Skip Headers

Oracle® Data Cartridge Developer's Guide
10g Release 1 (10.1)

Part Number B10800-01
Go to Documentation Home
Home
Go to Book List
Book List
Go to Table of Contents
Contents
Go to Index
Index
Go to Master Index
Master Index
Go to Feedback page
Feedback

Go to previous page
Previous
Go to next page
Next
View PDF

10 Using Extensible Optimizer

This chapter contains these topics:

Overview of Query Optimization

Query Optimization is the process of choosing the most efficient way to execute a SQL statement. When the cost-based optimizer was offered for the first time with Oracle7, Oracle supported only standard relational data. The introduction of objects extended the supported datatypes and functions. The Extensible Indexing feature discussed in Chapter 9, " Defining Operators" introduces user-defined access methods.


See Also:


The extensible optimizer feature allows authors of user-defined functions and indexes to create statistics collection, selectivity, and cost functions that are used by the optimizer in choosing a query plan. The optimizer cost model is extended to integrate information supplied by the user to assess CPU and the I/O cost, where CPU cost is the number of machine instructions used, and I/O cost is the number of data blocks fetched.

Specifically, you now can:

Please note that only the cost-based optimizer has been enhanced; Oracle has not altered the operation of the rule-based optimizer.

The optimizer generates an execution plan for SQL queries and DML statements—SELECT, INSERT, UPDATE, or DELETE statements. For simplicity, we describe the generation of an execution plan in terms of a SELECT statement, but the process for DML statements is similar.

An execution plan includes an access method for each table in the FROM clause, and an ordering, called the join order, of the tables in the FROM clause. System-defined access methods include indexes, hash clusters, and table scans. The optimizer chooses a plan by generating a set of join orders, or permutations, by computing the cost of each, and then by selecting the process with the lowest cost. For each table in the join order, the optimizer computes the cost of each possible access method and join method and chooses the one with the lowest cost. The cost of the join order is the sum of the access method and join method costs. The costs are calculated using algorithms which together comprise the cost model. The cost model includes varying level of detail about the physical environment in which the query is executed.

The optimizer uses statistics about the objects referenced in the query to compute the selectivity and costs. The statistics are gathered using the DBMS_STATS package. The selectivity of a predicate is the fraction of rows in a table that is chosen by the predicate, and it is a number between 0 and 1.


See Also:

PL/SQL Packages and Types Reference for information about DBMS_STATS

The Extensible Indexing feature allows users to define new operators, indextypes, and domain indexes. For user-defined operators and domain indexes, the Extensible Optimizer feature enables you to control the three main components used by the optimizer to select an execution plan:

In the following sections, we describe each of these components in greater detail.

Statistics

Statistics for tables and indexes can be generated by using the DBMS_STATS package. In general, the more accurate the statistics, the better the execution plan generated by the optimizer.

User-Defined Statistics

The Extensible Optimizer feature lets you define statistics collection functions for domain indexes, indextypes, datatypes, individual table columns, and partitions. This means that whenever a domain index is analyzed, a call is made to the user-specified statistics collection function. The database does not know the representation and meaning of the user-collected statistics.

In addition to domain indexes, Oracle supports user-defined statistics collection functions for individual columns of a table, and for user-defined datatypes. In the former case, whenever a column is analyzed, the user-defined statistics collection function is called to collect statistics in addition to any standard statistics that the database collects. If a statistics collection function exists for a datatype, it is called for each column of the table being analyzed that has the required type.

The cost of evaluating a user-defined function depends on the algorithm and the statistical properties of its arguments. It is not practical to store statistics for all possible combinations of columns that could be used as arguments for all functions. Therefore, Oracle maintains only statistics on individual columns. It is also possible that function costs depend on the different statistical properties of each argument. Every column could require statistics for every argument position of every applicable function. Oracle does not support such a proliferation of statistics and cost functions because it would decrease performance.

A user-defined function to drop statistics is required whenever there is a user-defined statistics collection function.

User-Defined Statistics for Partitioned Objects

Since domain indexes cannot be partitioned, a user-defined statistics collection function collects only global statistics on the non-partitioned index. When you specify a list of partitions through the DBMS_STATS package, the information is not passed to user-defined statistics collection functions.

Selectivity

The optimizer uses statistics to calculate the selectivity of predicates. The selectivity is the fraction of rows in a table or partition that is chosen by the predicate. It is a number between 0 and 1. The selectivity of a predicate is used to estimate the cost of a particular access method; it is also used to determine the optimal join order. A poor choice of join order by the optimizer could result in a very expensive execution plan.

