Unsupported versions: 7.1
This documentation is for an unsupported version of PostgreSQL.
You may want to view the same page for the current version, or one of the other supported versions listed above instead.

1.4. The SQL Language

As is the case with most modern relational languages, SQL is based on the tuple relational calculus. As a result every query that can be formulated using the tuple relational calculus (or equivalently, relational algebra) can also be formulated using SQL. There are, however, capabilities beyond the scope of relational algebra or calculus. Here is a list of some additional features provided by SQL that are not part of relational algebra or calculus:

  • Commands for insertion, deletion or modification of data.

  • Arithmetic capability: In SQL it is possible to involve arithmetic operations as well as comparisons, e.g.

    A < B + 3.
           
    
    Note that + or other arithmetic operators appear neither in relational algebra nor in relational calculus.
  • Assignment and Print Commands: It is possible to print a relation constructed by a query and to assign a computed relation to a relation name.

  • Aggregate Functions: Operations such as average, sum, max, etc. can be applied to columns of a relation to obtain a single quantity.

1.4.1. Select

The most often used command in SQL is the SELECT statement, used to retrieve data. The syntax is:

SELECT [ ALL | DISTINCT [ ON ( expression [, ...] ) ] ]
    * | expression [ AS output_name ] [, ...]
    [ INTO [ TEMPORARY | TEMP ] [ TABLE ] new_table ]
    [ FROM from_item [, ...] ]
    [ WHERE condition ]
    [ GROUP BY expression [, ...] ]
    [ HAVING condition [, ...] ]
    [ { UNION | INTERSECT | EXCEPT [ ALL ] } select ]
    [ ORDER BY expression [ ASC | DESC | USING operator ] [, ...] ]
    [ FOR UPDATE [ OF class_name [, ...] ] ]
    [ LIMIT { count | ALL } [ { OFFSET | , } start ]]
     

Now we will illustrate the complex syntax of the SELECT statement with various examples. The tables used for the examples are defined in The Suppliers and Parts Database.

1.4.1.1. Simple Selects

Here are some simple examples using a SELECT statement:

Example 1-4. Simple Query with Qualification

To retrieve all tuples from table PART where the attribute PRICE is greater than 10 we formulate the following query:

SELECT * FROM PART
    WHERE PRICE > 10;
        
and get the table:
 PNO |  PNAME  |  PRICE
-----+---------+--------
  3  |  Bolt   |   15
  4  |  Cam    |   25
        

Using "*" in the SELECT statement will deliver all attributes from the table. If we want to retrieve only the attributes PNAME and PRICE from table PART we use the statement:

SELECT PNAME, PRICE 
    FROM PART
    WHERE PRICE > 10;
        
In this case the result is:
                      PNAME  |  PRICE
                     --------+--------
                      Bolt   |   15
                      Cam    |   25
        
Note that the SQL SELECT corresponds to the "projection" in relational algebra not to the "selection" (see Relational Algebra for more details).

The qualifications in the WHERE clause can also be logically connected using the keywords OR, AND, and NOT:

SELECT PNAME, PRICE 
    FROM PART
    WHERE PNAME = 'Bolt' AND
         (PRICE = 0 OR PRICE <= 15);
        
will lead to the result:
 PNAME  |  PRICE
--------+--------
 Bolt   |   15
        

Arithmetic operations may be used in the target list and in the WHERE clause. For example if we want to know how much it would cost if we take two pieces of a part we could use the following query:

SELECT PNAME, PRICE * 2 AS DOUBLE
    FROM PART
    WHERE PRICE * 2 < 50;
        
and we get:
 PNAME  |  DOUBLE
--------+---------
 Screw  |    20
 Nut    |    16
 Bolt   |    30
        
Note that the word DOUBLE after the keyword AS is the new title of the second column. This technique can be used for every element of the target list to assign a new title to the resulting column. This new title is often referred to as alias. The alias cannot be used throughout the rest of the query.

1.4.1.2. Joins

The following example shows how joins are realized in SQL.

