Zend_Db_SelectIntroduction
The Zend_Db_Select object represents a SQL SELECT
query statement. The class has methods for adding individual parts to the query. You can
specify some parts of the query using PHP methods and data structures, and the class
forms the correct SQL syntax for you. After you build a query, you can execute the query
as if you had written it as a string.
The value offered by Zend_Db_Select includes:
Object-oriented methods for specifying SQL queries in a piece-by-piece manner;
Database-independent abstraction of some parts of the SQL query;
Automatic quoting of metadata identifiers in most cases, to support identifiers
containing SQL reserved words and special characters;
Quoting identifiers and values, to help reduce risk of SQL injection attacks.
Using Zend_Db_Select is not mandatory. For very simple SELECT
queries, it is usually simpler to specify the entire SQL query as a string and execute
it using Adapter methods like query() or fetchAll(). Using
Zend_Db_Select is helpful if you need to assemble a SELECT query
procedurally, or based on conditional logic in your application.
Creating a Select Object
You can create an instance of a Zend_Db_Select object using the
select() method of a Zend_Db_Adapter_Abstract
object.
Example of the database adapter's select() methodselect();
]]>
Another way to create a Zend_Db_Select object is with its
constructor, specifying the database adapter as an argument.
Example of creating a new Select objectBuilding Select queries
When building the query, you can add clauses of the query one by one. There is a
separate method to add each clause to the Zend_Db_Select object.
Example of the using methods to add clausesselect();
// Add a FROM clause
$select->from( ...specify table and columns... )
// Add a WHERE clause
$select->where( ...specify search criteria... )
// Add an ORDER BY clause
$select->order( ...specify sorting criteria... );
]]>
You also can use most methods of the Zend_Db_Select object with a
convenient fluent interface. A fluent interface means that each method returns a
reference to the object on which it was called, so you can immediately call another
method.
Example of the using the fluent interfaceselect()
->from( ...specify table and columns... )
->where( ...specify search criteria... )
->order( ...specify sorting criteria... );
]]>
The examples in this section show usage of the fluent interface, but you can use the
non-fluent interface in all cases. It is often necessary to use the non-fluent
interface, for example, if your application needs to perform some logic before adding a
clause to a query.
Adding a FROM clause
Specify the table for this query using the from() method. You can
specify the table name as a simple string. Zend_Db_Select
applies identifier quoting around the table name, so you can use special characters.
Example of the from() methodselect()
->from( 'products' );
]]>
You can also specify the correlation name (sometimes called the "table alias") for
a table. Instead of a simple string, use an associative array mapping the
correlation name to the table name. In other clauses of the SQL query, use this
correlation name. If your query joins more than one table,
Zend_Db_Select generates unique correlation names based on
the table names, for any tables for which you don't specify the correlation name.
Example of specifying a table correlation nameselect()
->from( array('p' => 'products') );
]]>
Some RDBMS brands support a leading schema specifier for a table. You can specify
the table name as "schemaName.tableName", where
Zend_Db_Select quotes each part individually, or you may
specify the schema name separately. A schema name specified in the table name takes
precedence over a schema provided separately in the event that both are provided.
Example of specifying a schema nameselect()
->from( 'myschema.products' );
// or
$select = $db->select()
->from('products', '*', 'myschema');
]]>Adding Columns
In the second argument of the from() method, you can specify the
columns to select from the respective table. If you specify no columns, the default
is "*", the SQL wildcard for "all columns".
You can list the columns in a simple array of strings, or as an associative mapping
of column alias to column name. If you only have one column to query, and you don't
need to specify a column alias, you can list it as a plain string instead of an
array.
If you give an empty array as the columns argument, no columns from the respective
table are included in the result set. See a
code example
under the section on the join() method.
You can specify the column name as "correlationName.columnName".
Zend_Db_Select quotes each part individually. If you don't
specify a correlation name for a column, it uses the correlation name for the table
named in the current from() method.
