Resolving database upgrade issues with multiple database drivers by Copilot · Pull Request #965 · objectstack-ai/framework

Copilot · 2026-03-23T07:55:34Z

Pull request created by AI Agent

…MySQL database auto-creation The core issue was a naming mismatch between: - FQN in SchemaRegistry: sys__user (double underscore) - Physical table name from ObjectSchema.create(): sys_user (single underscore) - Table name expected by auth adapter (SystemObjectName): sys_user Changes: 1. SchemaRegistry.getObject() - Added fallback to match by tableName property 2. ObjectQL.resolveObjectName() - Returns tableName instead of FQN 3. ObjectQLPlugin.syncRegisteredSchemas() - Uses tableName for DDL operations 4. SqlDriver.ensureDatabaseExists() - Extended for MySQL support 5. SqlDriver.createDatabase() - Added MySQL-specific database creation Co-authored-by: hotlong <50353452+hotlong@users.noreply.github.com> Agent-Logs-Url: https://github.com/objectstack-ai/spec/sessions/79aaa91e-29db-42ee-8a83-2041d708a8f7

…MySQL config handling - Removed unnecessary `(... as any)` type casts since ServiceObject already includes tableName - Fixed syncRegisteredSchemas to pass original obj instead of spreading with name override - Fixed MySQL createDatabase to use destructuring instead of setting database: undefined Co-authored-by: hotlong <50353452+hotlong@users.noreply.github.com> Agent-Logs-Url: https://github.com/objectstack-ai/spec/sessions/79aaa91e-29db-42ee-8a83-2041d708a8f7

vercel · 2026-03-23T07:55:39Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
objectstack-play	Ready	Preview, Comment	Mar 23, 2026 7:57am

1 Skipped Deployment

Project	Deployment	Actions	Updated (UTC)
spec	Ignored		Mar 23, 2026 7:57am

Copilot

Pull request overview

This PR addresses database upgrade/sync problems across multiple SQL drivers by aligning schema/table naming (FQN vs physical table name) and expanding automatic database creation beyond PostgreSQL.

Changes:

Extend SqlDriver.ensureDatabaseExists() / createDatabase() to handle MySQL and explicitly no-op for SQLite.
Add tableName-based resolution in SchemaRegistry.getObject() and prefer tableName in ObjectQL driver operations / schema sync.
Add tests for tableName resolution + syncSchema argument behavior; update CHANGELOG.md.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
packages/plugins/driver-sql/src/sql-driver.ts	Add MySQL support to DB existence checks/DB creation; skip SQLite.
packages/plugins/driver-sql/src/sql-driver-schema.test.ts	Add a SQLite “ensureDatabaseExists is a no-op” test.
packages/objectql/src/registry.ts	Add `tableName` fallback in `SchemaRegistry.getObject()`.
packages/objectql/src/registry.test.ts	Add tests for resolving objects by `tableName`.
packages/objectql/src/plugin.ts	Use `obj.tableName` (when present) for `syncSchema()` DDL target.
packages/objectql/src/plugin.integration.test.ts	Assert `syncSchema()` is called with `tableName` rather than FQN.
packages/objectql/src/engine.ts	Prefer `schema.tableName` when resolving object names for driver operations.
CHANGELOG.md	Document fixes around naming mismatch and MySQL DB auto-creation.

Copilot · 2026-03-23T08:01:38Z

packages/objectql/src/plugin.ts

+      // Use the physical table name (e.g., 'sys_user') for DDL operations
+      // instead of the FQN (e.g., 'sys__user'). ObjectSchema.create()
+      // auto-derives tableName as {namespace}_{name}.
+      const tableName = obj.tableName || obj.name;
+
      try {
-        await driver.syncSchema(obj.name, obj);
+        await driver.syncSchema(tableName, obj);
        synced++;


syncRegisteredSchemas() now passes tableName as the first argument to driver.syncSchema(), but SqlDriver.syncSchema() currently ignores that argument and uses schema.name when calling initObjects(). Since SchemaRegistry.registerObject() overwrites schema.name with the FQN (e.g. sys__user), SqlDriver will still create/alter the FQN table instead of the physical tableName (e.g. sys_user). To make this change effective, either update SqlDriver.syncSchema to use the object parameter as the physical table name, or pass a schema copy with name rewritten to tableName for drivers that derive DDL names from schema.name.

Copilot · 2026-03-23T08:01:39Z

packages/plugins/driver-sql/src/sql-driver.ts

    const adminKnex = knex(adminConfig);
    try {
-      await adminKnex.raw(`CREATE DATABASE "${dbName}"`);
+      if (this.isPostgres) {
+        await adminKnex.raw(`CREATE DATABASE "${dbName}"`);
+      } else if (this.isMysql) {
+        await adminKnex.raw(`CREATE DATABASE IF NOT EXISTS \`${dbName}\``);
+      }


createDatabase() interpolates dbName directly into raw SQL (CREATE DATABASE "${dbName}" / ``CREATE DATABASE IF NOT EXISTS `${dbName}```), but dbName comes from the connection string/object and is not validated/escaped. This can fail for empty/undefined names and also risks SQL injection if the connection config can be influenced. Prefer Knex identifier bindings (e.g., `??`) or a strict identifier allowlist validation before executing the raw statement, and explicitly error if `dbName` is missing.

