INDEX
Create an index file
Syntax
INDEX ON <expKey> [TAG <cOrderName>] [TO <cOrderBagName>]
[FOR <lCondition>] [ALL]
[WHILE <lCondition>] [NEXT <nNumber>]
[RECORD <nRecord>] [REST]
[EVAL <bBlock>] [EVERY <nInterval>]
[UNIQUE] [ASCENDING|DESCENDING]
[USECURRENT] [ADDITIVE]
[CUSTOM] [NOOPTIMIZE]
Arguments
<expKey> is an expression that returns the key value to place in the index for each record in the current work area. <expKey> can be character, date, logical, or numeric type. The maximum length of the index key expression is determined by the driver.
TAG <cOrderName> is the name of the order to be created. <cOrderName> can be any Harbour expression that evaluates to a string constant.
TO <cOrderBagName> is the name of a disk file containing one or more orders. The active RDD determines the order capacity of an order bag. The default DBFNTX driver only supports single-order bags, while other RDDs may support multiple-order bags (e.g., the DBFCDX and DBFMDX drivers). You may specify <cOrderBagName> as the file name with or without a path name or extension. If an extension is not provided as part of <cOrderBagName>, Harbour will use the default extension of the current RDD.
Both the TAG and the TO clauses are optional, but you must use at least one of them.
FOR <lCondition> specifies the conditional set of records on which to create the order. Only those records that meet the condition are included in the resulting order. <lCondition> is an expression that may be no longer than 250 characters under the DBFNTX and DBFNDX drivers. The maximum value for these expressions is determined by the RDD. The FOR condition is stored as part of the order bag and used when updating or recreating the index using the REINDEX command. Duplicate key values are not added to the order bag.
Drivers that do not support the FOR condition will produce an “unsupported” error.
The FOR clause provides the only scoping that is maintained for all database changes. All other scope conditions create orders that do not reflect database updates.
ALL specifies all orders in the current or specified work area. ALL is the default scope of INDEX .
WHILE <lCondition> specifies another condition that must be met by each record as it is processed. As soon as a record is encountered that causes the condition to fail, the INDEX command terminates. If a WHILE clause is specified, the data is processed in the controlling order. The WHILE condition is transient (i.e., it is not stored in the file and not used for index updates and REINDEXing purposes). The WHILE clause creates temporary orders, but these orders are not updated.
Drivers that do not support the WHILE condition will produce an “unsupported” error.
Using the WHILE clause is more efficient and faster than using the FOR clause. The WHILE clause only processes data for which <lCondition> is true (.T.) from the current position. The FOR clause, however, processes all data in the data source.
NEXT <nNumber> specifies the portion of the database to process. If you specify NEXT, the database is processed in the controlling order for the <nNumber> number of identities. The scope is transient (i.e., it is not stored in the order and not used for REINDEXing purposes).
RECORD <nRecord> specifies the processing of the specified record.
REST specifies the processing of all records from the current position of the record pointer to the end of file (EOF).
EVAL <bBlock> evaluates a code block every <nInterval>, where <nInterval> is a value specified by the EVERY clause. The default value is 1. This is useful in producing a status bar or odometer that monitors the indexing progress. The return value of <bBlock> must be a logical data type. If <bBlock> returns false (.F.), indexing halts.
EVERY <nInterval> is a clause containing a numeric expression that modifies the number of times <bBlock> is EVALuated. The EVERY option of the EVAL clause offers a performance enhancement by evaluating the condition for every nth record instead of evaluating every record ordered. The EVERY keyword is ignored if you specify no EVAL condition.
UNIQUE specifies that the key value of each record inserted into the order be unique. Duplicate key values are not added to the order.
ASCENDING specifies that the keyed pairs be sorted in increasing order of value. If neither ASCENDING nor DESCENDING is specified, ASCENDING is assumed. Although not stored as an explicit part of the file, ASCENDING is an implicit file attribute that is understood by the REINDEX command.
Drivers that do not support the ASCENDING condition will produce an “unsupported” error. The following keywords are new to Harbour 5.3.
DESCENDING specifies that the keyed pairs be sorted in decreasing order of value. Using this keyword is the same as specifying the DESCEND() function within <expKey>, but without the performance penalty during order updates. If you create a DESCENDING index, you will not need to use the DESCEND() function during a SEEK. DESCENDING is an attribute of the file, where it is stored and used for REINDEXing purposes.
Drivers that do not support the DESCENDING condition will produce an “unsupported” error.
USECURRENT specifies that only records in the controlling order–and within the current range as specified by ORDSETSCOPE()–will be included in this order. This is useful when you have already created a conditional order and want to reorder the records which meet that condition, and/or to further restrict the records meeting a condition. If not specified, all records in the database file are included in the order.
