Transform your SQL!
SQL Transformer
formats SQL to more organized and readable code. Heavy focus was made on improving developer productivity and clarity. It works with common table expressions (CTEs) as well as complicated inline SQL.
The code is heavily commented in case you are eager to understand how it works, or want to make updates or contributions. This has been tested with SQL that ranges from tens of lines to thousands of lines of SQL containing complex CTE and inline logic.
This release is ready for broad use on most SQL, and has been tested with SQL that ranges from tens of lines to thousands of lines of SQL containing complex CTE and inline logic. Feedback is welcomed!
- Select your SQL code block.
- Execute
SQL Transformer: Format SQL
from the command palette. - Or, use a keyboard shortcut:
- Windows:
Shift
+Alt
+T
- Mac:
Shift
+Ctrl
+T
- Windows:
There are four user settings for configuring SQL Transformer output:
- Block Comments: boolean for specifying block or dash comments
- This is currently defaulted to block margins
- Targeting dash comments for the default, however it hasn't been tested as heavily yet.
- See style example(s) below
- Caveats with dash comments:
- care has been taken to ensure that comments are on their own separate lines
- that said, any comments that are on the same line as SQL may break your SQL
- be mindful of your code placement and validate that it is separated from your comments
- Outcome Comments: boolean to include/exclude OUTCOME comments on CTE queries
- This is currently defaulted to include OUTCOME comments
- See style example(s) below
- Starting Width: integer for specifying starting select margin
- Current default margin is set to 5
- Setting to 5 or greater allows for aligning the query on the first attribute, while capitalizing on VSCode's margin collapse functionality
- Where One One: boolean to include/exclude 'WHERE 1=1' filters
- Current default is to include
Formatted by SQL Transformer
with multiple single-line comment blocks.
INSERT INTO target_tab
( monthly
, a
, max_b
, max_c
, sum_cnt
, some_case
)
WITH
cte_1 AS (SELECT a.k_1
, a.k_2
, max(a.col) AS max_col
FROM tab_d a
JOIN tab_f b
ON(a.k_1 = b.k_1
AND a.k_2 = b.k_2
AND a.k_3 = b.k_3
AND a.k_4 = b.k_4)
GROUP BY a.k_1
, a.k_2
LIMIT 10
),
cte_2 AS (SELECT a.k
, max(b.v) AS max_v
, avg(c.v) AS avg_v
, min(d.v) AS min_v
, count(e.v) AS cnt
FROM tab_e a
JOIN tab_f b
ON a.k = b.k
LEFT JOIN tab_g c
ON a.k = c.k
LEFT JOIN tab_h d
ON a.k = d.k
LEFT JOIN tab_i e
ON a.k = e.k
GROUP BY a)
/* Outcome */
SELECT from_timestamp(from_utc_timestamp(from_unixtime(cast(dttm/1000 AS bigint)), 'ROK'), 'yyyy-mm') AS monthly
, a.a
/* double-dashed comments will be replaced like C-style comments. */
, max(a.b) AS max_b
/* * */
/* * multi */
/* * line */
/* * comments */
/* * will be transformed into multiple single-line comments. */
, max(a.c) AS max_c
, sum(b.cnt) AS sum_cnt
, nvl(cast(max(CASE WHEN a.cnt > 0
AND b.max_v > 0 THEN 'A'
WHEN a.cnt > 0
AND b.max_v < 0 THEN 'B'
WHEN a.cnt < 0 THEN 'C'
WHEN a.cnt < 0 THEN 'D'
ELSE (SELECT max(CASE WHEN a.k_1 is null
THEN 'A'
WHEN a.k_1 > 'V'
THEN 'A'
ELSE 'B'
END) AS v
FROM cte_1 a
WHERE 1=1
AND a.max_col > 120
AND a.type IN (SELECT type
FROM tab_t)
AND a.type IN('A', 'B', 'C', 'D')
)
END) AS STRING), '') AS some_case
FROM (SELECT a.k
/* Reserved words in comments are treated as comments. */
, b.v
, c.max_v
, c.avg_v
, c.min_v
, c.cnt
FROM tab_a a
JOIN tab_b b
ON(a.k = b.k
AND b.d = 'dd')
/* + shuffle */
LEFT JOIN cte_2 c
ON a.c = c.a
WHERE 1=1
AND a.col_1 >= 'filter'
/* comment */
AND b.col_2 BETWEEN 'fil'
AND 'ter'
AND a.col_2 > 0) a
LEFT JOIN (SELECT a.k
, max(v) AS max_v
FROM tab_c a
WHERE 1=1
AND a.d >= 'filter'
GROUP BY a.k
HAVING count(1) > 0
) b
ON a.k = b.k
GROUP BY 1, 2
ORDER BY 1, 2
;
Formatted by SQL Transformer
with dash comment line blocks.