Examples of specifying columnsselect()
->from(array('p' => 'products'),
array('product_id', 'product_name'));
// Build the same query, specifying correlation names:
// SELECT p."product_id", p."product_name"
// FROM "products" AS p
$select = $db->select()
->from(array('p' => 'products'),
array('p.product_id', 'p.product_name'));
// Build this query with an alias for one column:
// SELECT p."product_id" AS prodno, p."product_name"
// FROM "products" AS p
$select = $db->select()
->from(array('p' => 'products'),
array('prodno' => 'product_id', 'product_name'));
]]>Adding Expression Columns
Columns in SQL queries are sometimes expressions, not simply column names from a
table. Expressions should not have correlation names or quoting applied. If your
column string contains parentheses, Zend_Db_Select recognizes
it as an expression.
You also can create an object of type Zend_Db_Expr
explicitly, to prevent a string from being treated as a column name.
Zend_Db_Expr is a minimal class that contains a single
string. Zend_Db_Select recognizes objects of type
Zend_Db_Expr and converts them back to string, but does not
apply any alterations, such as quoting or correlation names.
Using Zend_Db_Expr for column names is not necessary if
your column expression contains parentheses;
Zend_Db_Select recognizes parentheses and treats the
string as an expression, skipping quoting and correlation names.
Examples of specifying columns containing expressionsselect()
->from(array('p' => 'products'),
array('product_id', 'LOWER(product_name)'));
// Build this query:
// SELECT p."product_id", (p.cost * 1.08) AS cost_plus_tax
// FROM "products" AS p
$select = $db->select()
->from(array('p' => 'products'),
array('product_id',
'cost_plus_tax' => '(p.cost * 1.08)')
);
// Build this query using Zend_Db_Expr explicitly:
// SELECT p."product_id", p.cost * 1.08 AS cost_plus_tax
// FROM "products" AS p
$select = $db->select()
->from(array('p' => 'products'),
array('product_id',
'cost_plus_tax' =>
new Zend_Db_Expr('p.cost * 1.08'))
);
]]>
In the cases above, Zend_Db_Select does not alter the string
to apply correlation names or identifier quoting. If those changes are necessary to
resolve ambiguity, you must make the changes manually in the string.
If your column names are SQL keywords or contain special characters, you should use
the Adapter's quoteIdentifier() method and interpolate the result into
the string. The quoteIdentifier() method uses SQL quoting to delimit
the identifier, which makes it clear that it is an identifier for a table or a
column, and not any other part of SQL syntax.
Your code is more database-independent if you use the
quoteIdentifier() method instead of typing quotes literally in your
string, because some RDBMS brands use nonstandard symbols for quoting identifiers.
The quoteIdentifier() method is designed to use the appropriate
quoting symbols based on the adapter type. The quoteIdentifier()
method also escapes any quote characters that appear within the identifier name
itself.
Examples of quoting columns in an expressionselect()
->from(array('p' => 'products'),
array('origin' =>
'(p.' . $db->quoteIdentifier('from') . ' + 10)')
);
]]>Adding columns to an existing FROM or JOIN table
There may be cases where you wish to add columns to an existing FROM or JOIN table
after those methods have been called. The columns() method allows you
to add specific columns at any point before the query is executed. You can supply
the columns as either a string or Zend_Db_Expr or as an array
of these elements. The second argument to this method can be omitted, implying that
the columns are to be added to the FROM table, otherwise an existing correlation
name must be used.
Examples of adding columns with the columns() methodselect()
->from(array('p' => 'products'), 'product_id')
->columns('product_name');
// Build the same query, specifying correlation names:
// SELECT p."product_id", p."product_name"
// FROM "products" AS p
$select = $db->select()
->from(array('p' => 'products'), 'p.product_id')
->columns('product_name', 'p');
// Alternatively use columns('p.product_name')
]]>Adding Another Table to the Query with JOIN
Many useful queries involve using a JOIN to combine rows from multiple
tables. You can add tables to a Zend_Db_Select query using
the join() method. Using this method is similar to the
from() method, except you can also specify a join condition in most
cases.