To join the three tables SUPPLIER, PART and SELLS over their common attributes we formulate the following statement:

SELECT S.SNAME, P.PNAME
    FROM SUPPLIER S, PART P, SELLS SE
    WHERE S.SNO = SE.SNO AND
          P.PNO = SE.PNO;
      
and get the following table as a result:
 SNAME | PNAME
-------+-------
 Smith | Screw
 Smith | Nut
 Jones | Cam
 Adams | Screw
 Adams | Bolt
 Blake | Nut
 Blake | Bolt
 Blake | Cam
      

In the FROM clause we introduced an alias name for every relation because there are common named attributes (SNO and PNO) among the relations. Now we can distinguish between the common named attributes by simply prefixing the attribute name with the alias name followed by a dot. The join is calculated in the same way as shown in An Inner Join. First the Cartesian product SUPPLIER × PART × SELLS is derived. Now only those tuples satisfying the conditions given in the WHERE clause are selected (i.e. the common named attributes have to be equal). Finally we project out all columns but S.SNAME and P.PNAME.

Another way to perform joins is to use the SQL JOIN syntax as follows:

select sname, pname from supplier
        JOIN sells USING (sno)
        JOIN part USING (pno);
        
giving again:
 sname | pname
-------+-------
 Smith | Screw
 Adams | Screw
 Smith | Nut
 Blake | Nut
 Adams | Bolt
 Blake | Bolt
 Jones | Cam
 Blake | Cam
(8 rows) 
         

A joined table, created using JOIN syntax, is a table reference list item that occurs in a FROM clause and before any WHERE, GROUP BY, or HAVING clause. Other table references, including table names or other JOIN clauses, may be included in the FROM clause if separated by commas. JOINed tables are logically like any other table listed in the FROM clause.

SQL JOINs come in two main types, CROSS JOINs (unqualified joins) and qualified JOINs. Qualified joins can be further subdivided based on the way in which the join condition is specified (ON, USING, or NATURAL) and the way in which it is applied (INNER or OUTER join).

Join Types

CROSS JOIN

{ T1 } CROSS JOIN { T2 }

A cross join takes two tables T1 and T2 having N and M rows respectively, and returns a joined table containing all N*M possible joined rows. For each row R1 of T1, each row R2 of T2 is joined with R1 to yield a joined table row JR consisting of all fields in R1 and R2. A CROSS JOIN is equivalent to an INNER JOIN ON TRUE.

Qualified JOINs

{ T1 } [ NATURAL ] [ INNER | { LEFT | RIGHT | FULL } [ OUTER ] ] JOIN { T2 } { ON search condition | USING ( join column list ) }

A qualified JOIN must specify its join condition by providing one (and only one) of NATURAL, ON, or USING. The ON clause takes a search condition, which is the same as in a WHERE clause. The USING clause takes a comma-separated list of column names, which the joined tables must have in common, and joins the tables on equality of those columns. NATURAL is shorthand for a USING clause that lists all the common column names of the two tables. A side-effect of both USING and NATURAL is that only one copy of each joined column is emitted into the result table (compare the relational-algebra definition of JOIN, shown earlier).


[ INNER ] JOIN

For each row R1 of T1, the joined table has a row for each row in T2 that satisfies the join condition with R1.

Tip: The words INNER and OUTER are optional for all JOINs. INNER is the default. LEFT, RIGHT, and FULL imply an OUTER JOIN.


LEFT [ OUTER ] JOIN

First, an INNER JOIN is performed. Then, for each row in T1 that does not satisfy the join condition with any row in T2, an additional joined row is returned with null fields in the columns from T2.

Tip: The joined table unconditionally has a row for each row in T1.


RIGHT [ OUTER ] JOIN

First, an INNER JOIN is performed. Then, for each row in T2 that does not satisfy the join condition with any row in T1, an additional joined row is returned with null fields in the columns from T1.

Tip: The joined table unconditionally has a row for each row in T2.


FULL [ OUTER ] JOIN

First, an INNER JOIN is performed. Then, for each row in T1 that does not satisfy the join condition with any row in T2, an additional joined row is returned with null fields in the columns from T2. Also, for each row in T2 that does not satisfy the join condition with any row in T1, an additional joined row is returned with null fields in the columns from T1.

