|
|
|
Access augmentation or optimizing |
Method and system for using dynamically generated code to perform index record retrieval in certain circumstances in a relational database manager6353820
Abstract
A system and method that enhances the index processing performance of a multi-layer relational database manager by expanding the code generation component layer of the database manager to include an index processing performance enhancing subroutine designed to execute functions performed by lower component layers substantially faster than if the functions were executed by such lower component layers. The subroutine includes logic for establishing the conditions under which the particular subroutine is invoked during the execution of a SQL request. The output code generated to execute a specific SQL query, including calls to the index processing subroutine in place of normally included calls to the lower component layer. This enables the generated code to perform lower component layer functions with specialized code designed to increase performance.
Claims
What is claimed is:
1. A method for enhancing the performance of a data manager that manages a relational database, the manager having a number of operatively coupled component layers for performing functions required for accessing the relational database file records in response to user query statements, the number of component layers including a first component layer for generating output code used in executing the query statement in response to input representing an optimized form of the query statement and a number of lower component layers, the method comprising the steps of:
(a) adding an index processing performance enhancing subroutine to a subroutine library included within the first component layer, the index processing performance enhancing subroutine being designed to execute index processing functions performed by lower component layers substantially faster than if the function were executed by the lower component layers, and including code for determining conditions under which the particular subroutine is to be invoked during query execution time;
(b) including calls in the generated output code to the index processing enhancing subroutine in place of normally included calls to the lower component layers for inserting the index processing enhancing subroutine into the generated code; and,
(c) including logic within the index processing enhancing subroutine for determining if the index processing enhancing subroutine can be reliably invoked based on a predetermined set of conditions.
2. The method of claim 1 wherein the calls included in the generated output code are in place of normally included calls to a prior index processing subroutine that unconditionally invoked lower component layers.
3. The method of claim 1 wherein the predetermined set of conditions indicate that there were no changes in index currency and in a fine level index used for retrieving the database file records.
4. The method of claim 3 wherein the fine level index contains entries that are lightly packed adjacent to one another in sorted order enabling the logic in the index processing enhancing subroutine to walk through the fine level index taking entries right out from under control of one of the number of lower component layers as search next requests are being processed.
5. The method of claim 4 wherein the first component layer includes a code generation component and the one of the lower component layers includes a relational file manager (RFM) component having a library for storing a number of routines for performing the lower layer index functions.
6. The method of claim 5 wherein the number of lower component layers further includes a third lower component layer operatively coupled to the first component layer, to the second lower component layer and to the relational database for executing input/output operations required for executing the query statement.
7. The method of claim 6 wherein in step (c), the first component layer invokes the index processing performance subroutine to call conditionally, the third lower component layer function based on the setting of a result processing indicator.
8. The method of claim 7 wherein the third lower component layer includes an input/output controller component that operatively couples to the relational database.
9. The method of claim 1 wherein the index processing enhancing subroutine is coded for performing a number of different index state tests for determining the set of predetermined conditions.
10. A data manager for managing a relational database, the manager having a number of operatively coupled component layers for performing functions required for accessing database file records of the relational database in response to user query statements, the number of component layers including a first component layer for generating output code used in executing the query statement in response to input representing an optimized form of the query statement and a number of lower component layers operatively coupled to the first component layer, to each other and to the relational database, the lower component layers, the first component layer further including:
(a) a subroutine library including an index processing performance enhancing subroutine designed to execute functions performed by lower component layers substantially faster than if the functions were executed by the lower component layers, the index processing subroutine including coded logic for determining conditions under which the subroutine is to be executed;
(b) the generated output code being stored in memory and including calls to the index processing subroutine stored in the subroutine library in place of normally included calls to a prior index processing subroutine that unconditionally invoked lower component layers; and,
(c) the generated output code when accessed from memory during execution of the query invoking the index processing subroutine to perform the lower layer functions.
11. The data manager of claim 10 wherein the particular conditions indicate there were no changes in index currency and in a fine level index used in retrieving the database file records.
12. The data manager of claim 11 wherein the fine level index contains entries that are lightly packed adjacent to one another in sorted order enabling the coded logic to walk through the fine coded level index taking entries out from under control of the second lower component layer.
13. The data manager of claim 12 wherein the first component layer includes a code generation component and the second lower component layer includes a relational file manager (RFM) component having a library for storing a number of routines for performing the index processing functions utilizing the fine level index.
14. The data manager of claim 13 wherein the number of lower component layers further includes a third lower component layer operatively coupled to the first component layer, to the second lower component layer and to the relational database for executing input/output operations required for executing the query statement.