ADDITIVE specifies that any open orders should remain open. If not specified, all open orders are closed before creating the new one. Note, however, that the production index file is never closed.
CUSTOM specifies that a custom built order will be created for RDDs that support them. A custom built order is initially empty, giving you complete control over order maintenance. The system does not automatically add and delete keys from a custom built order. Instead, you explicitly add and delete keys using ORDKEYADD() and ORDKEYDEL(). This capability is excellent for generating pick lists of specific records and other custom applications.
NOOPTIMIZE specifies that the FOR condition will not be optimized. If NOOPTIMIZE is not specified, the FOR condition will be optimized if the RDD supports optimization.
Description
The INDEX command adds a set of keyed pairs, ordered by <expKey> to a file specified by <cOrderBagName> using the database open in the current work area.
In RDDs that support production or structural indexes (e.g., DBFCDX, DBFMDX), if you specify a tag but do not specify an order bag, the tag is created and added to the order bag. If no production or structural index exists, it will be created and the tag will be added to it.
When using RDDs that support multiple order bags, you must explicitly SET ORDER (or ORDSETFOCUS()) to the desired controlling order. If you do not specify a controlling order, the data file will be viewed in natural order.
If <cOrderBagName> does not exist, it is created in accordance with the RDD in the current or specified work area.
If <cOrderBagName> exists and the RDD specifies that order bags can only contain a single order, <cOrderBagName> is erased and the new order is added to the order bag and to the order list in the current or specified work area.
If <cOrderBagName> exists and the RDD specifies that order bags can contain multiple tags, <cOrderName> is created if it does not already exist; otherwise, <cOrderName> is replaced in <cOrderBagName> and the order is added to the order list in the current or specified work area.
ASCENDING or DESCENDING specifies the sequence of keyed pairs in the order. If neither clause is specified, the default is ASCENDING.
If you specify the UNIQUE clause, the resulting order will contain only unique records. Some RDDs may do this by only including record references to a key value once. Others may produce a runtime recoverable error as a non-unique key insertion is attempted.
The EVAL clause lets you specify a code block to be evaluated as each record is placed in the order. The EVERY clause lets you modify how often <bBlock> is called. Instead of evaluation as each record is placed in the order, evaluation only occurs as every <nInterval> records are placed in the order.
The INDEX command accepts certain clauses that let the user create conditional and partial orders. Some orders are intended to be maintained across the application, others are considered “temporary” orders.
The FOR clause provides the only order scoping that is permanent and can be maintained across the life of the application. The string passed as the FOR condition is stored within the order for later use in maintaining the order. Though only accessing part of a database, orders created using this clause exist as long as the database is active. The FOR clause lets you create maintainable scoped orders.
The WHILE, NEXT, REST and RECORD clauses process data from the current position of the database cursor in the default or specified work area. If you specify these clauses, the order list remains open and the active order is used to organize the database while it is being created. These clauses let you create temporary (non-maintainable) orders. Orders created using these clauses contain records in which <lCondition> is true (.T.) at the location of the record pointer.
Notes
RDD support: Not all RDDs support all aspects of the INDEX command. See the “Replaceable Database Driver Architecture” chapter in the Drivers Guide for details on a particular RDD.
Although both the TAG and the TO clauses are optional, you must specify at least one of them.
Examples
. The following example creates a simple order (index) based on
one field (Acct):
USE Customer NEW
INDEX ON Customer->Acct TO CuAcct
. This example creates a conditional order (index) based on a
FOR clause. This index will contain only records whose field
TransDate contains a date greater than or equal to January 1, 1995:
USE Invoice NEW
INDEX ON Invoice->TransDate ;
TO InDate ;
FOR ( Invoice->TransDate >= CTOD( "01/01/95" ) )
. This example creates an order in a multiple-order bag (i.e., a
tag in an index that can support multiple tags in an index file):
USE Customer NEW
INDEX ON Customer->Acct TAG CuAcct TO Customer
. The following example creates an order that calls a routine,
MyMeter, during its creation:
#define MTR_INCREMENT 10
USE Customer NEW
INDEX ON Customer->Acct TO CuAcct EVAL ;
{|| MYMETER() } EVERY MTR_INCREMENT
FUNCTION MYMETER()
STATIC nRecsDone := 0
nRecsDone := += MTR_INCREMENT
? ( nRecsDone/LASTREC() ) * 100
RETURN (.T.)
Seealso
CLOSE, DBCREATEIND(), DBORDERINFO(), DBREINDEX()