Tip: The joined table unconditionally has a row for every row of T1 and a row for every row of T2.

JOINs of all types can be chained together or nested where either or both of T1 and T2 may be JOINed tables. Parenthesis can be used around JOIN clauses to control the order of JOINs which are otherwise processed left to right.

1.4.1.3. Aggregate Operators

SQL provides aggregate operators (e.g. AVG, COUNT, SUM, MIN, MAX) that take an expression as argument. The expression is evaluated at each row that satisfies the WHERE clause, and the aggregate operator is calculated over this set of input values. Normally, an aggregate delivers a single result for a whole SELECT statement. But if grouping is specified in the query, then a separate calculation is done over the rows of each group, and an aggregate result is delivered per group (see next section).

Example 1-5. Aggregates

If we want to know the average cost of all parts in table PART we use the following query:

SELECT AVG(PRICE) AS AVG_PRICE
    FROM PART;
        

The result is:

 AVG_PRICE
-----------
   14.5
        

If we want to know how many parts are defined in table PART we use the statement:

SELECT COUNT(PNO)
    FROM PART;
        
and get:
 COUNT
-------
   4
        

1.4.1.4. Aggregation by Groups

SQL allows one to partition the tuples of a table into groups. Then the aggregate operators described above can be applied to the groups --- i.e. the value of the aggregate operator is no longer calculated over all the values of the specified column but over all values of a group. Thus the aggregate operator is evaluated separately for every group.

The partitioning of the tuples into groups is done by using the keywords GROUP BY followed by a list of attributes that define the groups. If we have GROUP BY A1, &tdot;, Ak we partition the relation into groups, such that two tuples are in the same group if and only if they agree on all the attributes A1, &tdot;, Ak.

Example 1-6. Aggregates

If we want to know how many parts are sold by every supplier we formulate the query:

SELECT S.SNO, S.SNAME, COUNT(SE.PNO)
    FROM SUPPLIER S, SELLS SE
    WHERE S.SNO = SE.SNO
    GROUP BY S.SNO, S.SNAME;
        
and get:
 SNO | SNAME | COUNT
-----+-------+-------
  1  | Smith |   2
  2  | Jones |   1
  3  | Adams |   2
  4  | Blake |   3
        

Now let's have a look of what is happening here. First the join of the tables SUPPLIER and SELLS is derived:

 S.SNO | S.SNAME | SE.PNO
-------+---------+--------
   1   |  Smith  |   1
   1   |  Smith  |   2
   2   |  Jones  |   4
   3   |  Adams  |   1
   3   |  Adams  |   3
   4   |  Blake  |   2
   4   |  Blake  |   3
   4   |  Blake  |   4
        

Next we partition the tuples into groups by putting all tuples together that agree on both attributes S.SNO and S.SNAME:

 S.SNO | S.SNAME | SE.PNO
-------+---------+--------
   1   |  Smith  |   1
                 |   2
--------------------------
   2   |  Jones  |   4
--------------------------
   3   |  Adams  |   1
                 |   3
--------------------------
   4   |  Blake  |   2
                 |   3
                 |   4
        

In our example we got four groups and now we can apply the aggregate operator COUNT to every group leading to the final result of the query given above.

Note that for a query using GROUP BY and aggregate operators to make sense the target list can only refer directly to the attributes being grouped by. Other attributes may only be used inside the argument of an aggregate function. Otherwise there would not be a unique value to associate with the other attributes.

Also observe that it makes no sense to ask for an aggregate of an aggregate, e.g., AVG(MAX(sno)), because a SELECT only does one pass of grouping and aggregation. You can get a result of this kind by using a temporary table or a sub-SELECT in the FROM clause to do the first level of aggregation.

1.4.1.5. Having

The HAVING clause works much like the WHERE clause and is used to consider only those groups satisfying the qualification given in the HAVING clause. Essentially, WHERE filters out unwanted input rows before grouping and aggregation are done, whereas HAVING filters out unwanted group rows post-GROUP. Therefore, WHERE cannot refer to the results of aggregate functions. On the other hand, there's no point in writing a HAVING condition that doesn't involve an aggregate function! If your condition doesn't involve aggregates, you might as well write it in WHERE, and thereby avoid the computation of aggregates for groups that you're just going to throw away anyway.