Currently, the optimizer uses a standard algorithm to estimate the selectivity of selection and join predicates. However, the algorithm does not always work well in cases in which predicates contain functions or type methods. In addition, predicates can contain user-defined operators about which the optimizer does not have any information. In that case the optimizer cannot compute an accurate selectivity.

User-Defined Selectivity

For greater control over the optimizer's selectivity estimation, this feature lets you specify user-defined selectivity functions for predicates containing user-defined operators, standalone functions, package functions, or type methods. The user-defined selectivity function is called by the optimizer whenever it encounters a predicate with one of the following forms:

operator(...) relational_operator <constant>

<constant> relational_operator operator(...)

operator(...) LIKE <constant>

where

  • operator(...) is a user-defined operator, standalone function, package function, or type method,

  • relational_operator is one of {<, <=, =, >=, >}, and

  • <constant> is a constant value expression or bind variable.

For such cases, users can define selectivity functions associated with operator(...). The arguments to operator can be columns, constants, bind variables, or attribute references. When optimizer encounters such a predicate, it calls the user-defined selectivity function and passes the entire predicate as an argument (including the operator, function, or type method and its arguments, the relational operator relational_operator, and the constant expression or bind variable). The return value of the user-defined selectivity function must be expressed as a percent, and be between 0 and 100 inclusive; the optimizer ignores values outside this range.

Wherever possible, the optimizer uses user-defined selectivity values. However, this is not possible in the following cases:

  • The user-defined selectivity function returns an invalid value (less than 0 or greater than 100)

  • There is no user-defined selectivity function defined for the operator, function, or method in the predicate

  • The predicate does not have one of the preceding forms: for example, operator(...) + 3 relational_operator <constant>

In each of these cases, the optimizer uses heuristics to estimate the selectivity.

Cost

The optimizer estimates the cost of various access paths to choose an optimal plan. For example, it computes the CPU and I/O cost of using an index and a full table scan to choose between the two. However, with regard to domain indexes, the optimizer does not know the internal storage structure of the index, and so it cannot compute a good estimate of the cost of a domain index.

User-Defined Cost

For greater flexibility, the cost model has been extended to let you define costs for domain indexes, index partitions, and user-defined standalone functions, package functions, and type methods. The user-defined costs can be in the form of default costs that the optimizer looks up, or they can be full-fledged cost functions which the optimizer calls to compute the cost.

Like user-defined selectivity statistics, user-defined cost statistics are optional. If no user-defined cost is available, the optimizer uses heuristics to compute an estimate. However, in the absence of sufficient useful information about the storage structures in user-defined domain indexes and functions, such estimates can be very inaccurate and result in the choice of a sub-optimal execution plan.

User-defined cost functions for domain indexes are called by the optimizer only if a domain index is a valid access path for a user-defined operator (for details regarding when this is true, see the discussion of user-defined indexing in the previous chapter). User-defined cost functions for functions, methods and domain indexes are only called when a predicate has one of the following forms:

operator(...) relational_operator <constant>

<constant> relational_operator operator(...)

operator(...) LIKE <constant>

where

  • operator(...) is a user-defined operator, standalone function, package function, or type method,

  • relational_operator is one of {<, <=, =, >=, >}, and

  • <constant> is a constant value expression or bind variable.

This is, of course, identical to the conditions for user-defined selectivity functions.

User-defined cost functions can return three cost values, each value representing the cost of a single execution of a function or domain index implementation:

  • CPU — the number of machine cycles executed by the function or domain index implementation. This does not include the overhead of invoking the function.

  • I/O — the number of data blocks read by the function or domain index implementation. For a domain index, this does not include accesses to the Oracle table. The multiblock I/O factor is not passed to the user-defined cost functions.

  • NETWORK — the number of data blocks transmitted. This is valid for distributed queries as well as functions and domain index implementations. For Oracle this cost component is not used and is ignored; however, as described in the following sections, the user is required to stipulate a value so that backward compatibility is facilitated when this feature is introduced.

The optimizer computes a composite cost from these cost values.

The package DBMS_ODCI contains a function estimate_cpu_units to help get the CPU and I/O cost from input consisting of the elapsed time of a user function. estimate_cpu_units measures CPU units by multiplying the elapsed time by the processor speed of the machine and returns the approximate number of CPU instructions associated with the user function. (For a multiprocessor machine, estimate_cpu_units considers the speed of a single processor.)


See Also:

PL/SQL Packages and Types Reference for information about package DBMS_ODCI


Optimizer Parameters

The cost of a query is a function of the cost values discussed in the preceding section. The settings of optimizer initialization parameters determine which cost to minimize. If optimizer_mode is first_rows, the resource cost of returning a single row is minimized, and the optimizer mode is passed to user-defined cost functions. Otherwise, the resource cost of returning all rows is minimized.

