exasol-sql-dialect

name	exasol-sql-dialect
description	Exasol SQL dialect specifics: syntax, data types, functions, and common pitfalls for generating correct Exasol SQL.
tags	["exasol","sql","dialect"]

Exasol SQL Dialect

Identifiers

• Identifiers (schema, table, column names) are case-insensitive by default and stored uppercase.
• To preserve case or use reserved words as identifiers, wrap them in double quotes: "myColumn".
• Embed a literal double-quote inside a quoted identifier by doubling it: "MY""COLUMN".

String Literals

• Use single quotes: 'value'.
• Embed a literal single-quote by doubling it: 'it''s a value'.

Data Types

Type	Notes
VARCHAR(n)	Variable-length string, up to 2,000,000 characters; n counts characters (not bytes); empty string is NULL
CHAR(n)	Fixed-length, blank-padded; up to 2,000 characters
DECIMAL(p, s)	Exact numeric; also written NUMERIC(p, s); default DECIMAL(18, 0)
DOUBLE PRECISION	64-bit floating point; also FLOAT or DOUBLE; NaN is treated as NULL; Infinity is not supported
BOOLEAN	TRUE / FALSE
DATE	Calendar date only (no time component)
TIMESTAMP(p)	Date + time; precision p 0–9, default 3 (milliseconds)
TIMESTAMP WITH LOCAL TIME ZONE	Stored in UTC, displayed in session time zone
INTERVAL YEAR TO MONTH	Period expressed as years and months
INTERVAL DAY TO SECOND	Period expressed as days through seconds
HASHTYPE(n BYTE)	Fixed-length hash/binary; 1–1,024 bytes (default 16); accepts hex, UUID, base64
GEOMETRY	Spatial data type

Row Limiting

Exasol uses LIMIT/OFFSET. FETCH FIRST … ROWS ONLY is not supported.

SELECT * FROM my_table ORDER BY id LIMIT 100;
SELECT * FROM my_table ORDER BY id LIMIT 20 OFFSET 10;

OFFSET requires ORDER BY. LIMIT is not allowed in correlated subqueries.

String Operations

• Concatenation: use ||, not +
• SUBSTR(str, pos, len) — 1-based positions
• LENGTH(str) — character count
• TRIM, LTRIM, RTRIM — whitespace or specific characters
• UPPER, LOWER, INITCAP
• REPLACE(str, search, replacement)
• LPAD(str, len, pad), RPAD(str, len, pad)

Regular Expressions

-- Boolean match (REGEXP_LIKE is not supported; use REGEXP_INSTR or REGEXP_COUNT instead)
WHERE REGEXP_INSTR(column, '^[A-Z]+$') > 0

-- Count matches
SELECT REGEXP_COUNT(column, '[0-9]+')

-- Replace with regex
SELECT REGEXP_REPLACE(column, '[0-9]+', '#')

-- Extract a match
SELECT REGEXP_SUBSTR(column, '[0-9]+', 1, 1)

-- Find position of first match
SELECT REGEXP_INSTR(column, '[0-9]+')

Date and Time Functions

-- Parse strings to date/timestamp
TO_DATE('2024-01-15', 'YYYY-MM-DD')
TO_TIMESTAMP('2024-01-15 10:30:00', 'YYYY-MM-DD HH24:MI:SS')

-- Format dates as strings
TO_CHAR(my_date, 'YYYY-MM-DD')

-- Arithmetic
ADD_DAYS(my_date, 7)
ADD_MONTHS(my_date, 3)
MONTHS_BETWEEN(date1, date2)

-- Truncate to period
TRUNC(my_timestamp, 'MM')   -- first day of month
TRUNC(my_timestamp, 'DD')   -- midnight

-- Current time
CURRENT_DATE
CURRENT_TIMESTAMP
SYSTIMESTAMP    -- like CURRENT_TIMESTAMP but uses database time zone (not session time zone)

NULL Handling

• Any arithmetic or string operation involving NULL returns NULL.
• Use IS NULL / IS NOT NULL, never = NULL.
• NVL(expr, default) — return default if expr is NULL
• NVL2(expr, val_if_not_null, val_if_null)
• NULLIF(a, b) — returns NULL if a equals b
• COALESCE(a, b, c, ...) — first non-NULL value

Aggregation and GROUP BY

• Columns in SELECT that are not aggregated must appear in GROUP BY.
• GROUP BY accepts column positions: GROUP BY 1, 2.
• HAVING filters groups after aggregation.
• ROLLUP, CUBE, GROUPING SETS are supported.
• COUNT(DISTINCT col) is supported.

Window Functions

Fully supported. Standard syntax:

SELECT
    col,
    ROW_NUMBER() OVER (PARTITION BY grp ORDER BY col) AS rn,
    SUM(val) OVER (PARTITION BY grp ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS running_sum
FROM my_table;

Joins

Standard INNER JOIN, LEFT/RIGHT/FULL OUTER JOIN, CROSS JOIN. NATURAL JOIN is not supported.

Set Operations

UNION, UNION ALL, INTERSECT, and EXCEPT work as in standard SQL. MINUS is an Oracle-compatible alias for EXCEPT:

SELECT * FROM a EXCEPT SELECT * FROM b;  -- standard
SELECT * FROM a MINUS SELECT * FROM b;   -- identical, Oracle-compatible

MERGE

MERGE INTO target t
USING source s ON t.id = s.id
WHEN MATCHED THEN
    UPDATE SET t.val = s.val
WHEN NOT MATCHED THEN
    INSERT (id, val) VALUES (s.id, s.val);

Hierarchical Queries (CONNECT BY)

SELECT level, id, parent_id, name
FROM hierarchy
START WITH parent_id IS NULL
CONNECT BY PRIOR id = parent_id
ORDER SIBLINGS BY name;

Subqueries and CTEs

-- Common Table Expression
WITH ranked AS (
    SELECT *, ROW_NUMBER() OVER (ORDER BY val DESC) AS rn
    FROM my_table
)
SELECT * FROM ranked WHERE rn <= 10;

-- Scalar subquery
SELECT name, (SELECT COUNT(*) FROM orders WHERE orders.cust_id = c.id) AS order_count
FROM customers c;

CASE Expression

-- Searched form
CASE WHEN score >= 90 THEN 'A' WHEN score >= 80 THEN 'B' ELSE 'C' END

-- Simple form
CASE status WHEN 1 THEN 'active' WHEN 0 THEN 'inactive' END

Type Casting

-- Explicit cast
CAST(my_col AS VARCHAR(100))
CAST('42' AS INTEGER)

-- Implicit coercion: Exasol coerces VARCHAR to numeric in arithmetic,
-- but be explicit to avoid surprises

Common Pitfalls

• Case of stored names: querying EXA_ALL_COLUMNS returns uppercase names unless the table was created with quoted identifiers.
• VARCHAR size is in characters: VARCHAR(10) means 10 characters (not bytes), regardless of UTF-8 encoding.
• Empty string is NULL: '' is stored and returned as NULL in VARCHAR columns.
• Timestamp precision: default TIMESTAMP (i.e. TIMESTAMP(3)) has millisecond precision; use TIMESTAMP(6) for microseconds. Valid precision values are 0–9; values outside this range cause an error.