Example 1-7. Having

If we want only those suppliers selling more than one part we use the query:

SELECT S.SNO, S.SNAME, COUNT(SE.PNO)
    FROM SUPPLIER S, SELLS SE
    WHERE S.SNO = SE.SNO
    GROUP BY S.SNO, S.SNAME
    HAVING COUNT(SE.PNO) > 1;
        
and get:
 SNO | SNAME | COUNT
-----+-------+-------
  1  | Smith |   2
  3  | Adams |   2
  4  | Blake |   3
        

1.4.1.6. Subqueries

In the WHERE and HAVING clauses the use of subqueries (subselects) is allowed in every place where a value is expected. In this case the value must be derived by evaluating the subquery first. The usage of subqueries extends the expressive power of SQL.

Example 1-8. Subselect

If we want to know all parts having a greater price than the part named 'Screw' we use the query:

SELECT * 
    FROM PART 
    WHERE PRICE > (SELECT PRICE FROM PART
                   WHERE PNAME='Screw');
        

The result is:

 PNO |  PNAME  |  PRICE
-----+---------+--------
  3  |  Bolt   |   15
  4  |  Cam    |   25
        

When we look at the above query we can see the keyword SELECT two times. The first one at the beginning of the query - we will refer to it as outer SELECT - and the one in the WHERE clause which begins a nested query - we will refer to it as inner SELECT. For every tuple of the outer SELECT the inner SELECT has to be evaluated. After every evaluation we know the price of the tuple named 'Screw' and we can check if the price of the actual tuple is greater. (Actually, in this example the inner query need only be evaluated once, since it does not depend on the state of the outer query.)

If we want to know all suppliers that do not sell any part (e.g. to be able to remove these suppliers from the database) we use:

SELECT * 
    FROM SUPPLIER S
    WHERE NOT EXISTS
        (SELECT * FROM SELLS SE
         WHERE SE.SNO = S.SNO);
        

In our example the result will be empty because every supplier sells at least one part. Note that we use S.SNO from the outer SELECT within the WHERE clause of the inner SELECT. Here the subquery must be evaluated afresh for each tuple from the outer query, i.e. the value for S.SNO is always taken from the current tuple of the outer SELECT.

1.4.1.7. Subqueries in FROM

A somewhat different way of using subqueries is to put them in the FROM clause. This is a useful feature because a subquery of this kind can output multiple columns and rows, whereas a subquery used in an expression must deliver just a single result. It also lets us get more than one round of grouping/aggregation without resorting to a temporary table.

Example 1-9. Subselect in FROM

If we want to know the highest average part price among all our suppliers, we can't write MAX(AVG(PRICE)), but we can write:

SELECT MAX(subtable.avgprice)
    FROM (SELECT AVG(P.PRICE) AS avgprice
          FROM SUPPLIER S, PART P, SELLS SE
          WHERE S.SNO = SE.SNO AND
                P.PNO = SE.PNO
          GROUP BY S.SNO) subtable;
        
The subquery returns one row per supplier (because of its GROUP BY) and then we aggregate over those rows in the outer query.

1.4.1.8. Union, Intersect, Except

These operations calculate the union, intersection and set theoretic difference of the tuples derived by two subqueries.

Example 1-10. Union, Intersect, Except

The following query is an example for UNION:

SELECT S.SNO, S.SNAME, S.CITY
    FROM SUPPLIER S
    WHERE S.SNAME = 'Jones'
UNION
    SELECT S.SNO, S.SNAME, S.CITY
    FROM SUPPLIER S
    WHERE S.SNAME = 'Adams';    
        
gives the result:
 SNO | SNAME |  CITY
-----+-------+--------
  2  | Jones | Paris
  3  | Adams | Vienna
        

Here is an example for INTERSECT:

SELECT S.SNO, S.SNAME, S.CITY
    FROM SUPPLIER S
    WHERE S.SNO > 1
INTERSECT
    SELECT S.SNO, S.SNAME, S.CITY
    FROM SUPPLIER S
    WHERE S.SNO < 3;
        
gives the result:
 SNO | SNAME |  CITY
-----+-------+--------
  2  | Jones | Paris
        
The only tuple returned by both parts of the query is the one having SNO=2.

