Manual › Data tools

Import, export and back up data

Use an export for a file of rows, a data transfer for a mapped copy, or a dump for a database backup. You need permission to read the source and, for imports and restores, to write the destination.

Choose a tool

Your goal	Use
Paste selected values into a spreadsheet	Copy from the result grid
Save query rows as CSV or JSON	Export result
Load a file into a table	Import from the table's explorer menu
Copy a table to another connection	Transfer data… in the command palette
Preserve database structure and data for recovery	Dump and restore

Copying into another grid creates a changeset: review the column mapping in Paste as new rows…, then use Apply to write the rows. See Results and data editing.

Export query results

Run the query and open its result.
Open the export action from the result strip or command palette.
Choose the file format and destination.
Review the sample and format options.
Start the export and watch its progress.

You can export CSV, TSV, a JSON array, JSON Lines, SQL INSERT statements or a Markdown table. Delimited formats let you choose the header, delimiter, quote character, NULL text and line endings. SQL export also needs the destination table name.

For a partially loaded result, choose between its loaded rows and rerunning the whole statement. The whole-statement option requires a read statement. Rerunning reads the database again, so the export can reflect changes made since the result appeared.

Exports use raw values unless Use for export is enabled under Data formats in Preferences. A successful export replaces its temporary output with the finished file; cancelling or failing removes the partial output.

Export dialog with output format and file options

Import a file into a table

Right-click the destination table in the explorer and open its import action.
Choose a CSV, TSV, JSON array or JSON Lines file.
Check the sample rows and detected columns.
Map each file column to its destination column. Leave automatic IDs unmapped when the database should generate them.
Review the batch and error options, then start the import.

An unmapped destination column uses its default or nullability rules. Conversion errors identify the affected row and column. JSON Lines expects one object per line; a JSON array expects one array containing objects.

CSV import preview showing sample orders and column mapping

Copy between connections

Open the command palette with ⌘⇧P and choose Transfer data…. The wizard keeps your choices when you move back between steps.

Data transfer wizard with source and target choices

In Source, choose a table, read-only query or file. Check the preview before continuing. File sources include CSV, TSV, JSON, JSON Lines, Excel .xlsx sheets and Parquet.
In Target, choose an existing table, a new table or a file. For a new table, enter its name and review the generated column types.
In Mapping, choose the destination for each source column, or skip it. You can trim text, change case, turn blank text into NULL, supply a constant or request a type conversion. This step appears for table targets.
In Options, choose the batch size and write behavior.
In Review, check both endpoints, column mappings, warnings and any new-table DDL. Choose Run transfer.

Table and query sources cannot be copied into a table on the same saved connection through this wizard. Use INSERT … SELECT in an editor for that case. A table target needs a transaction-capable, writable session with no open transaction.

New-table type conversion can lose information. For example, a timestamp with a time zone may become a UTC timestamp without its zone, and an unfamiliar type may become text. Review every warning and the editable DDL before running.

Control writes and rejected rows

Option	Effect
Rows per batch	Sets the requested batch size; the engine writes at most 10,000 rows per batch
Commit after each batch (a cancel keeps the committed batches)	Keeps completed batches if you stop later; otherwise one commit covers the transfer
Delete the target's rows first	Removes existing data before loading; requires confirmation
Update rows that exist, by key	Updates matching rows instead of inserting duplicates; choose the identifying key columns
Switch foreign keys off during the load	Temporarily relaxes foreign-key checks; requires confirmation
Skip bad rows and report them	Records rejected rows instead of failing on the first one

When skipping bad rows, choose a Rejected rows file. It contains the row position, reason and source values. The dialog shows the first 100 rejections, and the transfer stops at 1,000 rejected rows. Keep this file private if the source contains sensitive values.

Use Stop transfer to cancel. Read the final counts of rows read, written and rejected before retrying, especially if you enabled per-batch commits.

Dump a database

Start from a database node in the explorer, or SQLite's main schema, and choose the dump action.

Database	Backup formats	Native tools
PostgreSQL	SQL, custom archive, tar, directory	`pg_dump`
MySQL and MariaDB	SQL	`mysqldump`
SQLite	SQL or database file	`sqlite3` for SQL; online backup for a database file

Choose the output path and format.
Select schema-only or data-only output if needed.
Review the database-specific options and displayed command. PostgreSQL directory output supports parallel jobs; MySQL offers routines, events, triggers and a consistent snapshot.
Choose Dump and follow the log.

sqlr finds native tools on PATH and in common installation directories. If discovery fails, install the database's client tools or set Executable in the dialog. Passwords are passed through the tool's environment or a private temporary file, and SSH connections use the existing tunnel.

SQLite dump dialog with output file and backup format options

Restore a backup

Open the restore action on the destination database.
Select the backup and its format.
Review the destination and options, including whether existing objects will be removed.
Choose Restore and answer any confirmation.
Read the outcome and refresh the explorer after completion.

PostgreSQL SQL scripts use psql; archive formats use pg_restore. MySQL SQL uses mysql. SQLite accepts a SQL script or a database-file backup.

Warning: Restoring a SQLite database file replaces the destination's contents. Cancelling a restore stops the tool but may leave changes it already wrote. Check the destination before retrying.

These dialogs configure a single run. They do not currently offer a Tasks panel or a Save as task action.

Results and data editing: copy rows and apply pasted changesets.
Profiles and transactions: understand write protection and commit modes.
Schema tools: compare structures or generate sample data.

← Profiles and transactions · Schema tools →