Example of the join() methodselect()
->from(array('p' => 'products'),
array('product_id', 'product_name'))
->join(array('l' => 'line_items'),
'p.product_id = l.product_id');
]]>
The second argument to join() is a string that is the join condition.
This is an expression that declares the criteria by which rows in one table match
rows in the the other table. You can use correlation names in this expression.
No quoting is applied to the expression you specify for the join condition; if
you have column names that need to be quoted, you must use
quoteIdentifier() as you form the string for the join condition.
The third argument to join() is an array of column names, like that
used in the from() method. It defaults to "*", supports
correlation names, expressions, and Zend_Db_Expr in the same
way as the array of column names in the from() method.
To select no columns from a table, use an empty array for the list of columns. This
usage works in the from() method too, but typically you want some
columns from the primary table in your queries, whereas you might want no columns
from a joined table.
Example of specifying no columnsselect()
->from(array('p' => 'products'),
array('product_id', 'product_name'))
->join(array('l' => 'line_items'),
'p.product_id = l.product_id',
array() ); // empty list of columns
]]>
Note the empty array() in the above example in place of a list of
columns from the joined table.
SQL has several types of joins. See the list below for the methods to support
different join types in Zend_Db_Select.
INNER JOIN with the
join(table, join, [columns]) or
joinInner(table, join, [columns]) methods.
This may be the most common type of join. Rows from each table are compared
using the join condition you specify. The result set includes only the rows
that satisfy the join condition. The result set can be empty if no rows
satisfy this condition.
All RDBMS brands support this join type.
LEFT JOIN with the
joinLeft(table, condition, [columns]) method.
All rows from the left operand table are included, matching rows from the
right operand table included, and the columns from the right operand table
are filled with NULLs if no row exists matching the left table.
All RDBMS brands support this join type.
RIGHT JOIN with the
joinRight(table, condition, [columns]) method.
Right outer join is the complement of left outer join. All rows from the
right operand table are included, matching rows from the left operand table
included, and the columns from the left operand table are filled with NULLs
if no row exists matching the right table.
Some RDBMS brands don't support this join type, but in general any right
join can be represented as a left join by reversing the order of the
tables.
FULL JOIN with the
joinFull(table, condition, [columns]) method.
A full outer join is like combining a left outer join and a right outer
join. All rows from both tables are included, paired with each other on the
same row of the result set if they satisfy the join condition, and
otherwise paired with NULLs in place of columns from the other table.
Some RDBMS brands don't support this join type.
CROSS JOIN with the
joinCross(table, [columns]) method.
A cross join is a Cartesian product. Every row in the first table is
matched to every row in the second table. Therefore the number of rows in
the result set is equal to the product of the number of rows in each table.
You can filter the result set using conditions in a WHERE clause; in this
way a cross join is similar to the old SQL-89 join syntax.
The joinCross() method has no parameter to specify the join
condition. Some RDBMS brands don't support this join type.
NATURAL JOIN with the
joinNatural(table, [columns]) method.
A natural join compares any column(s) that appear with the same name in
both tables. The comparison is equality of all the column(s); comparing the
columns using inequality is not a natural join. Only natural inner joins
are supported by this API, even though SQL permits natural outer joins as
well.
The joinNatural() method has no parameter to specify the join
condition.
In addition to these join methods, you can simplify your queries by using the
JoinUsing methods. Instead of supplying a full condition to your join, you simply
pass the column name on which to join and the Zend_Db_Select
object completes the condition for you.
