling in album artists insertion
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: improve participants foreign key tests
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: clarify comments in album artists insertion query
Signed-off-by: Deluan <deluan@navidrome.org>
* test: add cleanup to album repository tests
Added individual test cleanup to 4 album repository tests that create temporary
artists and albums. This ensures proper test isolation by removing test data
after each test completes, preventing test interference when running with
shuffle mode. Each test now cleans up its own temporary data from the artist,
library_artist, album, and album_artists tables using direct SQL deletion.
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: refactor participant JSON handling for simpler and improved SQL processing
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: update test command description in Makefile for clarity
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: refactor album repository tests to use albumRepository type directly
Signed-off-by: Deluan <deluan@navidrome.org>
---------
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: prevent foreign key constraint error in tag UpdateCounts
Added JOIN clause with tag table in UpdateCounts SQL query to filter out
tag IDs from JSON that don't exist in the tag table. This prevents
'FOREIGN KEY constraint failed' errors when the library_tag table
tries to reference non-existent tag IDs during scanner operations.
The fix ensures only valid tag references are counted while maintaining
data integrity and preventing scanner failures during library updates.
* test(tag): add regression tests for foreign key constraint fix
Add comprehensive regression tests to prevent the foreign key constraint
error when tag IDs in JSON data don't exist in the tag table. Tests cover
both album and media file scenarios with non-existent tag IDs.
- Test UpdateCounts() with albums containing non-existent tag IDs
- Test UpdateCounts() with media files containing non-existent tag IDs
- Verify operations complete without foreign key errors
Signed-off-by: Deluan <deluan@navidrome.org>
---------
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: enable library access for headless processes
Fixed multi-library filtering to allow headless processes (shares, external providers) to access data by skipping library restrictions when no user context is present.
Previously, the library filtering system returned empty results (WHERE 1=0) for processes without user authentication, breaking functionality like public shares and external service integrations.
Key changes:
- Modified applyLibraryFilter methods to skip filtering when user.ID == invalidUserId
- Refactored tag repository to use helper method for library filtering logic
- Fixed SQL aggregation bug in tag statistics calculation across multiple libraries
- Added comprehensive test coverage for headless process scenarios
- Updated genre repository to support proper column mappings for aggregated data
This preserves the secure "safe by default" approach for authenticated users while restoring backward compatibility for background processes that need unrestricted data access.
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: resolve SQL ambiguity errors in share queries
Fixed SQL ambiguity errors that were breaking share links after the Multi-library PR.
The Multi-library changes introduced JOINs between album and library tables,
both of which have 'id' columns, causing 'ambiguous column name: id' errors
when unqualified column references were used in WHERE clauses.
Changes made:
- Updated core/share.go to use 'album.id' instead of 'id' in contentsLabelFromAlbums
- Updated persistence/share_repository.go to use 'album.id' in album share loading
- Updated persistence/sql_participations.go to use 'artist.id' for consistency
- Added regression tests to prevent future SQL ambiguity issues
This resolves HTTP 500 errors that users experienced when accessing existing
share URLs after the Multi-library feature was merged.
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: improve headless library access handling
Added proper user context validation and reordered joins in applyLibraryFilterToArtistQuery to ensure library filtering works correctly for both authenticated and headless operations. The user_library join is now only applied when a valid user context exists, while the library_artist join is always applied to maintain proper data relationships. (+1 squashed commit)
Squashed commits:
[a28c6965b] fix: remove headless library access guard
Removed the invalidUserId guard condition in applyLibraryFilterToArtistQuery that was preventing proper library filtering for headless operations. This fix ensures that library filtering joins are always applied consistently, allowing headless library access to work correctly with the library_artist junction table filtering.
The previous guard was skipping all library filtering when no user context was present, which could cause issues with headless operations that still need to respect library boundaries through the library_artist relationship.