Defining Statistics, Selectivity, and Cost Functions

You can compute and store user-defined statistics for domain indexes and columns. User-defined selectivity and cost functions for functions and domain indexes can use both standard and user-defined statistics in their computation. The internal representation of these statistics need not be known to Oracle, but you must provide methods for their collection. You are solely responsible for defining the representation of such statistics and for maintaining them. Note that user-collected statistics are used only by user-defined selectivity and cost functions; the optimizer uses only its standard statistics.

User-defined statistics collection, selectivity, and cost functions must be defined in a user-defined type. Depending on the functionality you want it to support, this type must implement as methods some or all of the functions defined in the system interface ODCIStats, Oracle Data Cartridge interface Statistics, in Chapter 21, " Extensible Optimizer Interface".

The following example shows a type definition (or the outline of one) that implements all the functions in the ODCIStats interface.

CREATE TYPE my_statistics AS OBJECT (

   -- Function to get current interface
   FUNCTION ODCIGetInterfaces(ifclist OUT ODCIObjectList) RETURN NUMBER,

   -- User-defined statistics functions
   FUNCTION ODCIStatsCollect(col ODCIColInfo, options ODCIStatsOptions,
      statistics OUT RAW, env ODCIEnv) RETURN NUMBER,
   FUNCTION ODCIStatsCollect(ia ODCIIndexInfo, options ODCIStatsOptions,
      statistics OUT RAW, env ODCIEnv) RETURN NUMBER,
   FUNCTION ODCIStatsDelete(col ODCIColInfo, statistics OUT RAW, env ODCIEnv) 
      RETURN NUMBER,
   FUNCTION ODCIStatsDelete(ia ODCIIndexInfo, statistics OUT RAW, env ODCIEnv) 
      RETURN NUMBER,
   
   -- User-defined selectivity function
   FUNCTION ODCIStatsSelectivity(pred ODCIPredInfo, sel OUT NUMBER, args
      ODCIArgDescList, start <function_return_type>,
      stop <function_return_type>, <list of function arguments>, 
      env ODCIEnv) RETURN NUMBER,

   -- User-defined cost function for functions and type methods
   FUNCTION ODCIStatsFunctionCost(func ODCIFuncInfo, cost OUT ODCICost,
      args ODCIArgDescList, <list of function arguments>) RETURN NUMBER,

   -- User-defined cost function for domain indexes
   FUNCTION ODCIStatsIndexCost(ia ODCIIndexInfo, sel NUMBER,
      cost OUT ODCICost, qi ODCIQueryInfo, pred ODCIPredInfo,         
      args ODCIArgDescList, start <operator_return_type>,
      stop <operator_return_type>, <list of operator value arguments>, 
      env ODCIEnv) RETURN NUMBER
)

The object type that you define, referred to as a statistics type, need not implement all the functions from ODCIStats. User-defined statistics collection, selectivity, and cost functions are optional, so a statistics type may contain only a subset of the functions in ODCIStats. Table 10-1 lists the type methods and default statistics associated with different kinds of schema objects.

Table 10-1 Statistics Methods and Default Statistics for Various Schema Objects

ASSOCIATE STATISTICS WITH Statistics Type Methods Used Default Statistics Used
column
ODCIStatsCollect, ODCIStatsDelete

object type
ODCIStatsCollect, ODCIStatsDelete, ODCIStatsFunctionCost, ODCIStatsSelectivity
cost, selectivity
function
ODCIStatsFunctionCost, ODCIStatsSelectivity
cost, selectivity
package
ODCIStatsFunctionCost, ODCIStatsSelectivity
cost, selectivity
index
ODCIStatsCollect, ODCIStatsDelete, ODCIIndexCost
cost
indextype
ODCIStatsCollect, ODCIStatsDelete, ODCIIndexCost
cost

The types of the parameters of statistics type methods are system-defined ODCI datatypes. These are described in Chapter 20 and Chapter 21.

The selectivity and cost functions must not change any database or package state. Consequently, no SQL DDL or DML operations are permitted in the selectivity and cost functions. If such operations are present, the functions will not be called by the optimizer.

User-Defined Statistics Functions

There are two user-defined statistics collection functions, one for collecting statistics and the other for deleting them.

The first, ODCIStatsCollect, is used to collect user-defined statistics; its interface depends on whether a column or domain index is being analyzed. It is called when analyzing a column of a table or a domain index and takes two parameters:

  • col for the column being analyzed, or ia for the domain index being analyzed;

  • options for options specified in the DBMS_STATS package.

As mentioned, the database does not interpret statistics collected by ODCIStatsCollect. You can store output in a user-managed format or in a dictionary table, as described in Chapter 21, " Extensible Optimizer Interface". The statistics collected by the ODCIStatsCollect functions are returned in the output parameter, statistics, as a RAW datatype.