Example of the joinUsing() methodselect()
->from('table1')
->joinUsing('table2', 'column1')
->where('column2 = ?', 'foo');
]]>
Each of the applicable join methods in the Zend_Db_Select
component has a corresponding 'using' method.
joinUsing(table, join, [columns]) and
joinInnerUsing(table, join, [columns])joinLeftUsing(table, join, [columns])joinRightUsing(table, join, [columns])joinFullUsing(table, join, [columns])Adding a WHERE Clause
You can specify criteria for restricting rows of the result set using the
where() method. The first argument of this method is a SQL expression,
and this expression is used in a SQL WHERE clause in the query.
Example of the where() method 100.00
$select = $db->select()
->from('products',
array('product_id', 'product_name', 'price'))
->where('price > 100.00');
]]>
No quoting is applied to expressions given to the where() or
orWhere() methods. If you have column names that need to be
quoted, you must use quoteIdentifier() as you form the string for
the condition.
The second argument to the where() method is optional. It is a value
to substitute into the expression. Zend_Db_Select quotes the
value and substitutes it for a question-mark ("?") symbol in the
expression.
This method accepts only one parameter. If you have an expression into which you
need to substitute multiple variables, you must format the string manually,
interpolating variables and performing quoting yourself.
Example of a parameter in the where() method 100.00)
$minimumPrice = 100;
$select = $db->select()
->from('products',
array('product_id', 'product_name', 'price'))
->where('price > ?', $minimumPrice);
]]>
You can invoke the where() method multiple times on the same
Zend_Db_Select object. The resulting query combines the
multiple terms together using AND between them.
Example of multiple where() methods 100.00)
// AND (price < 500.00)
$minimumPrice = 100;
$maximumPrice = 500;
$select = $db->select()
->from('products',
array('product_id', 'product_name', 'price'))
->where('price > ?', $minimumPrice)
->where('price < ?', $maximumPrice);
]]>
If you need to combine terms together using OR, use the
orWhere() method. This method is used in the same way as the
where() method, except that the term specified is preceded by
OR, instead of AND.
Example of the orWhere() method 500.00)
$minimumPrice = 100;
$maximumPrice = 500;
$select = $db->select()
->from('products',
array('product_id', 'product_name', 'price'))
->where('price < ?', $minimumPrice)
->orWhere('price > ?', $maximumPrice);
]]>Zend_Db_Select automatically puts parentheses around each
expression you specify using the where() or orWhere()
methods. This helps to ensure that Boolean operator precedence does not cause
unexpected results.
Example of parenthesizing Boolean expressions 500.00)
// AND (product_name = 'Apple')
$minimumPrice = 100;
$maximumPrice = 500;
$prod = 'Apple';
$select = $db->select()
->from('products',
array('product_id', 'product_name', 'price'))
->where("price < $minimumPrice OR price > $maximumPrice")
->where('product_name = ?', $prod);
]]>
In the example above, the results would be quite different without the parentheses,
because AND has higher precedence than OR.
Zend_Db_Select applies the parentheses so the effect is that
each expression in successive calls to the where() bind more tightly
than the AND that combines the expressions.
Adding a GROUP BY Clause
In SQL, the GROUP BY clause allows you to reduce the rows of a query
result set to one row per unique value found in the column(s) named in the
GROUP BY clause.
In Zend_Db_Select, you can specify the column(s) to use for
calculating the groups of rows using the group() method. The argument
to this method is a column or an array of columns to use in the
GROUP BY clause.
Example of the group() methodselect()
->from(array('p' => 'products'),
array('product_id'))
->join(array('l' => 'line_items'),
'p.product_id = l.product_id',
array('line_items_per_product' => 'COUNT(*)'))
->group('p.product_id');
]]>
Like the columns array in the from() method, you can use correlation
names in the column name strings, and the column is quoted as an identifier unless
the string contains parentheses or is an object of type
Zend_Db_Expr.
Adding a HAVING Clause
In SQL, the HAVING clause applies a restriction condition on groups of
rows. This is similar to how a WHERE clause applies a restriction
condition on rows. But the two clauses are different because WHERE
conditions are applied before groups are defined, whereas HAVING
conditions are applied after groups are defined.
In Zend_Db_Select, you can specify conditions for restricting
groups using the having() method. Its usage is similar to that of the
where() method. The first argument is a string containing a SQL
expression. The optional second argument is a value that is used to replace a
positional parameter placeholder in the SQL expression. Expressions given in
multiple invocations of the having() method are combined using the
Boolean AND operator, or the OR operator if you use the
orHaving() method.
Example of the having() method 10
$select = $db->select()
->from(array('p' => 'products'),
array('product_id'))
->join(array('l' => 'line_items'),
'p.product_id = l.product_id',
array('line_items_per_product' => 'COUNT(*)'))
->group('p.product_id')
->having('line_items_per_product > 10');
]]>
No quoting is applied to expressions given to the having() or
orHaving() methods. If you have column names that need to be
quoted, you must use quoteIdentifier() as you form the string for
the condition.
Adding an ORDER BY Clause
In SQL, the ORDER BY clause specifies one or more columns or
expressions by which the result set of a query is sorted. If multiple columns are
listed, the secondary columns are used to resolve ties; the sort order is
determined by the secondary columns if the preceding columns contain identical
values. The default sorting is from least value to greatest value. You can also
sort by greatest value to least value for a given column in the list by specifying
the keyword DESC after that column.
In Zend_Db_Select, you can use the order()
method to specify a column or an array of columns by which to sort. Each element of
the array is a string naming a column. Optionally with the ASCDESC keyword following it, separated by a space.
Like in the from() and group() methods, column names are
quoted as identifiers, unless they contain contain parentheses or are an object of
type Zend_Db_Expr.
Example of the order() methodselect()
->from(array('p' => 'products'),
array('product_id'))
->join(array('l' => 'line_items'),
'p.product_id = l.product_id',
array('line_items_per_product' => 'COUNT(*)'))
->group('p.product_id')
->order(array('line_items_per_product DESC',
'product_id'));
]]>Adding a LIMIT Clause
Some RDBMS brands extend SQL with a query clause known as the LIMIT
clause. This clause reduces the number of rows in the result set to at most a
number you specify. You can also specify to skip a number of rows before starting
to output. This feature makes it easy to take a subset of a result set, for example
when displaying query results on progressive pages of output.
In Zend_Db_Select, you can use the limit()
method to specify the count of rows and the number of rows to skip. The first
argument to this method is the desired count of rows. The second argument is the
number of rows to skip.
Example of the limit() methodselect()
->from(array('p' => 'products'),
array('product_id', 'product_name'))
->limit(10, 20);
]]>
The LIMIT syntax is not supported by all RDBMS brands. Some RDBMS
require different syntax to support similar functionality. Each
Zend_Db_Adapter_Abstract class includes a method to
produce SQL appropriate for that RDBMS.
Use the limitPage() method for an alternative way to specify row count
and offset. This method allows you to limit the result set to one of a series of
fixed-length subsets of rows from the query's total result set. In other words, you
specify the length of a "page" of results, and the ordinal number of the single
page of results you want the query to return. The page number is the first argument
of the limitPage() method, and the page length is the second argument.
Both arguments are required; they have no default values.
Example of the limitPage() methodselect()
->from(array('p' => 'products'),
array('product_id', 'product_name'))
->limitPage(2, 10);
]]>Adding the DISTINCT Query Modifier
The distinct() method enables you to add the DISTINCT
keyword to your SQL query.
Example of the distinct() methodselect()
->distinct()
->from(array('p' => 'products'), 'product_name');
]]>Adding the FOR UPDATE Query Modifier
The forUpdate() method enables you to add the FOR UPDATE
modifier to your SQL query.