INSERT INTO target_tab
( monthly
, a
, max_b
, max_c
, sum_cnt
, some_case
)
WITH
cte_1 AS (SELECT a.k_1
, a.k_2
, max(a.col) AS max_col
FROM tab_d a
JOIN tab_f b
ON(a.k_1 = b.k_1
AND a.k_2 = b.k_2
AND a.k_3 = b.k_3
AND a.k_4 = b.k_4)
GROUP BY a.k_1
, a.k_2
LIMIT 10
),
cte_2 AS (SELECT a.k
, max(b.v) AS max_v
, avg(c.v) AS avg_v
, min(d.v) AS min_v
, count(e.v) AS cnt
FROM tab_e a
JOIN tab_f b
ON a.k = b.k
LEFT JOIN tab_g c
ON a.k = c.k
LEFT JOIN tab_h d
ON a.k = d.k
LEFT JOIN tab_i e
ON a.k = e.k
GROUP BY a)
-- Outcome
SELECT from_timestamp(from_utc_timestamp(from_unixtime(cast(dttm/1000 AS bigint)), 'ROK'), 'yyyy-mm') AS monthly
, a.a
-- double-dashed comments will be replaced like C-style comments.
, max(a.b) AS max_b
-- *
-- * multi
-- * line
-- * comments
-- * will be transformed into multiple single-line comments.
, max(a.c) AS max_c
, sum(b.cnt) AS sum_cnt
, nvl(cast(max(CASE WHEN a.cnt > 0
AND b.max_v > 0 THEN 'A'
WHEN a.cnt > 0
AND b.max_v < 0 THEN 'B'
WHEN a.cnt < 0 THEN 'C'
WHEN a.cnt < 0 THEN 'D'
ELSE (SELECT max(CASE WHEN a.k_1 is null
THEN 'A'
WHEN a.k_1 > 'V'
THEN 'A'
ELSE 'B'
END) AS v
FROM cte_1 a
WHERE 1=1
AND a.max_col > 120
AND a.type IN (SELECT type
FROM tab_t)
AND a.type IN('A', 'B', 'C', 'D')
)
END) AS STRING), '') AS some_case
FROM (SELECT a.k
-- Reserved words in comments are treated as comments.
, b.v
, c.max_v
, c.avg_v
, c.min_v
, c.cnt
FROM tab_a a
JOIN tab_b b
ON(a.k = b.k
AND b.d = 'dd')
-- + shuffle
LEFT JOIN cte_2 c
ON a.c = c.a
WHERE 1=1
AND a.col_1 >= 'filter'
-- comment
AND b.col_2 BETWEEN 'fil'
AND 'ter'
AND a.col_2 > 0) a
LEFT JOIN (SELECT a.k
, max(v) AS max_v
FROM tab_c a
WHERE 1=1
AND a.d >= 'filter'
GROUP BY a.k
HAVING count(1) > 0
) b
ON a.k = b.k
GROUP BY 1, 2
ORDER BY 1, 2
;
Update and enhance as necessary to:
- include VSCode user settings for adjusting output preferences
- improve Snowflake SQL compatibility
- work seamlessly with DBT (Data Build Tool)
Run npm run test
to test code changes on our sample.sql file
Thanks for the inspiration and examples: