SQL syntax

Queries can be nested so that the results of one query can be used in another query via a relational operator or aggregation function. A nested query is also known as a subquery. While joins and other table operations provide computationally superior (i.e. faster) alternatives in many cases, the use of subqueries introduces a hierarchy in execution that can be useful or necessary. In the following example, the aggregation function AVG receives as input the result of a subquery:

SELECT isbn,
       title,
       price
 FROM  Book
 WHERE price < (SELECT AVG(price) FROM Book)
 ORDER BY title;

A subquery can use values from the outer query, in which case it is known as a correlated subquery.

Since 1999 the SQL standard allows WITH clauses for subqueries, i.e. named subqueries, usually called common table expressions (also called subquery factoring). CTEs can also be recursive by referring to themselves; the resulting mechanism allows tree or graph traversals (when represented as relations), and more generally fixpoint computations.

A derived table is the use of referencing an SQL subquery in a FROM clause. Essentially, the derived table is a subquery that can be selected from or joined to. The derived table functionality allows the user to reference the subquery as a table. The inline view is also referred to as an inline view or a subselect.

In the following example, the SQL statement involves a join from the initial "Book" table to the derived table "sales". This derived table captures associated book sales information using the ISBN to join to the "Book" table. As a result, the derived table provides the result set with additional columns (the number of items sold and the company that sold the books):

SELECT b.isbn, b.title, b.price, sales.items_sold, sales.company_nm
FROM Book b
  JOIN (SELECT SUM(Items_Sold) Items_Sold, Company_Nm, ISBN
        FROM Book_Sales
        GROUP BY Company_Nm, ISBN) sales
  ON sales.isbn = b.isbn

The concept of Null allows SQL to deal with missing information in the relational model. The word NULL is a reserved keyword in SQL, used to identify the Null special marker. Comparisons with Null, for instance equality (=) in WHERE clauses, results in an Unknown truth value. In SELECT statements SQL returns only results for which the WHERE clause returns a value of True; i.e., it excludes results with values of False and also excludes those whose value is Unknown.

Along with True and False, the Unknown resulting from direct comparisons with Null thus brings a fragment of three-valued logic to SQL. The truth tables SQL uses for AND, OR, and NOT correspond to a common fragment of the Kleene and Lukasiewicz three-valued logic (which differ in their definition of implication, however SQL defines no such operation).[6]

There are however disputes about the semantic interpretation of Nulls in SQL because of its treatment outside direct comparisons. As seen in the table above, direct equality comparisons between two NULLs in SQL (e.g. {{{1}}}) return a truth value of Unknown. This is in line with the interpretation that Null does not have a value (and is not a member of any data domain) but is rather a placeholder or "mark" for missing information. However, the principle that two Nulls aren't equal to each other is effectively violated in the SQL specification for the UNION and INTERSECT operators, which do identify nulls with each other.[7] Consequently, these set operations in SQL may produce results not representing sure information, unlike operations involving explicit comparisons with NULL (e.g. those in a WHERE clause discussed above). In Codd's 1979 proposal (which was basically adopted by SQL92) this semantic inconsistency is rationalized by arguing that removal of duplicates in set operations happens "at a lower level of detail than equality testing in the evaluation of retrieval operations".[6] However, computer-science professor Ron van der Meyden concluded that "The inconsistencies in the SQL standard mean that it is not possible to ascribe any intuitive logical semantics to the treatment of nulls in SQL."[7]

Additionally, because SQL operators return Unknown when comparing anything with Null directly, SQL provides two Null-specific comparison predicates: IS NULL and IS NOT NULL test whether data is or is not Null.[8] SQL does not explicitly support universal quantification, and must work it out as a negated existential quantification.[9][10][11] There is also the "<row value expression> IS DISTINCT FROM <row value expression>" infixed comparison operator, which returns TRUE unless both operands are equal or both are NULL. Likewise, IS NOT DISTINCT FROM is defined as "NOT (<row value expression> IS DISTINCT FROM <row value expression>)". SQL:1999 also introduced BOOLEAN type variables, which according to the standard can also hold Unknown values if it is nullable. In practice, a number of systems (e.g. PostgreSQL) implement the BOOLEAN Unknown as a BOOLEAN NULL, which the standard says that the NULL BOOLEAN and UNKNOWN "may be used interchangeably to mean exactly the same thing".C. Date (2011). SQL and Relational Theory: How to Write Accurate SQL Code. O'Reilly Media, Inc. p. 83. ISBN 978-1-4493-1640-2.</ref>[12]