User-collected statistics are deleted by calling the ODCIStatsDelete function whose interface depends on whether the statistics for a column or domain index are being dropped. It takes a single parameter: col, for the column whose user-defined statistics need to be deleted, or ia, for the domain index whose statistics are to be deleted.

If a user-defined ODCIStatsCollect function is present in a statistics type, the corresponding ODCIStatsDelete function must also be present.

The return values of the ODCIStatsCollect and ODCIStatsDelete functions must be Success (indicating success), Error (indicating an error), or Warning (indicating a warning); these return values are defined in a system package ODCIConst.

User-Defined Selectivity Functions

You will recall that user-defined selectivity functions are used only for predicates of the following forms:

operator(...) relational_operator <constant>

<constant> relational_operator operator(...)

operator(...) LIKE <constant>

A user-defined selectivity function, ODCIStatsSelectivity, takes five sets of input parameters that describe the predicate:

  • pred describing the function operator and the relational operator relational_operator;

  • args describing the start and stop values (that is, <constant>) of the function and the actual arguments to the function (operator());

  • start whose datatype is the same as that of the function's return value, describing the start value of the function;

  • stop whose datatype is the same as that of the function's return value, describing the stop value of the function;

  • and a list of function arguments whose number, position, and type must match the arguments of the function operator.

The computed selectivity is returned in the output parameter sel as a number between 0 and 100 (inclusive) that represents a percentage. The optimizer ignores numbers less than 0 or greater than 100 as invalid values.

The return value of the ODCIStatsSelectivity function must be

  • Success indicating success, or

  • Error indicating an error, or

  • Warning indicating a warning.

As an example, consider a function myFunction defined as follows:

myFunction (a NUMBER, b VARCHAR2(10)) return NUMBER

A user-defined selectivity function for the function myFunction would be as follows:

ODCIStatsSelectivity(pred ODCIPredInfo, sel OUT NUMBER, args ODCIArgDescList,
   start NUMBER, stop NUMBER, a NUMBER, b VARCHAR2(10), env ODCIEnv) 
   return NUMBER

If the function myFunction is called with literal arguments, for example,

myFunction (2, 'TEST') > 5 

then the selectivity function is called as follows:

ODCIStatsSelectivity(<ODCIPredInfo constructor>, sel,
   <ODCIArgDescList constructor>, 5, NULL, 2, 'TEST', <ODCIEnv flag>)

If, on the other hand, the function myFunction is called with some non-literals—for example:

myFunction(Test_tab.col_a, 'TEST')> 5 

where col_a is a column in table Test_tab, then the selectivity function is called as follows:

ODCIStatsSelectivity(<ODCIPredInfo constructor>, sel,
   <ODCIArgDescList constructor>, 5, NULL, NULL, 'TEST', <ODCIEnv flag>)

In other words, the start, stop, and function argument values are passed to the selectivity function only if they are literals; otherwise they are NULL. The ODCIArgDescList descriptor describes all its following arguments.

User-Defined Cost Functions for Functions

As already mentioned, user-defined cost functions are only used for predicates of the following forms:

operator(...) relational_operator <constant>

<constant> relational_operator operator(...)

operator(...) LIKE <constant>

You can define a function, ODCIStatsFunctionCost, for computing the cost of standalone functions, package functions, or type methods. This function takes three sets of input parameters describing the predicate:

  • func describing the function operator

  • args describing the actual arguments to the function operator

  • a list of function arguments whose number, position, and type must match the arguments of the function operator

The ODCIStatsFunctionCost function returns its computed cost in the cost parameter. As mentioned, the returned cost can have two components, CPU and I/O, which are combined by the optimizer to compute a composite cost. The costs returned by user-defined cost functions must be positive whole numbers. Invalid values are ignored by the optimizer.

The return value of the ODCIStatsFunctionCost function must be one of:

  • Success indicating success

  • Error indicating an error

  • Warning indicating a warning

Consider a function myFunction defined as follows:

myFunction (a NUMBER, b VARCHAR2(10)) return NUMBER

A user-defined cost function for the function myFunction would be coded as follows:

ODCIStatsFunctionCost(func ODCIFuncInfo, cost OUT ODCICost,
   args ODCIArgDescList, a NUMBER, b VARCHAR2(10), env ODCIEnv) return NUMBER

If the function myFunction is called with literal arguments,

myFunction(2, 'TEST') > 5, 

then the cost function is called in this manner:

ODCIStatsFunctionCost(<ODCIFuncInfo constructor>, cost,
   <ODCIArgDescList constructor>, 2, 'TEST', <ODCIEnv flag>)

If, on the other hand, the function myFunction is called with some non-literals,

myFunction(Test_tab.col_a, 'TEST') > 5 

where col_a is a column in table Test_tab, then the cost function is called like this:

ODCIStatsFunctionCost(<ODCIFuncInfo constructor>, cost,
   <ODCIArgDescList constructor>, NULL, 'TEST', <ODCIEnv flag>)

In summary, function argument values are passed to the cost function only if they are literals; otherwise, they are NULL. ODCIArgDescList describes all the arguments that follow it.

