v0.8.2

Highlights

• Preserves transient in-memory fields when entities are rehydrated after
  create and update flows, preventing non-persisted state from being lost.

Improvements

• Tightens rehydration behavior so mutation flows return fully refreshed
  entities without discarding caller-managed transient data.

API cleanup

• No API changes in this release.

Internal changes

• Reorganizes association coverage into dedicated create, update, delete,
  tx, and shared fixture test files for easier maintenance.

v0.8.1

Highlights

• This release does not introduce new user-facing features; it prepares the repository for easier downstream testing and interface-based integration work.

Improvements

• No behavior changes are included beyond the previously published v0.8.0 release.

API cleanup

• No public API changes are included in this release.

Internal changes

• Added generated mocks for repository interfaces to improve test ergonomics and make internal verification of repository interactions easier to maintain.

v0.8.0

Highlights

• Created entities now reload through the same rehydration path as updates, so requested preloads and auto-preloaded relations are returned from persisted data immediately after create operations.
• Repository and transaction write flows now hydrate belongs-to and many-to-many associations consistently after inserts, reducing follow-up queries and mismatched post-write state.

Improvements

• Create operations now honor post-write preload behavior consistently with update operations, making relation loading after inserts more predictable.
• Shared mutation preload validation and reload helpers reduce divergence between repository and transaction code paths.

API cleanup

• No public API surface was removed, but create-time reload behavior now matches update semantics more closely when preloads or auto-preloads are configured.

Internal changes

• Refactored mutation reload handling into shared helpers reused by both create and update flows.
• Expanded repository, transaction, query builder, and association test coverage for create-side reload and preload scenarios.

v0.7.0

Highlights

• Added post-update preloading support so update workflows can immediately hydrate requested relations after persistence.
• Updated entities now reload automatically when sync behavior or explicit preloads are configured, making update results more consistent with create and query flows.

Improvements

• Improved repository update behavior to return fully refreshed entities when association syncing or preload options are in use.
• Tightened update execution around prepared statements by using context-aware query execution paths.

API cleanup

• No breaking API changes in this release.

Internal changes

• Expanded repository and transaction test coverage for update preload hydration and association reload behavior.
• Refined the update pipeline internals to support post-update reloads without changing the public release workflow.

v0.6.0

Highlights

• Split repository metadata into dedicated sqlr struct tags, making entity definitions clearer and keeping column mapping in db tags.
• Added schema-level association sync defaults and aligned sync: tag syntax with sqlh, including support for nested sync:delete defaults.
• Added Create support for omitting all associations in a single call, and made owned-association cleanup cascade by default during deletes.

Improvements

• Made many-to-many updates link-only by default to avoid unintended row mutations in related tables.
• Fixed has-one updates so cleared relations are persisted correctly.
• Updated prepared statement handling to work with sqlc v0.3.0 and the newer client API.
• Tightened schema parsing by rejecting unexported fields with sqlr metadata.

API cleanup

• sqlr metadata is now expected in sqlr tags instead of being mixed into db tags.
• sqlr tag options must now be separated with semicolons.
• Association sync behavior has been normalized around the sync: option model, which may require tag updates in existing entities.

Internal changes

• Simplified schema parsing helpers and cleaned up lint issues.
• Expanded repository and schema coverage around association lifecycle, delete behavior, preload handling, and prepared statements.
• Updated the bundled sqlc dependency to v0.3.0.

v0.5.0

Highlights

• Add fixture writer helpers for gosoline fixture sets, making it easier for applications built on sqlr to load deterministic integration-test data with caller-managed IDs and timestamps.
• Add DisableAutoUpdates write controls so create and update flows can preserve caller-supplied IDs, timestamps, and association fields when needed.
• Export schema parsing helpers and add type-based schema parsing, making it easier to build reflection-driven integrations on top of sqlr.
• Introduce canonical relation path resolution so preloads, joins, and association sync follow the same nested path rules.

Improvements

• Restore entity IDs, timestamps, and foreign keys after failed writes so in-memory entities do not remain partially mutated.
• Tighten automatic relationship detection to reduce false positives for untagged value objects and incomplete foreign key setups.
• Reject conflicting scalar join rows while still deduplicating exact duplicates for more predictable join hydration.