The Data Manipulation Language (DML) is the subset of SQL used to add, update and delete data:

• INSERT adds rows (formally tuples) to an existing table, e.g.:

INSERT INTO example
 (column1, column2, column3)
 VALUES
 ('test', 'N', NULL);

• UPDATE modifies a set of existing table rows, e.g.:

UPDATE example
 SET column1 = 'updated value'
 WHERE column2 = 'N';

• DELETE removes existing rows from a table, e.g.:

DELETE FROM example
 WHERE column2 = 'N';

• MERGE is used to combine the data of multiple tables. It combines the INSERT and UPDATE elements. It is defined in the SQL:2003 standard; prior to that, some databases provided similar functionality via different syntax, sometimes called "upsert".

 MERGE INTO table_name USING table_reference ON (condition)
 WHEN MATCHED THEN
 UPDATE SET column1 = value1 [, column2 = value2 ...]
 WHEN NOT MATCHED THEN
 INSERT (column1 [, column2 ...]) VALUES (value1 [, value2 ...])

Transactions, if available, wrap DML operations:

• START TRANSACTION (or BEGIN WORK, or BEGIN TRANSACTION, depending on SQL dialect) marks the start of a database transaction, which either completes entirely or not at all.
• SAVE TRANSACTION (or SAVEPOINT) saves the state of the database at the current point in transaction

CREATE TABLE tbl_1(id int);
 INSERT INTO tbl_1(id) VALUES(1);
 INSERT INTO tbl_1(id) VALUES(2);
COMMIT;
 UPDATE tbl_1 SET id=200 WHERE id=1;
SAVEPOINT id_1upd;
 UPDATE tbl_1 SET id=1000 WHERE id=2;
ROLLBACK to id_1upd;
 SELECT id from tbl_1;

• COMMIT makes all data changes in a transaction permanent.
• ROLLBACK discards all data changes since the last COMMIT or ROLLBACK, leaving the data as it was prior to those changes. Once the COMMIT statement completes, the transaction's changes cannot be rolled back.

COMMIT and ROLLBACK terminate the current transaction and release data locks. In the absence of a START TRANSACTION or similar statement, the semantics of SQL are implementation-dependent. The following example shows a classic transfer of funds transaction, where money is removed from one account and added to another. If either the removal or the addition fails, the entire transaction is rolled back.

START TRANSACTION;
 UPDATE Account SET amount=amount-200 WHERE account_number=1234;
 UPDATE Account SET amount=amount+200 WHERE account_number=2345;

IF ERRORS=0 COMMIT;
IF ERRORS<>0 ROLLBACK;

The Data Definition Language (DDL) manages table and index structure. The most basic items of DDL are the CREATE, ALTER, RENAME, DROP and TRUNCATE statements:

• CREATE creates an object (a table, for example) in the database, e.g.:

CREATE TABLE example(
 column1 INTEGER,
 column2 VARCHAR(50),
 column3 DATE NOT NULL,
 PRIMARY KEY (column1, column2)
);

• ALTER modifies the structure of an existing object in various ways, for example, adding a column to an existing table or a constraint, e.g.:

ALTER TABLE example ADD column4 INTEGER DEFAULT 25 NOT NULL;

• TRUNCATE deletes all data from a table in a very fast way, deleting the data inside the table and not the table itself. It usually implies a subsequent COMMIT operation, i.e., it cannot be rolled back (data is not written to the logs for rollback later, unlike DELETE).