User-Defined Cost Functions for Domain Indexes

User-defined cost functions for domain indexes are used for the same type of predicates mentioned previously, except that operator must be a user-defined operator for which a valid domain index access path exists.

The ODCIStatsIndexCost function takes these sets of parameters:

  • ia describing the domain index

  • sel representing the user-computed selectivity of the predicate

  • cost giving the computed cost

  • qi containing additional information about the query

  • pred describing the predicate

  • args describing the start and stop values (that is, <constant>) of the operator and the actual arguments to the operator operator

  • start, whose datatype is the same as that of the operator's return value, describing the start value of the operator

  • stop whose datatype is the same as that of the operator's return value, describing the stop value of the operator

  • a list of operator value arguments whose number, position, and type must match the arguments of the operator operator. The value arguments of an operator are the arguments excluding the first argument.

  • env, an environment flag set by the server to indicate which call is being made in cases where multiple calls are made to the same routine. The flag is reserved for future use; currently it is always set to 0.

The computed cost of the domain index is returned in the output parameter, cost.

ODCIStatsIndexCost returns

  • Success indicating success, or

  • Error indicating an error, or

  • Warning indicating a warning.

Consider an operator

Contains(a_string VARCHAR2(2000), b_string VARCHAR2(10))

that returns 1 or 0 depending on whether or not the string b_string is contained in the string a_string. Further, assume that the operator is implemented by a domain index. A user-defined index cost function for this domain index would be coded like this:

ODCIStatsIndexCost(ia ODCIIndexInfo, sel NUMBER, cost OUT ODCICost,
   qi ODCIQueryInfo, pred ODCIPredInfo, args ODCIArgDescList,
   start NUMBER, stop NUMBER, b_string VARCHAR2(10), env ODCIEnv) return NUMBER

Note that the first argument, a_string, of Contains does not appear as a parameter of ODCIStatsIndexCost. This is because the first argument to an operator must be a column for the domain index to be used, and this column information is passed in through the ODCIIndexInfo parameter. Only the operator arguments after the first (the "value" arguments) must appear as parameters to the ODCIStatsIndexCost function.

If the operator is called,

Contains(Test_tab.col_c,'TEST') <= 1

then the index cost function is called like this:

ODCIStatsIndexCost(<ODCIIndexInfo constructor>, sel, cost,
   <ODCIQueryInfo constructor>, <ODCIPredInfo constructor>, 
   <ODCIArgDescList constructor>, NULL, 1, 'TEST', <ODCIEnv flag>)

In summary, the start, stop, and operator argument values are passed to the index cost function only if they are literals; otherwise they are NULL. The ODCIArgDescList descriptor describes all its following arguments.

Using User-Defined Statistics, Selectivity, and Cost

Statistics types act as interfaces for user-defined functions that influence the choice of an execution plan by the optimizer. However, for the optimizer to be able to use a statistics type requires a mechanism to bind the statistics type to a database object (column, standalone function, object type, index, indextype or package; you cannot associate a statistics type with a partition of a table or a partition of a domain index). Creating this association is the job of the ASSOCIATE STATISTICS command. The following sections describe this command in more detail.

User-Defined Statistics

User-defined statistics functions are relevant for columns (both standard SQL datatypes and object types) and domain indexes. The functions ODCIStatsSelectivity, ODCIStatsFunctionCost, and ODCIStatsIndexCost are not used for user-defined statistics, so statistics types used only to collect user-defined statistics need not implement these functions. The following sections describe how column and index user-defined statistics are collected.

User-collected statistics can either be stored in some predefined dictionary tables or users could create their own tables. The latter approach requires that privileges on these tables be administered properly, backup and restoration of these tables be done along with other dictionary tables, and point-in-time recovery considerations be resolved.

Statistics are stored in a predefined system table for use by user-defined selectivity and cost functions. Three system views of this table are available:

Column Statistics

Consider a table Test_tab defined as follows:

CREATE TABLE Test_tab (
   col_a    NUMBER,
   col_b    typ1,
   col_c    VARCHAR2(2000)
)

where typ1 is an object type. Suppose that stat is a statistics type that implements ODCIStatsCollect and ODCIStatsDelete functions. User-defined statistics are collected by the DBMS_STATS package for the column col_b if we bind a statistics type with the column as follows:

ASSOCIATE STATISTICS WITH COLUMNS Test_tab.col_b USING stat

A list of columns can be associated with the statistics type stat. Note that Oracle supports only associations with top-level columns, not attributes of object types; if you wish, the ODCIStatsCollect function can collect individual attribute statistics by traversing the column.