* fix: simplify genre selection query in genre repository
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: enhance tag library filtering tests for headless access
Signed-off-by: Deluan <deluan@navidrome.org>
* test: add comprehensive test coverage for headless library access
Added extensive test coverage for headless library access improvements including:
- Added 17 new tests across 4 test files covering headless access scenarios
- artist_repository_test.go: Added headless process tests for GetAll, Count,
Get operations and explicit library_id filtering functionality
- genre_repository_test.go: Added library filtering tests for headless processes
including GetAll, Count, ReadAll, and Read operations
- sql_base_repository_test.go: Added applyLibraryFilter method tests covering
admin users, regular users, and headless processes with/without custom table names
- share_repository_test.go: Added headless access tests and SQL ambiguity
verification for the album.id vs id fix in loadMedia function
- Cleaned up test setup by replacing log.NewContext usage with GinkgoT().Context()
and removing unnecessary configtest.SetupConfig() calls for better test isolation
These tests ensure that headless processes (background operations without user context)
can access all libraries while respecting explicit filters, and verify that the SQL
ambiguity fixes work correctly without breaking existing functionality.
* revert: remove user context handling from scrobble buffer getParticipants
Reverts commit 5b8ef74f05.
The artist repository no longer requires user context for proper library
filtering, so the workaround of temporarily injecting user context into
the scrobbleBufferRepository.Next method is no longer needed.
This simplifies the code and removes the dependency on fetching user
information during background scrobbling operations.
* fix: improve library access filtering for artists
Enhanced artist repository filtering to properly handle library access restrictions
and prevent artists with no accessible content from appearing in results.
Backend changes:
- Modified roleFilter to use direct JSON_EXTRACT instead of EXISTS subquery for better performance
- Enhanced applyLibraryFilterToArtistQuery to filter out artists with empty stats (no content)
- Changed from LEFT JOIN to INNER JOIN with library_artist table for stricter filtering
- Added condition to exclude artists where library_artist.stats = '{}' (empty content)
Frontend changes:
- Added null-checking in getCounter function to prevent TypeError when accessing undefined records
- Improved optional chaining for safer property access in role-based statistics display
These changes ensure that users only see artists that have actual accessible content
in their permitted libraries, fixing issues where artists appeared in the list
despite having no albums or songs available to the user.
* fix: update library access logic for non-admin users and enhance test coverage
Signed-off-by: Deluan <deluan@navidrome.org>
* fix: refine library artist query and implement cleanup for empty entries
Signed-off-by: Deluan <deluan@navidrome.org>
* refactor: consolidate artist repository tests to eliminate duplication
Significantly refactored artist_repository_test.go to reduce code duplication and
improve maintainability by ~27% (930 to 680 lines). Key improvements include:
- Added test helper functions createTestArtistWithMBID() and createUserWithLibraries()
to eliminate repetitive test data creation
- Consolidated duplicate MBID search tests using DescribeTable for parameterized testing
- Removed entire 'Permission-Based Behavior Comparison' section (~150 lines) that
duplicated functionality already covered in other test contexts
- Reorganized search tests into cohesive 'MBID and Text Search' section with proper
setup/teardown and shared test infrastructure
- Streamlined missing artist tests and moved them to dedicated section
- Maintained 100% test coverage while eliminating redundant test patterns
All tests continue to pass with identical functionality and coverage.
---------
Signed-off-by: Deluan <deluan@navidrome.org>
Add EnsureCompiled calls in plugin test BeforeEach blocks to wait for
WebAssembly compilation before loading plugins. This prevents race conditions
where tests would attempt to load plugins before compilation completed,
causing flaky test failures in CI environments.
The race condition occurred because ScanPlugins() registers plugins
synchronously but compiles them asynchronously in background goroutines
with a concurrency limit of 2. Tests that immediately called LoadPlugin()
or LoadMediaAgent() after ScanPlugins() could fail if compilation wasn't
finished yet.
Fixed in both adapter_media_agent_test.go and manager_test.go which had
multiple tests vulnerable to this timing issue.
