Output formats

Pick a format with --format / -F on the command line, \T in the REPL, or table_format in the config. \T alone lists them.

Table formats

For reading. They respect max_field_width, switch to vertical layout when expanded = "auto" and the table is too wide, and print a status line (3 rows · 0.4 ms) in the REPL.

Format	Also accepted	Looks like
rounded (default)	fancy_grid, fancy, round	╭──┬──╮ with rounded corners
unicode	psql_unicode, box	┌──┬──┐ single lines
double	double_grid	╔══╦══╗ double lines
ascii	grid, mysql	+----+----+, like the mysql client
psql		Like psql: a -+- rule under the header, no outer border
simple		A ---- rule under the header
minimal		Column gaps and a ─ rule under the header
plain		Columns separated by spaces only
markdown	github, md, pipe	A GitHub-flavoured Markdown table
vertical	expanded	One name: value line per column, one block per row, like mysql's \G

The same result in four of them:

rounded                   ascii                     markdown              psql
╭────────┬────────╮       +--------+--------+       | status | orders |    status | orders
│ status │ orders │       | status | orders |       |--------|-------:|   --------+--------
├────────┼────────┤       +--------+--------+       | new    |    666 |    new    |    666
│ new    │    666 │       | new    |    666 |       | paid   |    667 |    paid   |    667
├────────┼────────┤       +--------+--------+
│ paid   │    667 │       | paid   |    667 |
╰────────┴────────╯       +--------+--------+

In the boxed formats (rounded, unicode, double, ascii) a line separates the rows. Set row_lines = false in the config for a more compact table.

Machine formats

For other programs. They're never coloured, cut or turned vertical, and print only the data.

Format	Also accepted	Output
csv		Header row, then one line per row. NULL is an empty field.
tsv		Header row, tab-separated. Tabs, newlines and backslashes are escaped (\t, \n, \\). NULL is null_string.
json		An array of objects. Numbers and booleans stay typed, JSON columns are embedded, NULL is null.
jsonl	jsonlines, ndjson	One JSON object per line
html		A <table> with <thead> and <tbody>
sql-insert	sql_insert	INSERT INTO table (…) VALUES (…); per row
sql-update	sql_update	UPDATE table SET … WHERE first_column = …; per row

When quarry's output goes to a pipe or a file and you haven't chosen a format, it uses tsv.

In the REPL and in scripts, sql-insert and sql-update use the placeholder table name table; replace it with the real name. The TUI's export fills in the real name.

SQLite .mode

On SQLite, .mode accepts sqlite3's names too:

.mode	quarry format
line	vertical
list	plain
tabs	tsv
column	simple
table	ascii
insert	sql-insert
qbox	unicode

Vertical output

Vertical output shows one record at a time, one line per column, which suits wide rows. With expanded = "auto" (the default) a table that's wider than the terminal switches to it on its own; \x off (or expanded = "off") keeps the table and lets the pager scroll it sideways.

In the boxed formats (rounded, unicode, double, ascii) the records keep the table's look, each one numbered in the rule above it:

╭─ 1 ────────┬─────────────────────
│         id │ 1
│      email │ user1@example.com
│       name │ User 1
├─ 2 ────────┼─────────────────────
│         id │ 2
│      email │ user2@example.com
│       name │ User 2
╰────────────┴─────────────────────

\x cycles vertical output off → on → auto, and \x on, \x off, \x auto set it directly. End a single statement with \G instead of ; to show just that result vertically.

The other formats, and -F vertical (for scripts), use mysql's layout:

$ quarry shop.db -F vertical -e "select id, email, name from users where id = 1"
*************************** 1. row ***************************
   id: 1
email: user1@example.com
 name: User 1