Another way to collect user-defined statistics is to declare an association with a datatype as follows:

ASSOCIATE STATISTICS WITH TYPES typ1 USING stat_typ1

which declares stat_typ1 as the statistics type for the type typ1. When the table Test_tab is analyzed with this association, user-defined statistics are collected for the column col_b using the ODCIStatsCollect function of statistics type stat_typ1.

Individual column associations always have precedence over associations with types. Thus, in the preceding example, if both ASSOCIATE STATISTICS commands are issued, DBMS_STATS would use the statistics type stat (and not stat_typ1) to collect user-defined statistics for column col_b. It is also important to note that standard statistics, if possible, are collected along with user-defined statistics.

User-defined statistics are deleted using the ODCIStatsDelete function from the same statistics type that was used to collect the statistics.

Associations defined by the ASSOCIATE STATISTICS command are stored in a dictionary table called ASSOCIATION$.

Only user-defined datatypes can have statistics types associated with them; you cannot declare associations for standard SQL datatypes.

Domain Index Statistics

A domain index has an indextype. A statistics type for a domain index is defined by associating it either with the index or its indextype. Consider the following example using the table Test_tab we defined earlier:

CREATE INDEX Test_indx ON Test_tab(col_a)
INDEXTYPE IS indtype PARAMETERS('example');

CREATE OPERATOR userOp BINDING (NUMBER) RETURN NUMBER
USING userOp_func;

CREATE INDEXTYPE indtype
FOR userOp(NUMBER)
USING imptype;

Here, indtype is the indextype, userOp is a user-defined operator supported by indtype, userOp_func is the functional implementation of userOp, and imptype is the implementation type of the indextype indtype.

A statistics type stat_Test_indx can be associated with the index Test_indx as follows:

ASSOCIATE STATISTICS WITH INDEXES Test_indx USING stat_Test_indx 

When the domain index Test_indx is analyzed, user-defined statistics for the index are collected by calling the ODCIStatsCollect function of stat_Test_indx.

If a statistics type association is not defined for a specific index, Oracle looks for a statistics type association for the indextype of the index. In the preceding example, a statistics type stat_indtype can be associated with the indextype indtype as follows:

ASSOCIATE STATISTICS WITH INDEXTYPES indtype USING stat_indtype

When the domain index Test_indx is analyzed and no statistics type association has been defined for the index Test_indx, then user-defined statistics for the index are collected by calling the ODCIStatsCollect function of stat_indtype.

Thus, individual domain index associations always have precedence over associations with the corresponding indextypes.

Domain index statistics are dropped using the ODCIStatsDelete function from the same statistics type that was used to collect the statistics.

User-Defined Selectivity

Selectivity functions are used by the optimizer to compute the selectivity of predicates in a query. The predicates must have one of the appropriate forms and can contain user-defined operators, standalone functions, package functions, or type methods. Selectivity computation for each is described in the following sections.

User-Defined Operators

Consider the example laid out earlier, and suppose that the following association is declared:

ASSOCIATE STATISTICS WITH FUNCTIONS userOp_func USING stat_userOp_func

Now, if the following predicate

userOp(Test_tab.col_a) = 1

is encountered, the optimizer calls the ODCIStatsSelectivity function (if present) in the statistics type stat_userOp_func that is associated with the functional implementation of the userOp_func of the userOp operator.

Standalone Functions

If the association

ASSOCIATE STATISTICS WITH FUNCTIONS myFunction USING stat_MyFunction

is declared for a standalone function myFunction, then the optimizer calls the ODCIStatsSelectivity function (if present) in the statistics type stat_myFunction for the following predicate (for instance):

myFunction(Test_tab.col_a, 'TEST') = 1.

Package Functions

If the association

ASSOCIATE STATISTICS WITH PACKAGES Demo_pack USING stat_Demo_pack

is declared for a package Demo_pack, then the optimizer calls the ODCIStatsSelectivity function (if present) in the statistics type stat_Demo_pack for the following predicate (for instance):

Demo_pack.myDemoPackFunction(Test_tab.col_a, 'TEST') = 1

where myDemoPackFunction is a function in Demo_pack.

Type Methods

If the association

ASSOCIATE STATISTICS WITH TYPES Example_typ USING stat_Example_typ

is declared for a type Example_typ, then the optimizer calls the ODCIStatsSelectivity function (if present) in the statistics type stat_Example_typ for the following predicate (for instance):

myExampleTypMethod(Test_tab.col_b) = 1

where myExampleTypMethod is a method in Example_typ.