Example of forUpdate() methodselect()
->forUpdate()
->from(array('p' => 'products'));
]]>Executing Select Queries
This section describes how to execute the query represented by a
Zend_Db_Select object.
Executing Select Queries from the Db Adapter
You can execute the query represented by the Zend_Db_Select
object by passing it as the first argument to the query() method of a
Zend_Db_Adapter_Abstract object. Use the
Zend_Db_Select objects instead of a string query.
The query() method returns an object of type
Zend_Db_Statement or PDOStatement, depending on the adapter
type.
Example using the Db adapter's query() methodselect()
->from('products');
$stmt = $db->query($select);
$result = $stmt->fetchAll();
]]>Executing Select Queries from the Object
As an alternative to using the query() method of the adapter object,
you can use the query() method of the
Zend_Db_Select object. Both methods return an object of type
Zend_Db_Statement or PDOStatement, depending on the adapter
type.
Example using the Select object's query methodselect()
->from('products');
$stmt = $select->query();
$result = $stmt->fetchAll();
]]>Converting a Select Object to a SQL String
If you need access to a string representation of the SQL query corresponding to the
Zend_Db_Select object, use the __toString()
method.
Example of the __toString() methodselect()
->from('products');
$sql = $select->__toString();
echo "$sql\n";
// The output is the string:
// SELECT * FROM "products"
]]>Other methods
This section describes other methods of the Zend_Db_Select class
that are not covered above: getPart() and reset().
Retrieving Parts of the Select Object
The getPart() method returns a representation of one part of your SQL
query. For example, you can use this method to return the array of expressions for
the WHERE clause, or the array of columns (or column expressions) that
are in the SELECT list, or the values of the count and offset for the
LIMIT clause.
The return value is not a string containing a fragment of SQL syntax. The return
value is an internal representation, which is typically an array structure
containing values and expressions. Each part of the query has a different
structure.
The single argument to the getPart() method is a string that
identifies which part of the Select query to return. For example, the string
'from' identifies the part of the Select object that stores
information about the tables in the FROM clause, including joined
tables.
The Zend_Db_Select class defines constants you can use for
parts of the SQL query. You can use these constant definitions, or you can the
literal strings.
Constants used by getPart() and reset()ConstantString valueZend_Db_Select::DISTINCT'distinct'Zend_Db_Select::FOR_UPDATE'forupdate'Zend_Db_Select::COLUMNS'columns'Zend_Db_Select::FROM'from'Zend_Db_Select::WHERE'where'Zend_Db_Select::GROUP'group'Zend_Db_Select::HAVING'having'Zend_Db_Select::ORDER'order'Zend_Db_Select::LIMIT_COUNT'limitcount'Zend_Db_Select::LIMIT_OFFSET'limitoffset'
Example of the getPart() methodselect()
->from('products')
->order('product_id');
// You can use a string literal to specify the part
$orderData = $select->getPart( 'order' );
// You can use a constant to specify the same part
$orderData = $select->getPart( Zend_Db_Select::ORDER );
// The return value may be an array structure, not a string.
// Each part has a different structure.
print_r( $orderData );
]]>Resetting Parts of the Select Object
The reset() method enables you to clear one specified part of the SQL
query, or else clear all parts of the SQL query if you omit the argument.
The single argument is optional. You can specify the part of the query to clear,
using the same strings you used in the argument to the getPart()
method. The part of the query you specify is reset to a default state.
If you omit the parameter, reset() changes all parts of the query to
their default state. This makes the Zend_Db_Select object
equivalent to a new object, as though you had just instantiated it.
Example of the reset() methodselect()
->from(array('p' => 'products')
->order('product_name');
// Changed requirement, instead order by a different columns:
// SELECT p.*
// FROM "products" AS p
// ORDER BY "product_id"
// Clear one part so we can redefine it
$select->reset( Zend_Db_Select::ORDER );
// And specify a different column
$select->order('product_id');
// Clear all parts of the query
$select->reset();
]]>