TRUNCATE TABLE example;

• DROP deletes an object in the database, usually irretrievably, i.e., it cannot be rolled back, e.g.:

DROP TABLE example;

Each column in an SQL table declares the type(s) that column may contain. ANSI SQL includes the following data types.[13]

Character strings and national character strings

• CHARACTER(n) (or CHAR(n)): fixed-width n-character string, padded with spaces as needed
• CHARACTER VARYING(n) (or VARCHAR(n)): variable-width string with a maximum size of n characters
• CHARACTER LARGE OBJECT(n [ K | M | G | T ]) (or CLOB(n [ K | M | G | T ])): character large object with a maximum size of n [ K | M | G | T ] characters
• NATIONAL CHARACTER(n) (or NCHAR(n)): fixed width string supporting an international character set
• NATIONAL CHARACTER VARYING(n) (or NVARCHAR(n)): variable-width NCHAR string
• NATIONAL CHARACTER LARGE OBJECT(n [ K | M | G | T ]) (or NCLOB(n [ K | M | G | T ])): national character large object with a maximum size of n [ K | M | G | T ] characters

For the CHARACTER LARGE OBJECT and NATIONAL CHARACTER LARGE OBJECT data types, the multipliers K (1 024), M (1 048 576), G (1 073 741 824) and T (1 099 511 627 776) can be optionally used when specifying the length.

Binary

• BINARY(n): Fixed length binary string, maximum length n.
• BINARY VARYING(n) (or VARBINARY(n)): Variable length binary string, maximum length n.
• BINARY LARGE OBJECT(n [ K | M | G | T ]) (or BLOB(n [ K | M | G | T ])): binary large object with a maximum length n [ K | M | G | T ].

For the BINARY LARGE OBJECT data type, the multipliers K (1 024), M (1 048 576), G (1 073 741 824) and T (1 099 511 627 776) can be optionally used when specifying the length.

Boolean

• BOOLEAN

The BOOLEAN data type can store the values TRUE and FALSE.

Numerical

• INTEGER (or INT), SMALLINT and BIGINT
• FLOAT, REAL and DOUBLE PRECISION
• NUMERIC(precision, scale) or DECIMAL(precision, scale)
• DECFLOAT(precision)

For example, the number 123.45 has a precision of 5 and a scale of 2. The precision is a positive integer that determines the number of significant digits in a particular radix (binary or decimal). The scale is a non-negative integer. A scale of 0 indicates that the number is an integer. For a decimal number with scale S, the exact numeric value is the integer value of the significant digits divided by 10^S.

SQL provides the functions CEILING and FLOOR to round numerical values. (Popular vendor specific functions are TRUNC (Informix, DB2, PostgreSQL, Oracle and MySQL) and ROUND (Informix, SQLite, Sybase, Oracle, PostgreSQL, Microsoft SQL Server and Mimer SQL.))

Temporal (datetime)

• DATE: for date values (e.g. 2011-05-03).
• TIME: for time values (e.g. 15:51:36).
• TIME WITH TIME ZONE: the same as TIME, but including details about the time zone in question.
• TIMESTAMP: This is a DATE and a TIME put together in one variable (e.g. 2011-05-03 15:51:36.123456).
• TIMESTAMP WITH TIME ZONE: the same as TIMESTAMP, but including details about the time zone in question.

The SQL function EXTRACT can be used for extracting a single field (seconds, for instance) of a datetime or interval value. The current system date / time of the database server can be called by using functions like CURRENT_DATE, CURRENT_TIMESTAMP, LOCALTIME, or LOCALTIMESTAMP. (Popular vendor specific functions are TO_DATE, TO_TIME, TO_TIMESTAMP, YEAR, MONTH, DAY, HOUR, MINUTE, SECOND, DAYOFYEAR, DAYOFMONTH and DAYOFWEEK.)

Interval (datetime)

• YEAR(precision): a number of years
• YEAR(precision) TO MONTH: a number of years and months
• MONTH(precision): a number of months
• DAY(precision): a number of days
• DAY(precision) TO HOUR: a number of days and hours
• DAY(precision) TO MINUTE: a number of days, hours and minutes
• DAY(precision) TO SECOND(scale): a number of days, hours, minutes and seconds
• HOUR(precision): a number of hours
• HOUR(precision) TO MINUTE: a number of hours and minutes
• HOUR(precision) TO SECOND(scale): a number of hours, minutes and seconds
• MINUTE(precision): a number of minutes
• MINUTE(precision) TO SECOND(scale): a number of minutes and seconds

The Data Control Language (DCL) authorizes users to access and manipulate data. Its two main statements are:

• GRANT authorizes one or more users to perform an operation or a set of operations on an object.
• REVOKE eliminates a grant, which may be the default grant.

Example:

GRANT SELECT, UPDATE
 ON example
 TO some_user, another_user;

REVOKE SELECT, UPDATE
 ON example
 FROM some_user, another_user;