feat: SQLite DB adapter (sqlite3: ^3.5.1)

[gmpassos]

Adds DBSQLiteAdapter, a fourth DBSQLAdapter alongside PostgreSQL, MySQL and the in-memory one — an embedded, file-backed (or in-memory) SQL backend that needs no server and no native library to install: the sqlite3 package bundles SQLite 3.53.4 via Dart's build hooks.

import 'package:bones_api/bones_api_db_sqlite.dart';

var adapter = DBSQLiteAdapter('/var/lib/myapp/db.sqlite', generateTables: true);
var memory  = DBSQLiteAdapter(':memory:', generateTables: true);

db:
  sqlite:
    path: /var/lib/myapp/db.sqlite
    generateTables: true

Registered as sqlite, sqlite3, sql.sqlite, sql.sqlite3.

Reviewing this PR

Read the two commits separately.

• 0004437 chore: bump minimum Dart SDK to 3.10.0 — mostly mechanical, ~3k lines. sqlite3 ^3.5.1 requires SDK >= 3.10.0, and dart format picks its style from the resolved language version, so raising the floor reflows most of the package. master is format-clean under the old constraint and CI runs dart format --set-exit-if-changed, so the reflow has to land with the bump. Regenerated *.reflection.g.dart for the same reason. No behavior changes.
• 112c47d feat: SQLite DB adapter — the actual change.

Verified against the real engine

Each dialect decision below was checked by executing the generated SQL, not inferred:

• INTEGER PRIMARY KEY AUTOINCREMENT for auto-assigning IDs. SQLite has no SERIAL, and the generic SERIAL PRIMARY KEY is silently accepted (unknown type names get NUMERIC affinity) while leaving every ID NULL. AUTOINCREMENT is what matches PostgreSQL/MySQL semantics — a plain INTEGER PRIMARY KEY reuses the ID of a deleted row (measured: [1, 2] vs [1, 3]).
• ENUM emulated with VARCHAR CHECK (col IN (...)).
• ON CONFLICT DO NOTHING rather than MySQL's INSERT IGNORE.
• LIMIT -1 OFFSET n — a bare OFFSET is a syntax error, and the default offsetMaxLimitValue (2^64-1) overflows SQLite's int64.
• No implicit FK index; constraintSupport: false (no ALTER TABLE ... ADD CONSTRAINT); PRAGMA foreign_keys = ON (off by default, per connection); WAL + busy_timeout for file databases.
• Introspection via pragma_table_info / pragma_foreign_key_list instead of an information_schema.
• sqlite3 binds only int/double/String/List<int>/null, so bool, DateTime, Time, BigInt, DynamicInt, Decimal and enums are normalized at a single choke point.
• Constraint errors map to EntityFieldInvalid. SQLite's message names only table.column, so the rejected value is recovered from the statement's bound parameters (PostgreSQL and MySQL get it from the driver message).

One shared native handle

sqlite3 is a synchronous driver and SQLite allows a single writer, so the adapter shares one handle across every pooled connection wrapper. A second handle is a hazard rather than a benefit: one blocking on a lock stalls the isolate holding it, so the holder can never commit — measured, a second writer on a WAL file blocked for the full busy_timeout and then failed with database is locked. And there is no I/O wait to overlap.

Capping the pool at 1 would not have sufficed: Pool.catchFromPool force-creates elements beyond maxConnections under contention. Transactions opened while one is already active nest as a SAVEPOINT.

(Shared-cache in-memory URIs are not an option either — the bundled SQLite is compiled with OMIT_SHARED_CACHE.)

Changes outside the adapter

• New SQLDialect.returningAcceptsTableWildcard, defaulting to true so PostgreSQL/MySQL/memory emit exactly what they did before. SQLite rejects the table-qualified wildcard in DELETE ... RETURNING ("RETURNING may not use TABLE.* wildcards") and needs a bare RETURNING *. The MySQL temporary-table fallback isn't usable here — its CREATE TEMPORARY TABLE ... AS ( SELECT ... ) is invalid SQLite.
• Shared create-table assertion now also accepts AUTOINCREMENT after PRIMARY KEY. Strict generalization; no other adapter's expectation changes.
• README: documents the adapter, and corrects the class names in the SQLAdapter list, which still used the pre-DB* names (PostgreSQLAdapter → DBPostgreSQLAdapter, etc.).

⚠️ Breaking

Minimum Dart SDK is now 3.10.0 (was 3.7.0).

Testing

The adapter runs the same shared entity suite as PostgreSQL and MySQL (runAdapterTests) over the same reflection/generateTables/checkTables/populateSource matrix, plus an in-memory pass — and needs no Docker. New APITestConfigSQLite in package:bones_api/bones_api_test_sqlite.dart; new sqlite tag in dart_test.yaml.

Gate	Result
dart test test/bones_api_entity_db_sqlite_test.dart	91 passed
dart test --exclude-tags docker	778 passed
dart test --platform chrome	519 passed
dart analyze --fatal-infos --fatal-warnings .	clean
dart format -o none --set-exit-if-changed .	clean
dart pub publish --dry-run	0 warnings
dependency_validator	no issues

Not run locally: the Docker-tagged PostgreSQL and MySQL suites — no Docker daemon on this machine. They exercise generateDeleteSQL, the one shared function this PR touches, so they are worth confirming in CI. The new dialect flag defaults to the existing behavior, and the SQLDialect defaults test asserts that for the generic/PostgreSQL/MySQL dialects.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 71.75227% with 561 lines in your changes missing coverage. Please review.
✅ Project coverage is 68.12%. Comparing base (e92bc3e) to head (ca70200).

Files with missing lines	Patch %	Lines
lib/src/bones_api_entity.dart	54.09%	101 Missing ⚠️
lib/src/bones_api_entity_db_sql.dart	62.22%	85 Missing ⚠️
lib/src/bones_api_entity_db_sqlite.dart	81.05%	83 Missing ⚠️
lib/src/bones_api_server.dart	38.61%	62 Missing ⚠️
lib/src/bones_api_db_module.dart	10.00%	27 Missing ⚠️
lib/src/bones_api_entity_db_relational.dart	67.18%	21 Missing ⚠️
lib/src/bones_api_condition.dart	60.78%	20 Missing ⚠️
lib/src/bones_api_entity_db_object_directory.dart	40.74%	16 Missing ⚠️
lib/src/bones_api_entity_reference.dart	65.21%	16 Missing ⚠️
lib/src/bones_api_sql_builder.dart	82.66%	13 Missing ⚠️
... and 24 more

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #148      +/-   ##
==========================================
+ Coverage   67.77%   68.12%   +0.35%     
==========================================
  Files          64       66       +2     
  Lines       21607    22114     +507     
==========================================
+ Hits        14644    15066     +422     
- Misses       6963     7048      +85     

Flag	Coverage Δ
unittests	68.12% <71.75%> (+0.35%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.