Copilot · 2026-03-23T08:01:39Z

packages/plugins/driver-sql/src/sql-driver.ts

  protected async ensureDatabaseExists() {
-    if (!this.isPostgres) return;
+    // SQLite auto-creates database files — no need to check
+    if (this.isSqlite) return;
+
+    // Only PostgreSQL and MySQL support programmatic database creation
+    if (!this.isPostgres && !this.isMysql) return;

    try {
      await this.knex.raw('SELECT 1');
    } catch (e: any) {
-      if (e.code === '3D000') {
+      // PostgreSQL: '3D000' = database does not exist
+      // MySQL:      'ER_BAD_DB_ERROR' (errno 1049) = unknown database
+      if (
+        e.code === '3D000' ||
+        e.code === 'ER_BAD_DB_ERROR' ||
+        e.errno === 1049
+      ) {
        await this.createDatabase();
      } else {


The new MySQL handling in ensureDatabaseExists() (ER_BAD_DB_ERROR / errno 1049) and the MySQL branch in createDatabase() are not covered by tests. The added SQLite test still passes even with the old behavior (SQLite already skipped the PostgreSQL-only check), so it doesn’t validate the new MySQL upgrade fix. Consider adding unit-level tests that stub knex.raw() to throw MySQL/PG "unknown database" errors and assert createDatabase() is invoked (and not invoked for other error codes).

Copilot · 2026-03-23T08:01:39Z

packages/objectql/src/registry.ts

+    for (const fqn of this.objectContributors.keys()) {
+      const resolved = this.resolveObject(fqn);
+      if (resolved?.tableName === name) {
+        return resolved;


SchemaRegistry.getObject() now adds a second full scan over objectContributors and calls resolveObject() for each entry to match by tableName. Since getObject() is used on every engine operation (resolve + driver selection), resolving by physical table name can become O(n) per call and scale poorly as the number of objects grows. Consider maintaining an index/map from tableName -> fqn (updated on register/unregister/reset) so tableName lookups are O(1) and don’t require merging every object during resolution.

Suggested change

for (const fqn of this.objectContributors.keys()) {

const resolved = this.resolveObject(fqn);

if (resolved?.tableName === name) {

return resolved;

//

// To avoid repeatedly merging objects on every iteration, we derive the

// physical table name directly from the FQN using the convention:

// tableName = `${namespace}_${name}`

// and only resolve the object once we find a match.

for (const fqn of this.objectContributors.keys()) {

const { namespace, shortName } = parseFQN(fqn);

const tableName = `${namespace}_${shortName}`;

if (tableName === name) {

return this.resolveObject(fqn);

Copilot · 2026-03-23T08:01:39Z

packages/objectql/src/engine.ts

  private resolveObjectName(name: string): string {
    const schema = SchemaRegistry.getObject(name);
    if (schema) {
-      return schema.name; // FQN from registry (e.g., 'todo__task')
+      // Prefer the physical table name (e.g., 'sys_user') over the FQN
+      // (e.g., 'sys__user'). ObjectSchema.create() auto-derives tableName
+      // as {namespace}_{name} which matches the storage convention.
+      return schema.tableName || schema.name;
    }


Changing resolveObjectName() to return schema.tableName makes the engine’s canonical object identifier become the physical table name for objects that define tableName. This affects more than driver I/O: hook/middleware matching uses context.object string equality, and action keys are built from the object name, so existing code registering hooks/actions against the FQN (e.g. sys__user) will no longer match when operations run as sys_user. Consider keeping FQN as the canonical engine object identifier (for hooks/actions/registry) and passing tableName only at the driver boundary (or carrying both values in OperationContext) to avoid a breaking semantic change.

Copilot AI and others added 2 commits March 23, 2026 07:32

Copilot AI assigned Copilot and hotlong Mar 23, 2026

Copilot created this pull request from a session on behalf of hotlong March 23, 2026 07:55 View session

hotlong marked this pull request as ready for review March 23, 2026 07:56

Copilot AI review requested due to automatic review settings March 23, 2026 07:56

github-actions bot added documentation Improvements or additions to documentation tests size/m labels Mar 23, 2026

Copilot started reviewing on behalf of hotlong March 23, 2026 07:56 View session

vercel bot deployed to Preview – objectstack-play March 23, 2026 07:57 View deployment

Copilot AI reviewed Mar 23, 2026

View reviewed changes

hotlong merged commit fb97582 into main Mar 23, 2026
17 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Resolving database upgrade issues with multiple database drivers#965

Resolving database upgrade issues with multiple database drivers#965
hotlong merged 2 commits intomainfrom
copilot/fix-database-auto-upgrade-issue

Copilot AI commented Mar 23, 2026

Uh oh!

vercel bot commented Mar 23, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Copilot AI Mar 23, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

-    for (const fqn of this.objectContributors.keys()) {
-      const resolved = this.resolveObject(fqn);
-      if (resolved?.tableName === name) {
-        return resolved;
+    //
+    // To avoid repeatedly merging objects on every iteration, we derive the
+    // physical table name directly from the FQN using the convention:
+    //   tableName = `${namespace}_${name}`
+    // and only resolve the object once we find a match.
+    for (const fqn of this.objectContributors.keys()) {
+      const { namespace, shortName } = parseFQN(fqn);
+      const tableName = `${namespace}_${shortName}`;
+      if (tableName === name) {
+        return this.resolveObject(fqn);

Conversation

Copilot AI commented Mar 23, 2026

Uh oh!

vercel bot commented Mar 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 23, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

vercel bot commented Mar 23, 2026 •

edited

Loading