Data Transfer

functional

Copy table data to another connection, database, and schema with live progress.

A wizard for copying table data to another connection - including across engines, with column types mapped for the target and any lossy mappings reported in the summary. Pick source tables from a checkbox list, the target connection, database, and schema, then a write mode, Create table, Append, Truncate then insert, or Upsert on primary key, and an identifier casing (Keep names, lower_case, or UPPER_CASE, handy when moving between SQL Server and PostgreSQL naming conventions).

This is the one Tools dialog that runs directly rather than scripting first, it is moving rows, not changing definitions - so it is gated accordingly: it refuses production-marked targets, connections with writes disabled, relay-routed targets, and a target identical to the source. Selected tables copy in foreign-key dependency order with a live "Copying table, N rows (i/total)" readout, a Stop button, and a per-table row-count summary at the end. The dialog can also pop out into its own window so a long transfer does not hold the workspace hostage.

Targets: SQL Server and Azure SQL, PostgreSQL, MySQL/MariaDB, SQLite, libSQL / Turso (one server transaction per table), DuckDB, ClickHouse (where new tables are created as MergeTree), and Oracle Database (Oracle types such as VARCHAR2, NUMBER(p,s), TIMESTAMP, CLOB and BLOB, with a MERGE for upsert). Redis has no tables to transfer; use DUMP / RESTORE or MIGRATE in a query tab.

SQLly
SQLly Data Transfer wizardSQLly Data Transfer wizard
Data Transfer: choose source tables and a target connection/database/schema, then run with a progress readout.

Copying tables from SQL Server to PostgreSQL

  1. Right-click the source database and choose Data › Data Transfer.... To start from one table, right-click the table and choose Data › Transfer Data To… instead; that table arrives already checked.
  2. Source profile and Source database start on the connection you came from. Check the tables to copy, say dbo.Customers and dbo.Orders.
  3. Pick the Target profile and Target database, and type the Target schema (it defaults to dbo for SQL Server and public for PostgreSQL targets).
  4. Choose Mode Create table and Casing lower_case, so CustomerId becomes customerid.
  5. Read the summary box. It lists the table count and target, the mode, any type-mapping warnings, and any reason the run is blocked in red.
  6. Press Run transfer. Customers is copied before Orders because Orders references it.

When it finishes, the status bar says Data transfer finished (or failed, or stopped), a notification is added, and a Data Transfer output panel shows the mode, target engine, rows copied per table, and every warning.

Modes

ModeWhat runs on the target
Create tableA CREATE TABLE with mapped column types, NOT NULL where the source has it, and the primary key, then the inserts. Indexes, foreign keys, and defaults are not copied, and identity columns become plain columns (the summary warns). If the table already exists, the create fails and the transfer stops.
AppendInserts into the existing table.
Truncate then insertTRUNCATE TABLE first - or DELETE FROM on an engine without TRUNCATE, such as SQLite - then the inserts.
Upsert on primary keyUpdates matching rows and inserts new ones: MERGE on SQL Server and Oracle, ON CONFLICT ... DO UPDATE on PostgreSQL, SQLite, libSQL and DuckDB, ON DUPLICATE KEY UPDATE on MySQL and MariaDB. Tables without a primary key are skipped with a warning; ClickHouse has no upsert, so use Append or Truncate then insert there.

When writing into an existing SQL Server table that has an identity column, SQLly wraps the inserts in SET IDENTITY_INSERT ... ON / OFF so the source values are kept. Rows go in as parameterized multi-row inserts of up to 1,000 rows each, fewer when the table is wide, to stay inside each engine's parameter limit.

Failures, Stop, and transactions

  • Each table is copied in its own transaction on the target. ClickHouse has no transactions, so there each insert stands alone.
  • If a statement fails, the current table's transaction is rolled back and the transfer stops. Tables that already finished stay committed; the summary names the table and the error.
  • Stop (the run button while running) cancels the source query and rolls back the table in progress. Closing the dialog during a run stops it rather than hiding it.
  • Type-mapping notes take the form table.column: source type → target type (note), where the note is widened, narrowed, encoded as ..., or unknown type - stored as text.

When Run is disabled

The summary box shows the reason in red: Data transfer needs a direct connection to the target (relay-routed targets), Writes are disabled on ... by its safety settings, Writes are disabled on ... because it is marked production, or Source and target are the same table set. Data Transfer is a desktop feature and is not offered in the browser preview.