Finally an example for EXCEPT:

SELECT S.SNO, S.SNAME, S.CITY
    FROM SUPPLIER S
    WHERE S.SNO > 1
EXCEPT
    SELECT S.SNO, S.SNAME, S.CITY
    FROM SUPPLIER S
    WHERE S.SNO > 3;
        
gives the result:
 SNO | SNAME |  CITY
-----+-------+--------
  2  | Jones | Paris
  3  | Adams | Vienna
        

1.4.2. Data Definition

There is a set of commands used for data definition included in the SQL language.

1.4.2.1. Create Table

The most fundamental command for data definition is the one that creates a new relation (a new table). The syntax of the CREATE TABLE command is:

CREATE TABLE table_name
    (name_of_attr_1 type_of_attr_1
     [, name_of_attr_2 type_of_attr_2 
     [, ...]]);
      

Example 1-11. Table Creation

To create the tables defined in The Suppliers and Parts Database the following SQL statements are used:

CREATE TABLE SUPPLIER
    (SNO   INTEGER,
     SNAME VARCHAR(20),
     CITY  VARCHAR(20));
     
CREATE TABLE PART
    (PNO   INTEGER,
     PNAME VARCHAR(20),
     PRICE DECIMAL(4 , 2));
     
CREATE TABLE SELLS
    (SNO INTEGER,
     PNO INTEGER);
        

1.4.2.2. Data Types in SQL

The following is a list of some data types that are supported by SQL:

  • INTEGER: signed fullword binary integer (31 bits precision).

  • SMALLINT: signed halfword binary integer (15 bits precision).

  • DECIMAL (p[,q]): signed packed decimal number of up to p digits, with q digits to the right of the decimal point. If q is omitted it is assumed to be 0.

  • FLOAT: signed doubleword floating point number.

  • CHAR(n): fixed length character string of length n.

  • VARCHAR(n): varying length character string of maximum length n.

1.4.2.3. Create Index

Indices are used to speed up access to a relation. If a relation R has an index on attribute A then we can retrieve all tuples t having t(A) = a in time roughly proportional to the number of such tuples t rather than in time proportional to the size of R.

To create an index in SQL the CREATE INDEX command is used. The syntax is:

CREATE INDEX index_name 
    ON table_name ( name_of_attribute );
      

Example 1-12. Create Index

To create an index named I on attribute SNAME of relation SUPPLIER we use the following statement:

CREATE INDEX I ON SUPPLIER (SNAME);
      

The created index is maintained automatically, i.e. whenever a new tuple is inserted into the relation SUPPLIER the index I is adapted. Note that the only changes a user can perceive when an index is present are increased speed for SELECT and decreases in speed of updates.

1.4.2.4. Create View

A view may be regarded as a virtual table, i.e. a table that does not physically exist in the database but looks to the user as if it does. By contrast, when we talk of a base table there is really a physically stored counterpart of each row of the table somewhere in the physical storage.

Views do not have their own, physically separate, distinguishable stored data. Instead, the system stores the definition of the view (i.e. the rules about how to access physically stored base tables in order to materialize the view) somewhere in the system catalogs (see System Catalogs). For a discussion on different techniques to implement views refer to SIM98.

In SQL the CREATE VIEW command is used to define a view. The syntax is:

CREATE VIEW view_name
    AS select_stmt
      
where select_stmt is a valid select statement as defined in Select. Note that select_stmt is not executed when the view is created. It is just stored in the system catalogs and is executed whenever a query against the view is made.

