Amazon Redshift SQL Formatter

Redshift is PostgreSQL underneath, with its own additions

Redshift inherits most of its lexical rules from PostgreSQL 8, and PasteKit’s Amazon Redshift mode follows them:

• A single quote inside a string is written twice: 'it''s'. A backslash is not treated as an escape, so 'it\'s' is reported as an unterminated string.
• "table" or "Order Date" in double quotes is an identifier. That is how you query system views with reserved column names, as in svv_table_info.
• #recent_orders is read as one name, so temporary tables written with the hash prefix are not split.
• $1, $2 are the positional parameters used by prepared statements and the Data API. @ has no special meaning.

Choosing Standard SQL or MySQL instead would change these rules. MySQL mode, for example, reads #recent_orders as the start of a comment, so the rest of that line is never formatted.

DDL with distribution, sort keys and compression

CREATE TABLE statements are where Redshift differs most from other warehouses. The column list is indented one column per line, and arguments get normalised spacing: identity(1,1) comes out as IDENTITY (1, 1) and decimal(12,2) as DECIMAL(12, 2).

Two layout quirks are worth knowing. Each ENCODE clause is moved onto its own line under the column it belongs to, so customer_id INTEGER NOT NULL is followed by ENCODE AZ64 one line down. After the closing bracket, DISTSTYLE, DISTKEY and COMPOUND SORTKEY stay together on one line, and only some of those words are upper-cased (diststyle KEY distkey (...)). The DDL is still valid; tidy the casing by hand if it matters to you.

UNLOAD, COPY and maintenance commands

UNLOAD ('select ...') takes its query as a quoted string, so the inner SQL stays exactly as written, doubled quotes and all. The TO 's3://...' target, IAM_ROLE and FORMAT AS PARQUET follow on the closing line, and PARTITION BY starts a new one. If you want the unloaded query formatted too, paste it separately, format it, and put it back inside the quotes (doubling any single quotes).

COPY sales.orders FROM 's3://...' places FROM on its own line, followed by the credentials and options. ANALYZE and VACUUM format as one-liners, though VACUUM DELETE ONLY is split after VACUUM.

Words that Redshift treats as keywords are upper-cased even when you use them as column names. A column called region becomes REGION because it is also a COPY option. Redshift folds unquoted names to lower case, so this does not change what the query does; switch Keyword case to As written if the capitals bother you.

What will not format, and useful settings

Stored procedures written as CREATE PROCEDURE ... AS $$ ... $$ LANGUAGE plpgsql are rejected with "$$" is not valid Amazon Redshift syntax. Format the SQL statements from the procedure body on their own instead. JDBC-style ? placeholders are rejected too; use $1 or a sample value.

For long revenue and cohort queries, try Commas: Leading with Function case: UPPER, so aggregates like SUM, LISTAGG and DATEADD stand out from column names. Minify is handy before pasting SQL into a Lambda or Glue job string. For general conventions, see the SQL style guide.

Examples

create table sales.orders (order_id bigint identity(1,1), customer_id integer not null encode az64, order_date date not null encode az64, status varchar(20) encode zstd, amount decimal(12,2) encode az64) diststyle key distkey (customer_id) compound sortkey (order_date, customer_id)

CREATE TABLE sales.orders (
  order_id BIGINT IDENTITY (1, 1),
  customer_id INTEGER NOT NULL
  ENCODE AZ64,
  order_date date NOT NULL
  ENCODE AZ64,
  status VARCHAR(20)
  ENCODE ZSTD,
  amount DECIMAL(12, 2)
  ENCODE AZ64
) diststyle KEY distkey (customer_id) compound sortkey (order_date, customer_id)

unload ('select order_id, customer_id, amount, order_date from sales.orders where order_date >= ''2026-01-01''') to 's3://acme-exports/orders/' iam_role 'arn:aws:iam::123456789012:role/RedshiftUnload' format as parquet partition by (order_date)

UNLOAD (
  'select order_id, customer_id, amount, order_date from sales.orders where order_date >= ''2026-01-01'''
) TO 's3://acme-exports/orders/' IAM_ROLE 'arn:aws:iam::123456789012:role/RedshiftUnload' FORMAT AS PARQUET
PARTITION BY
  (order_date)

select customer_id, date_trunc('month', order_date) as month, sum(amount) as revenue, sum(sum(amount)) over (partition by customer_id order by date_trunc('month', order_date) rows unbounded preceding) as lifetime_revenue, listagg(distinct status, ',') within group (order by status) as statuses, nvl(max(discount_code), 'none') as last_code from sales.orders where order_date >= dateadd(month, -12, getdate()) group by 1, 2

SELECT
  customer_id
, DATE_TRUNC('month', order_date) AS month
, SUM(amount) AS revenue
, SUM(SUM(amount)) over (
    PARTITION BY
      customer_id
    ORDER BY
      DATE_TRUNC('month', order_date) rows unbounded preceding
  ) AS lifetime_revenue
, LISTAGG(DISTINCT status, ',') within GROUP (
    ORDER BY
      status
  ) AS statuses
, NVL(MAX(discount_code), 'none') AS last_code
FROM
  sales.orders
WHERE
  order_date >= DATEADD(month, -12, GETDATE())
GROUP BY
  1
, 2

create temp table #recent_orders as select order_id, customer_id, amount from sales.orders where order_date >= current_date - 7;
copy sales.orders from 's3://acme-raw/orders/2026/10/' iam_role 'arn:aws:iam::123456789012:role/RedshiftCopy' format as csv ignoreheader 1 region 'us-east-1';
analyze sales.orders;

CREATE TEMP TABLE #recent_orders AS
SELECT
  order_id,
  customer_id,
  amount
FROM
  sales.orders
WHERE
  order_date >= current_date - 7;

COPY sales.orders
FROM
  's3://acme-raw/orders/2026/10/' IAM_ROLE 'arn:aws:iam::123456789012:role/RedshiftCopy' FORMAT AS CSV IGNOREHEADER 1 REGION 'us-east-1';

ANALYZE sales.orders;

Common errors and how to fix them

Related tools

• SQL formatter
• sql formatter/postgresql
• sql formatter/snowflake
• sql formatter/bigquery
• sql validator
• sql minifier
• CSV to SQL
• sql unbalanced parentheses