Default Selectivity

An alternative to selectivity functions is user-defined default selectivity. The default selectivity is a value (between 0% and 100%) that is looked up by the optimizer instead of calling a selectivity function. Default selectivities can be used for predicates with user-defined operators, standalone functions, package functions, or type methods.

The following command:

ASSOCIATE STATISTICS WITH FUNCTIONS myFunction DEFAULT SELECTIVITY 20

declares that the following predicate, for instance,

myFunction(Test_tab.col_a) = 1

always has a selectivity of 20 percent (or 0.2) regardless of the parameters of myFunction, or the comparison operator =, or the constant 1. The optimizer uses this default selectivity instead of calling a selectivity function.

An association can be declared using either a statistics type or a default selectivity, but not both. Thus, the following statement is illegal:

ASSOCIATE STATISTICS WITH FUNCTIONS myFunction USING stat_myFunction
   DEFAULT SELECTIVITY 20

The following are some more examples of default selectivity declarations:

ASSOCIATE STATISTICS WITH PACKAGES Demo_pack DEFAULT SELECTIVITY 20
ASSOCIATE STATISTICS WITH TYPES Example_typ DEFAULT SELECTIVITY 20

User-Defined Cost

The optimizer uses user-defined cost functions to compute the cost of predicates in a query. The predicates must have one of the forms listed earlier and can contain user-defined operators, standalone functions, package functions, or type methods. In addition, user-defined cost functions are also used to compute the cost of domain indexes. Cost computation for each is described in the following sections.

User-defined Operators

Consider the example outlined in the preceding section, and suppose that the following associations are declared:

ASSOCIATE STATISTICS WITH INDEXES Test_indx USING stat_Test_indx
ASSOCIATE STATISTICS WITH FUNCTIONS userOp USING stat_userOp_func

Consider the following predicate:

userOp(Test_tab.col_a) = 1.

If the domain index Test_indx implementing userOp is being evaluated, the optimizer calls the ODCIStatsIndexCost function (if present) in the statistics type stat_Test_indx. If the domain index is not used, however, the optimizer calls the ODCIStatsFunctionCost function (if present) in the statistics type stat_userOp to compute the cost of the functional implementation of the operator userOp.

Standalone Functions

If the association

ASSOCIATE STATISTICS WITH FUNCTIONS myFunction USING stat_myFunction

is declared for a standalone function myFunction, then the optimizer calls the ODCIStatsFunctionCost function (if present) in the statistics type stat_myFunction for the following predicate (for instance):

myFunction(Test_tab.col_a, 'TEST') = 1

User-defined function costs do not influence the choice of access methods; they are only used for ordering predicates, described in Chapter 21, " Extensible Optimizer Interface".

Package Functions

If the association

ASSOCIATE STATISTICS WITH PACKAGES Demo_pack USING stat_Demo_pack;

is declared for a package Demo_pack, then the optimizer calls the ODCIStatsFunctionCost function, if present, in the statistics type stat_Demo_pack for the following predicate,

Demo_pack.myDemoPackFunction(Test_tab.col_a) = 1

where myDemoPackFunction is a function in Demo_pack.

Type Methods

If the association

ASSOCIATE STATISTICS WITH TYPES Example_typ USING stat_Example_typ;

is declared for a type Example_typ, then the optimizer calls the ODCIStatsFunctionCost function, if present, in the statistics type stat_Example_typ for the following predicate,

myExampleTypMethod(Test_tab.col_b) = 1

where myExampleTypMethod is a method in Example_typ.

Default Cost

Like default selectivity, default costs can be used for predicates with user-defined operators, standalone functions, package functions, or type methods. So, the following command

ASSOCIATE STATISTICS WITH INDEXES Test_indx DEFAULT COST (100, 5, 0)

declares that using the domain index Test_indx to implement the following predicate (to select one example)

userOp(Test_tab.col_a) = 1

always has a CPU cost of 100, I/O of 5, and network of 0 (the network cost is ignored in Oracle) regardless of the parameters of userOp, the comparison operator "=", or the constant "1". The optimizer uses this default cost instead of calling an ODCIStatsIndexCost cost function.