Let the following view definition be given (we use the tables from The Suppliers and Parts Database again):

CREATE VIEW London_Suppliers
    AS SELECT S.SNAME, P.PNAME
        FROM SUPPLIER S, PART P, SELLS SE
        WHERE S.SNO = SE.SNO AND
              P.PNO = SE.PNO AND
              S.CITY = 'London';
      

Now we can use this virtual relation London_Suppliers as if it were another base table:

SELECT * FROM London_Suppliers
    WHERE PNAME = 'Screw';
      
which will return the following table:
 SNAME | PNAME
-------+-------
 Smith | Screw                 
      

To calculate this result the database system has to do a hidden access to the base tables SUPPLIER, SELLS and PART first. It does so by executing the query given in the view definition against those base tables. After that the additional qualifications (given in the query against the view) can be applied to obtain the resulting table.

1.4.2.5. Drop Table, Drop Index, Drop View

To destroy a table (including all tuples stored in that table) the DROP TABLE command is used:

DROP TABLE table_name;
       

To destroy the SUPPLIER table use the following statement:

DROP TABLE SUPPLIER;
      

The DROP INDEX command is used to destroy an index:

DROP INDEX index_name;
      

Finally to destroy a given view use the command DROP VIEW:

DROP VIEW view_name;
      

1.4.3. Data Manipulation

1.4.3.1. Insert Into

Once a table is created (see Create Table), it can be filled with tuples using the command INSERT INTO. The syntax is:

INSERT INTO table_name (name_of_attr_1 
    [, name_of_attr_2 [,...]])
    VALUES (val_attr_1 [, val_attr_2 [, ...]]);
      

To insert the first tuple into the relation SUPPLIER (from The Suppliers and Parts Database) we use the following statement:

INSERT INTO SUPPLIER (SNO, SNAME, CITY)
    VALUES (1, 'Smith', 'London');
      

To insert the first tuple into the relation SELLS we use:

INSERT INTO SELLS (SNO, PNO)
    VALUES (1, 1);
      

1.4.3.2. Update

To change one or more attribute values of tuples in a relation the UPDATE command is used. The syntax is:

UPDATE table_name
    SET name_of_attr_1 = value_1 
        [, ... [, name_of_attr_k = value_k]]
    WHERE condition;
      

To change the value of attribute PRICE of the part 'Screw' in the relation PART we use:

UPDATE PART
    SET PRICE = 15
    WHERE PNAME = 'Screw';
      

The new value of attribute PRICE of the tuple whose name is 'Screw' is now 15.

1.4.3.3. Delete

To delete a tuple from a particular table use the command DELETE FROM. The syntax is:

DELETE FROM table_name
    WHERE condition;
      

To delete the supplier called 'Smith' of the table SUPPLIER the following statement is used:

DELETE FROM SUPPLIER
    WHERE SNAME = 'Smith';
      

1.4.4. System Catalogs

In every SQL database system system catalogs are used to keep track of which tables, views indexes etc. are defined in the database. These system catalogs can be queried as if they were normal relations. For example there is one catalog used for the definition of views. This catalog stores the query from the view definition. Whenever a query against a view is made, the system first gets the view definition query out of the catalog and materializes the view before proceeding with the user query (see Simkovics, 1998 for a more detailed description). For more information about system catalogs refer to Date, 1994.

1.4.5. Embedded SQL

In this section we will sketch how SQL can be embedded into a host language (e.g. C). There are two main reasons why we want to use SQL from a host language:

  • There are queries that cannot be formulated using pure SQL (i.e. recursive queries). To be able to perform such queries we need a host language with a greater expressive power than SQL.

  • We simply want to access a database from some application that is written in the host language (e.g. a ticket reservation system with a graphical user interface is written in C and the information about which tickets are still left is stored in a database that can be accessed using embedded SQL).

A program using embedded SQL in a host language consists of statements of the host language and of embedded SQL (ESQL) statements. Every ESQL statement begins with the keywords EXEC SQL. The ESQL statements are transformed to statements of the host language by a precompiler (which usually inserts calls to library routines that perform the various SQL commands).

When we look at the examples throughout Select we realize that the result of the queries is very often a set of tuples. Most host languages are not designed to operate on sets so we need a mechanism to access every single tuple of the set of tuples returned by a SELECT statement. This mechanism can be provided by declaring a cursor. After that we can use the FETCH command to retrieve a tuple and set the cursor to the next tuple.

For a detailed discussion on embedded SQL refer to Date and Darwen, 1997, Date, 1994, or Ullman, 1988.