How to Export Data from Dolt: Complete Guide to Table and Schema Export

Dolt provides two primary CLI commands for exporting data: dolt table export for row data (CSV, JSON, Parquet, etc.) and dolt schema export for DDL statements, both leveraging a pluggable mvdata pipeline architecture.

Dolt is a version-controlled SQL database that combines Git-like versioning with MySQL-compatible tables. When you need to extract data from a dolthub/dolt repository for analysis, migration, or backup purposes, the CLI offers robust export capabilities backed by a sophisticated data movement pipeline written in Go. This guide explains how to export both table rows and schema definitions using the official source code as reference.

Dolt Export CLI Commands Overview

Dolt separates data export into two distinct commands based on what you need to extract:

• dolt table export – Exports the rows of a table to various formats including CSV, PSV, JSON, JSONL, SQL, and Parquet
• dolt schema export – Exports the DDL (CREATE TABLE statements) for one or all tables

Both commands share a common architecture built on Dolt’s mvdata (move-data) package, providing consistent argument parsing, format detection, and streaming data movement.

How dolt table export Works

The dolt table export command, implemented in go/cmd/dolt/commands/tblcmds/export.go, follows a pipeline architecture that streams data from the Dolt storage engine to your chosen destination.

Command Syntax and Flags

The basic syntax follows this pattern:

dolt table export [-f] [--file-type <type>] <table> <file>

Key parameters include:

• -f – Force overwrite of existing files (skips the checkOverwrite validation)
• --file-type – Explicitly set the format when the file extension is ambiguous
• <table> – The source table name in the current database
• <file> – Destination path, or - for stdout streaming

The Export Pipeline Architecture

According to the source code in tblcmds/export.go, the export process follows these steps:

1. Argument parsing – The command builds an argparser.ArgParser that validates flags such as -f and --file-type (lines 78‑86)
2. Destination resolution – The user-supplied path and optional --file-type are turned into a mvdata.DataLocation via mvdata.NewDataLocation. This object determines the output format and whether the destination is a regular file or a stream (lines 98‑108)
3. Overwrite protection – If -f is omitted, exportOptions.checkOverwrite verifies whether the destination exists and aborts with an error (lines 65‑73)
4. SQL engine setup – A temporary SQL engine (engine.NewSqlEngineForEnv) initializes to read the table’s rows (lines 92‑104)
5. Reader creation – mvdata.NewSqlEngineReader builds a row reader that streams rows from the selected table (line 22)
6. Writer creation – exportOptions.dest.NewCreatingWriter produces a format-specific writer backed by either a file handle or os.Stdout (lines 66‑72)
7. Data movement – mvdata.NewDataMoverPipeline wires the reader and writer together; calling pipeline.Execute() streams the data from source to destination (lines 32‑35)

This architecture supports pluggable output formats including CSV, PSV, JSON, JSONL, SQL, and Parquet.

How dolt schema export Works

The dolt schema export command, implemented in go/cmd/dolt/commands/schcmds/export.go, extracts DDL rather than row data.

When you run this command, Dolt iterates over the specified tables (or all non-system tables via doltdb.GetNonSystemTableNames) and constructs CREATE TABLE statements using dsqle.GetCreateTableStmt. The resulting DDL is written sequentially to the specified file path or stdout.

Key implementation details from schcmds/export.go include:

• Argument parsing occurs at lines 65‑70
• Schema extraction logic at lines 92‑55
• Supports both single-table and full-database export modes

Practical Export Examples

Exporting Table Data to CSV

Export the "users" table to a CSV file (format inferred from extension):

dolt table export users users.csv

Under the hood, this creates a DataLocation with format CsvFile, builds a SQL reader for the users table, and streams rows to the file.

Exporting to JSON Lines with Force Overwrite

Export the "orders" table to JSONL format, overwriting if the file exists:

dolt table export -f --file-type jsonl orders orders.jsonl

The -f flag bypasses the overwrite check, while --file-type jsonl forces the format when the extension might be ambiguous.

Streaming to Standard Output

Export data directly to stdout for piping into other tools:

dolt table export -f mytable -

When the destination is -, the command creates a mvdata.StreamDataLocation that writes CSV (default) directly to stdout. Note that only CSV and PSV formats support streaming.

Exporting Schema DDL

Generate the CREATE TABLE statement for a specific table:

dolt schema export products products_schema.sql

This calls dsqle.GetCreateTableStmt for the "products" table and writes the DDL to the specified file.

Exporting All Schemas

Export DDL for all non-system tables in the database:

dolt schema export all_schemas.sql

With no table argument provided, Dolt iterates over all tables and writes each CREATE TABLE statement sequentially.

Exporting to Parquet Format

Export table data to Apache Parquet (experimental support):

dolt table export --file-type parquet mytable mytable.parquet

The mvdata pipeline recognizes Parquet as a valid format type and instantiates the appropriate writer via NewCreatingWriter.

Key Implementation Files

Understanding these source files helps when building custom tools or debugging export behavior:

• go/cmd/dolt/commands/tblcmds/export.go – Implements dolt table export including argument parsing, destination handling, and pipeline orchestration (lines 78‑108, 32‑35)
• go/cmd/dolt/commands/schcmds/export.go – Implements dolt schema export for DDL extraction (lines 65‑70, 92‑55)
• go/libraries/doltcore/mvdata/data_loc.go – Defines DataLocation, format inference logic, and writer factories used by both export commands
• go/libraries/doltcore/mvdata/data_loc_test.go – Unit tests illustrating supported formats (csv, psv, json, jsonl, parquet)

Summary

• Dolt provides two export commands: dolt table export for row data and dolt schema export for DDL
• The mvdata package provides a unified pipeline architecture using DataLocation, NewSqlEngineReader, and NewDataMoverPipeline
• Supported export formats include CSV, PSV, JSON, JSONL, SQL, and Parquet
• Use the -f flag to force overwrite existing files
• Export to stdout using - as the destination (CSV and PSV only)
• Schema export uses dsqle.GetCreateTableStmt to generate standard SQL DDL

Frequently Asked Questions

Does Dolt support exporting to Parquet format?

Yes, Dolt supports Apache Parquet export via dolt table export --file-type parquet <table> <file>. The mvdata pipeline in go/libraries/doltcore/mvdata/data_loc.go recognizes Parquet as a valid format type and instantiates the appropriate binary writer. This feature is marked as experimental in current versions.

Can I export Dolt data to standard output for piping?

Yes, specify - as the destination file to stream output to stdout: dolt table export mytable -. Only CSV and PSV formats support streaming via mvdata.StreamDataLocation. This is useful for piping data directly into other command-line tools without creating intermediate files.

What is the difference between dolt table export and dolt schema export?

dolt table export extracts the actual row data from tables into formats like CSV or JSON, while dolt schema export extracts only the DDL (CREATE TABLE statements) defining the table structures. The former uses mvdata.NewSqlEngineReader to stream rows, whereas the latter calls dsqle.GetCreateTableStmt to generate SQL definitions.

How do I force overwrite an existing export file?

Add the -f or --force flag to your export command. Without this flag, the exportOptions.checkOverwrite function in tblcmds/export.go (lines 65‑73) checks if the destination file exists and aborts with an error to prevent accidental data loss. The -f flag bypasses this safety check.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}