FROM Clause

(back to top)

• table-name a base table or temporary table. Tables owned by a different user can be qualified by specifying the user ID. Tables owned by groups to which the current user belongs are found by default without specifying the user ID.

• view-name specifies a view to include in the query. As with tables, views owned by a different user can be qualified by specifying the user ID. Views owned by groups to which the current user belongs are found by default without specifying the user ID. Although the syntax permits table hints on views, these hints have no effect.

• procedure-name a stored procedure that returns a result set. This clause applies to the FROM clause of SELECT statements only. The parentheses following the procedure name are required even if the procedure does not take parameters. DEFAULT can be specified in place of an optional parameter.

• parameter specifies a scalar-parameter or table-parameter clause. A scalar-parameter are any objects of a valid SQL datatype. A table-parameter can be specified using a table, view or common table-expression name which are treated as new instance of this object if the object is also used outside the table-parameter.

  This query illustrates a valid FROM clause where the two references to the same table T are treated as two different instances of the same table T.

  SELECT * FROM T, my_proc(TABLE(SELECT T.Z, T.X FROM T)
  OVER(PARTITION BY T.Z));

  Table Parameterized Function (TPF) Example—This query illustrates a valid FROM clause.

  SELECT * FROM R, SELECT * FROM my_udf(1);
  SELECT * FROM my_tpf(1, TABLE(SELECT c1, c2 FROM t))
     (my_proc(R.X, TABLE T OVER PARTITION BY T.X)) AS XX;

  If a subquery is used to define the TABLE parameter, then the following restrictions must hold:

  • The table-parameter clause must be of type IN.
  • PARTITION BY or ORDER BY clauses must refer to the columns of the derived table and outer references. An expression in the expression-list can be an integer K which refers to the Kth column of the TABLE input parameter.

  Note A Table UDF can only be referenced in a FROM clause of a SQL statement.

• PARTITION BY logically specifies how the invocation of the function will be performed by the execution engine. The execution engine must invoke the function for each partition and the function must process a whole partition in each invocation.

  PARTITION BY clause also specifies how the input data must be partitioned such that each invocation of the function will process exactly one partition of data. The function must be invoked the number of times equal to the number of partitions. For TPF, the parallelism characteristics are established through dynamic negotiation between the server and the UDF at the runtime. If the TPF can be executed in parallel, for N input partitions, the function can be instantiated M times, with M <=N. Each instantiation of the function can be invoked more than once, each invocation consuming exactly one partition.

  You can specify only one TABLE input parameter for PARTITION BY <expression-list> or PARTITION BY ANY clause. For all other TABLE input parameters you must specify, explicit or implicit PARTITION BY NONE clause.

  Note The execution engine can invoke the function in any order of the partitions and the function is assumed to return the same result sets regardless of the partitions order. Partitions cannot be split among two invocations of the function.

• ORDER BY specifies that the input data in each partition is expected to be sorted by <expression-list> by the execution engine. The UDF expects each partition to have this physical property. If only one partition exists, the whole input data is ordered based on the ORDER BY specification. ORDER BY clause can be specified for any of the TABLE input parameters with PARTITION BY NONE or without PARTITION BY clause.

• derived-table you can supply a SELECT statement instead of table or view name in the FROM clause. A SELECT statement used in this way is called a derived table, and it must be given an alias. For example, the following statement contains a derived table, MyDerivedTable, which ranks products in the Products table by UnitPrice.

  SELECT TOP 3 *
    FROM ( SELECT Description, 
           Quantity, 
           UnitPrice,
           RANK() OVER ( ORDER BY UnitPrice ASC ) 
           AS Rank 
           FROM Products ) AS MyDerivedTable
           ORDER BY Rank;

• join-expression, join-operator, join-type the join-type keywords are:

  Keyword	Description
  CROSS JOIN	Returns the Cartesian product (cross product) of the two source tables
  NATURAL JOIN	Compares for equality all corresponding columns with the same names in two tables (a special case equijoin; columns are of same length and data type)
  KEY JOIN	Restricts foreign-key values in the first table to be equal to the primary-key values in the second table
  INNER JOIN	Discards all rows from the result table that do not have corresponding rows in both tables
  LEFT OUTER JOIN	Preserves unmatched rows from the left table, but discards unmatched rows from the right table
  RIGHT OUTER JOIN	Preserves unmatched rows from the right table, but discards unmatched rows from the left table
  FULL OUTER JOIN	Retains unmatched rows from both the left and the right tables

  Do not mix comma-style joins and keyword-style joins in the FROM clause. The same query can be written two ways, each using one of the join styles. The ANSI syntax keyword style join is preferable.

  This query uses a comma-style join:

  SELECT *
    FROM Products pr, SalesOrders so, SalesOrderItems si
    WHERE pr.ProductID = so.ProductID
      AND pr.ProductID = si.ProductID;

  The same query can use the preferable keyword-style join:

  SELECT *
    FROM Products pr INNER JOIN SalesOrders so
      ON (pr.ProductID = so.ProductID)
    INNER JOIN SalesOrderItems si
      ON (pr.ProductID = si.ProductID);

  The ON clause filters the data of inner, left, right, and full joins. Cross joins do not have an ON clause. In an inner join, the ON clause is equivalent to a WHERE clause. In outer joins, however, the ON and WHERE clauses are different. The ON clause in an outer join filters the rows of a cross product and then includes in the result the unmatched rows extended with nulls. The WHERE clause then eliminates rows from both the matched and unmatched rows produced by the outer join. You must take care to ensure that unmatched rows you want are not eliminated by the predicates in the WHERE clause.

  You cannot use subqueries inside an outer join ON clause.

• openstring-expression Specify an OPENSTRING clause to query within a file or a BLOB, treating the content of these sources as a set of rows. When doing so, you also specify information about the schema of the file or BLOB for the result set to be generated, since you are not querying a defined structure such as a table or view. This clause applies to the FROM clause of a SELECT statement. It is not supported for UPDATE or DELETE statements.

• apply-expression Use this clause to specify a join condition where the right table-expression is evaluated for every row in the left table-expression. For example, you can use an apply expression to evaluate a function, procedure, or derived table for each row in a table expression.

• contains-expression Use the CONTAINS clause after a table name to filter the table, and return only those rows matching the full text query specified with contains-query. Every matching row of the table is returned, along with a score column that can be referred to using score-correlation-name, if it is specified. If score-correlation-name is not specified, then the score column can be referred to by the default correlation name, contains.

• dml-derived-table Supports the use of a DML statement (INSERT, UPDATE, or DELETE) as a table expression in a query's FROM clause.