Migrator

Create a new migrator instance.

Run the pending migrations at a given path.

Get the migration files eligible to run, respecting registered exclusions.

Get the migration files that have not yet run.

Get list of pending migrations to skip.

Run an array of migrations.

Run "up" a migration instance.

Rollback the last migration operation.

Get the migrations for a rollback operation.

Rollback the given migrations.

Rolls all of the currently applied migrations back.

Reset the given migrations.

Run "down" a migration instance.

Run a migration inside a transaction if the database supports it.

Pretend to run the migrations.

Get all of the queries that would be run for a migration.

Run a migration method on the given connection.

Sets the coroutine Context key so Schema/DB facade calls inside the migration body resolve to the correct connection. This handles both the migrator's --database override and per-migration $connection properties. Context is the single source of truth for the scoped default — no worker-global state is mutated.

Resolve a migration instance from a file.

Resolve a migration instance from a migration path.

Determine if the migration at the given path should be skipped.

Get the distinct connections declared by the migrations at the given paths.

Generate a migration class name based on the migration file name.

Get all of the migration files in a given path.

Require in all the migration files in a given path.

Get the name of the migration.

Register a custom migration path.

Paths remain on the shared migrator and affect every subsequent run.

Get all of the custom migration paths.

Set the pending migrations to skip.

The list persists in a static property for the worker lifetime and applies to every subsequent migration run.

Get the default connection name.

Resolve a connection name through the "migrations_connection" config key.

Allows pooled connections (PgBouncer, pgdog, Neon, Supabase, etc.) to declare an unpooled sibling that migration operations should route through. Session state required by migrations — advisory locks, LOCK TABLE, temp tables — is incompatible with transaction-pooling mode.

The target must be terminal: it may omit migrations_connection or reference itself, but it may not route migrations to another connection.

When $name is null, falls back to the "effective default connection" — the current coroutine's Context override first, then the configured default (database.default). This mirrors DatabaseManager::getDefaultConnection() so programmatic flows that wrap migrations in DB::usingConnection() or DatabaseManager::setDefaultConnection() route through the scoped default.

Defensively passes the name through when the container has no "config" binding so unit tests that construct Migrator without a booted framework still work.

Execute the given callback using the given connection as the default connection.

Snapshots the prior coroutine Context value and stored migrator connection independently, then restores both directly in finally. The two values can differ, so routing restoration through setConnection() would collapse them.

Set the default connection name.

Honors the target connection's "migrations_connection" config key. The swapped name is propagated via coroutine Context (so Schema/DB facade calls during migration resolve correctly), via the repository source (so the migrations table lands on the same target), and via the stored connection (so subsequent resolveConnection() calls use it). The migrator and database repository keep this mutable state in the current coroutine.

Resolve the database connection instance.

Applies the "migrations_connection" swap so per-migration connection overrides (via Migration::getConnection()) are also routed to the unpooled sibling when the named connection opts in.

Set a connection resolver callback.

The callback persists in a static property for the worker lifetime and runs on every migration's connection resolution.

Get the schema grammar out of a migration connection.

Get the migration repository instance.

Determine if the migration repository exists.

Determine if any migrations have been run.

Delete the migration repository data store.

Get the file system instance.

Set the output implementation that should be used by the console.

Write to the console's output.

Determine whether the given migration event has listeners.

Fire the given event for the migration.

Fetches the dispatcher from the container each time to ensure Event::fake() and other runtime swaps are respected (the Migrator may be constructed before fakes are set up).

Flush all static state.