Skip to content
QuerybaraQuerybara

Export

  • PostgreSQL
  • MySQL
  • MariaDB

The export wizard writes tables, or the result of a query, to files. Rows stream from the server to the file page by page in the job runner, so the size of the table does not matter and you can keep working.

Export wizard for tables of the shop schema with Parquet chosen as the format, one file per table, the ZSTD codec, and a note that integers, decimals, dates, timestamps and UUIDs keep their types.Export wizard for tables of the shop schema with Parquet chosen as the format, one file per table, the ZSTD codec, and a note that integers, decimals, dates, timestamps and UUIDs keep their types.
Export to CSV, JSON, Excel, SQL or typed Parquet for DuckDB, Spark and pandas.

Export tables

Section titled “Export tables”
  1. In the sidebar or the Objects tab, right-click a table and choose Export…, or right-click a schema (a database on MySQL and MariaDB) and choose Export tables….
  2. Tables: tick the tables to export.
  3. Format: pick the format and its options (see below), and how several tables are written.
  4. Destination: choose Choose file…, or Choose folder… when each table gets its own file, then Export.

On the Destination step, Schedule… sets the same export up to run on a schedule instead, writing a new file each run. A query with parameters cannot be scheduled, since no one is there to fill them in.

To export a query result, run the query in a query tab and choose Export results… above the results. The query runs again in the job runner, so every row is exported, not only the rows loaded in the grid. Statements that write do not offer it, since running them again would write again.

Formats

Section titled “Formats”
Format What it writes
CSV, TSV Delimited text
JSON An array of objects; bigints and decimals stay exact
JSON Lines One JSON object per line
Excel workbook (.xlsx) Typed cells: numbers, booleans, and dates as Excel dates; a worksheet per table
XML An <export> of <table> elements with one <row> per row; NULL is xsi:nil
Parquet Typed columns, one table per file (see below)
SQL INSERT statements Multi-row INSERT statements
SQL with DDL (CREATE TABLE and INSERTs) The CREATE TABLE, indexes and foreign keys, and the rows
HTML page A self-contained page with one table per result
Markdown tables Pipe tables

Parquet

Section titled “Parquet”

Parquet files keep the column types, for tools such as DuckDB, Spark and pandas:

  • integers, as 32-bit or 64-bit integers (MySQL’s unsigned BIGINT as an unsigned one)
  • numeric(p,s) as an exact DECIMAL(p,s), up to 76 digits
  • dates, timestamps (UTC-adjusted for timestamptz), PostgreSQL’s time, UUIDs and JSON with their own types
  • binary values as bytes

Other values stay text as the server wrote it, so no digit is lost: unbounded numeric, money, intervals, arrays, timetz and MySQL’s TIME. A value the column type cannot hold, such as PostgreSQL’s infinity date, fails the export and names the column and the row.

Each table goes into its own file: several tables export one file per table, or into a ZIP archive. Codec picks the page compression: Snappy (fast, read everywhere), the default, ZSTD (smaller files), GZIP (for older readers) or None. Parquet compresses its own pages, so gzip is not offered.

Options

Section titled “Options”
Option Formats
One file per table or One combined file Several tables; combined is not available for CSV, TSV, JSON Lines or Parquet
Header row with the column names CSV, TSV, Excel (bold, frozen, with filters)
Delimiter: comma, semicolon or pipe CSV
Write NULL as CSV, TSV
Decimals as Excel numbers when exact (up to 15 digits); otherwise decimals are text Excel
Pretty-print (indent each object) JSON
Rows per INSERT SQL
DROP TABLE IF EXISTS before each CREATE TABLE SQL with DDL
Codec Parquet
Compression: None, gzip, or a ZIP archive All (no gzip for Excel or Parquet)
Byte order mark (for Excel) and Encoding (UTF-8 or UTF-16LE) Text formats (not Excel or Parquet)

By default, Excel cells keep values exact: integers beyond what a double holds and decimals are written as text.

A failed or cancelled export removes the partial file. Saved settings keeps the wizard’s choices under a name (Save settings…); file paths are never saved.

From the command line

Section titled “From the command line”

querybara export writes the same formats. Several tables go to one file per table in a folder, into one ZIP archive with --zip, or into one file with --one-file (not for CSV, TSV, JSON Lines or Parquet). --codec picks the Parquet codec: snappy (the default), zstd, gzip or none.

Terminal window
querybara export prod --table shop.orders --format csv --out orders.csv
querybara export prod --table shop.orders --table shop.order_items --format sql-ddl --one-file --out shop.sql --gzip
querybara export prod --table shop.orders --table shop.order_items --format xlsx --one-file --out shop.xlsx
querybara export prod --table shop.orders --format parquet --codec zstd --out orders.parquet
querybara export prod --query "select id, email from shop.customers" --format json --out customers.json
Section titled “Related”

Documents Querybara 0.1.1 · built frombc9f5aa