API cleanup

• Standardize schema and relation-path handling around the exported parsing helpers for more consistent behavior across generic and reflection-based callers.
• No breaking API removals are included in this release.

Internal changes

• Split association internals into focused modules and centralize helper context handling to simplify maintenance.
• Expand coverage for preload failures, transactional repository paths, prepared statement lifecycle errors, and fixture loading workflows.

v0.4.0

Highlights

• Added selective association sync controls for create and update flows, allowing specific association paths to be synced or omitted with finer control.
• Improved association persistence so existing has-one and has-many children are linked correctly during create operations.
• Expanded relation handling with better support for pointer-based relations, pointer auto timestamps, and zero-valued relation keys.

Improvements

• Fixed joined reads to avoid truncating results when joins are involved.
• Fixed cases where joined queries could skip preloads.
• Improved preload execution safety by deduplicating paths and only parallelizing independent branches.
• Hardened schema validation and relation persistence behavior.
• Removed the hidden ID-setter requirement during create by falling back to reflective primary key assignment when needed.
• Added clearer error handling for nil entities and better guarding for nil relation conditions.
• Improved update-not-found handling and prepared statement validation in transactional flows.

API cleanup

• Removed unsupported RightJoin and CrossJoin builder APIs to align the public surface with supported behavior.

Internal changes

• Refactored association synchronization, persistence helpers, and relation wrappers to simplify the implementation.
• Expanded regression coverage across association, preload, schema, transaction, read, create, and update behavior.
• Applied general lint and test cleanups.

v0.3.0

Highlights

• update gosoline to v0.57.2 and sqlc to v0.2.0
• add GitHub Actions CI checks for build, test, and lint
• align lint and mockery configuration with the updated toolchain

Changes since v0.2.0

• refresh transitive Go module dependencies after the version bumps
• update .golangci.yml and .mockery.yml to match the current setup

v0.2.0

New Features

Global SchemaNameTransformer

A new package-level SchemaNameTransformer variable (default: toSnakeCase) controls how Go identifiers (struct type names, field names) are converted to database identifiers (table names, column names, M2M join-column names). It can be replaced at program startup to apply custom naming conventions.

// Use lowercase field names instead of snake_case:
sqlr.SchemaNameTransformer = strings.ToLower

Auto-Detection of Untagged Relationships

Public struct and slice-of-struct fields no longer require a db tag to be recognized as relationships:

• Non-slice struct field → BelongsTo, FK derived as SchemaNameTransformer(fieldName) + "_id"
• Slice-of-struct field → HasMany, FK derived as SchemaNameTransformer(parentTypeName) + "_id"

Standard library value types (time.Time, sql.NullString, net.IP, etc.) are excluded from auto-detection and treated as plain columns. The exclusion list (valueTypePackages) can be extended at startup.

type Post struct {
    Entity[int64]
    AuthorID int64   // column "author_id" via SchemaNameTransformer
    Author   Person  // auto-detected as BelongsTo, FK = "author_id"
}

type Person struct {
    Entity[int64]
    Posts []Post     // auto-detected as HasMany, FK = "person_id"
}

ManyToMany Join Table & Column Overrides

• Auto-derived join table name: when many2many: tag value is empty, the join table name is automatically derived by sorting both entity table names alphabetically and joining them with an underscore.
• parentKey: / relatedKey: tag options: new tag options allow overriding the join table column names that reference each side's primary key.

type Article struct {
    Entity[int64]
    Tags []Tag `db:"-,many2many:,parentKey:article_pk,relatedKey:tag_pk"`
}

Improvements

• toSnakeCase digit handling: correctly inserts underscores after digit sequences followed by uppercase letters (e.g., Uint64Article → uint64_article).
• Runtime tableNameForType: derives table names from reflect.Type at runtime, respecting both the TableNamer interface and SchemaNameTransformer in all reflection-driven code paths.
• M2M column name resolution: ManyToMany join column derivation now goes through SchemaNameTransformer consistently across both preload and association/create code paths.

Testing

• 15 new schema tests covering non-struct type errors, missing PK errors, pointer-to-struct unwrapping, string PKs in insert columns, HasOne, ManyToMany, mixed preload, and auto-detected relationships.
• Expanded preload, query builder, and read test coverage with doc comments on all test methods.