You can declare an association using either a statistics type or a default cost but not both. Thus, the following statement is illegal:

ASSOCIATE STATISTICS WITH INDEXES Test_indx USING stat_Test_indx
   DEFAULT COST (100, 5, 0)

The following are some more examples of default cost declarations:

ASSOCIATE STATISTICS WITH FUNCTIONS myFunction DEFAULT COST (100, 5, 0)
ASSOCIATE STATISTICS WITH PACKAGES Demo_pack DEFAULT COST (100, 5, 0)
ASSOCIATE STATISTICS WITH TYPES Example_typ DEFAULT COST (100, 5, 0)
ASSOCIATE STATISTICS WITH INDEXTYPES indtype DEFAULT COST (100, 5, 0)

Declaring a NULL Association for an Index or Column

An association of a statistics type defined for an indextype or object type is inherited by index instances of that indextype and by columns of that object type. An inherited association can be overridden by explicitly defining a different association for an index instance or column, but there may be occasions when you would prefer an index or column not to have any association at all. For example, for a particular query the benefit of a better plan may not outweigh the additional compilation time incurred by invoking the cost or selectivity functions. For cases like this, you can use the ASSOCIATE command to declare a NULL association for a column or index.

ASSOCIATE STATISTICS WITH COLUMNS <columns> NULL;
ASSOCIATE STATISTICS WITH INDEXES <indexes> NULL;

If the NULL association is specified, the schema object does not inherit any statistics type from the column type or the indextype. A NULL association also precludes default values.

How Statistics Are Affected by DDL Operations

Partition-level and schema object-level aggregate statistics are affected by DDL operations in the same way as standard statistics. Table 10-2 summarizes the effects.

Table 10-2 Effects of DDL on Partition and Global Statistics

Operation Effect on Partition Statistics Effect on Global Statistics
ADD PARTITION
None No Action
DROP PARTITION
Statistics deleted Statistics recalculated (if _minimal_stats_aggregation is FALSE, otherwise no effect)
SPLIT PARTITION
Statistics deleted None
MERGE PARTITION
Statistics deleted None
TRUNCATE PARTITION
Statistics deleted None
EXCHANGE PARTITION
Statistics deleted Statistics recalculated (if _minimal_stats_aggregation is FALSE, otherwise no effect)
REBUILD PARTITION
None None
MOVE PARTITION
None None
RENAME PARTITION
None None

If statistics for any partition are deleted, aggregate statistics for that object are deleted, and the aggregate statistics for the table or the index are recalculated.

If an existing partition is exchanged, or dropped with an ALTER TABLE DROP PARTITION statement, and the _minimal_stats_aggregation parameter is set to FALSE, the statistics for that partition are deleted, and the aggregate statistics of the table or index are recalculated.

Predicate Ordering

In the absence of an ORDERED_PREDICATES hint, predicates (except those used for index keys) are evaluated in the order specified by the following rules:

Dependency Model

The dependency model reflects the actions that are taken when you issue any of the SQL commands described in Table 10-3.

Table 10-3 Dependency Model for DDLs

Command Action
DROP statistics_type 
if an association is defined with statistics_type, the command fails, otherwise the type is dropped
DROP statistics_type FORCE
calls DISASSOCIATE FORCE for all objects associated with the statistics_type; drops statistics_type
DROP object
calls DISASSOCIATE, drops object_type if DISASSOCIATE succeeds
ALTER TABLE DROP COLUMN
if association is present for the column, this calls DISASSOCIATE FORCE with column; if no entry in ASSOCIATION$ but there are entries in type USATS$, then ODCIStatsDelete for the columns is invoked
DISASSOCIATE
if user-defined statistics collected with the statistics type are present, the command fails
DISASSOCIATE FORCE
deletes the entry in ASSOCIATION$ and calls ODCIStatsDelete
Delete index statistics using the DBMS_STATISTICS package the ODCIStatsDelete function is invoked; if any errors are raised, statistics deletion fails and an error is reported
ASSOCIATE
if an association or user-defined statistics are present for the associated object, the command fails

Restrictions and Suggestions

A statistics type is an ordinary object type. Since an object type must have at least one attribute, a statistics type also must have at least one attribute. This will be a dummy attribute, however, since it will never be set or accessed.

Parallel Query

Domain indexes are non-partitioned and serial. The optimizer computes the composite cost of a domain index access path assuming a serial execution.

Distributed Execution

Oracle's distributed implementation does not support adding functions to the remote capabilities list. All functions referencing remote tables are executed as filters. The placement of the filters occurs outside the optimizer. The cost model reflects this implementation and does not attempt to optimize placement of these predicates.

Since predicates are not shipped to the remote site, you cannot use domain indexes on remote tables. Therefore, the DESCRIBE protocol is unchanged, and remote domain indexes are not visible from the local site.

Performance

The cost of execution of the queries remains the same with the extensible optimizer if the same plan is chosen. If a different plan is chosen, the execution time should be better assuming that the user-defined cost, selectivity, and statistics collection functions are accurate. In light of this, you are strongly encouraged to provide statistics collection, selectivity, and cost functions for user-defined structures because the optimizer defaults can be inaccurate and lead to an expensive execution plan.