15. The data manager of claim 14 wherein in the first component layer during the execution of the query invokes the index processing subroutine to invoke the third lower component layer function when a result processing indicator has been set.
16. The data manager of claim 15 wherein the third lower component layer includes an input/output controller component that operatively couples to the relational database.
17. An RDMS program product including a data manager for managing a relational database stored on a media as groups of program instructions, the instructions corresponding to a number of operatively coupled component layers for performing functions required for accessing database file records of the relational database in response to user query statements, the number of component layers including a first component layer for generating code used in executing the query statement in response to input representing an optimized form of the query statement and a number of lower component layers operatively coupled to the first component layer, to each other and to the relational database, the first component layer further including:
(a) a first group of instructions corresponding to a subroutine library that includes an index processing performance enhancing subroutine designed to execute functions performed by lower component layers substantially faster than if the functions were executed by the number of lower component layers, the index processing performance enhancing subroutine including code containing instructions for determining conditions under which the particular subroutine is to be executed;
(b) a second group of instructions corresponding to a code generation component for providing output code containing calls to the index processing performance enhancing subroutine in place of normally included calls to lower component layers that inserts the index processing subroutine into the output code; and,
(c) another group of instructions included within the index processing subroutine for causing the execution of the index processing subroutine to perform the second lower layer functions based on a particular set of conditions resulting in increased performance.
18. A memory for storing a data manager for accessing data records of a relational database during the execution of a query, the memory comprising:
(a) first component layer for generating code used in executing the query statement in response to input representing an optimized form of the query statement and a number of lower component layers operatively coupled to the first component layer, to each other and to the relational database, the lower component layers, the first component layer contained in the memory further including:
(1) a subroutine library including an index processing performance enhancing subroutine designed to execute functions performed by lower component layers substantially faster than if the functions were executed by the lower component layers, the row retrieval performance enhancing subroutine including code for determining conditions under which the particular subroutine is to be invoked during index processing;
(2) generated output code including calls to the index processing performance enhancing subroutine in place of normally included calls to the lower component layers for incorporating the index processing subroutine into the generated output code for execution of the query being processed; and,
(3) a memory area being assigned for storing the output code, the output code accessed during execution of the query causing the index processing subroutine to be executed for performing the lower layer functions as a function of a particular set of conditions so as to result in increased performance.
Description
BACKGROUND OF THE INVENTION
1. Field of Use
The present invention relates to data processing systems and more particularly to database management systems.
2. Prior Art
Typically, today's enterprise or legacy systems store large quantities of data in database systems accessed by database management system (DBMS) software. In such database systems, data is logically organized into relations or tables wherein each relation can be viewed as a table where each row is a tuple and each column is a component of the relation designating an attribute. It has become quite common to use relational database management systems (RDMS) for enabling users to enter queries derived from a database query language, such as SQL, into the database in order to obtain or extract requested data.
In compiling type database management systems, an application program containing database queries is processed for compilation prior to run time. This can be done and more frequently is done at run time by users of the INTEREL product discussed herein. Users of other database products such as DB2, do such processing prior to run time.
During compilation, database queries are passed to the database management system for compilation by a database management system compiler. The compiler translates the queries contained in the application program into machine language. Generally, a database compiler component referred to, as a query optimizer is included in the database management system to select the manner in which queries will be processed. The reason is because most users do not input queries in formats that suggest the most efficient way for the database management system to address the query. The query optimizer component analyzes how best to conduct the user's query of the database in terms of optimum speed in accessing the requested data. That is, the optimizer typically transforms a user query into an equivalent query that can be computed more efficiently. This operation is performed at compile time, in advance of execution.
A major component of the RDBMS is the database services component or module that supports the functions of SQL language, such as definition, access control, retrieval and update of user and system data. Such components may utilize a multilayer structure containing submodules or components for carrying out the required functions. For example, one such system includes a series of components or conceptually, a series of layers for carrying out the required functions for accessing data from the relational database. More specifically, a first layer functions as a SQL director component that handles requests at the interface to the requesting or calling application program. A second layer consists of two major components, an optimizer for optimizing the query and a RAM code generation component. The optimizer processes the query by for example, by determining the appropriate access plan strategy. The code generation component (Codgen) generates code according to such plan for accessing and processing the requested data. The access plan defines the type of access to each table, order of access, whether any sorts or joins are performed along with other related information.
The generated code is passed to a third layer that functions as a relational file manager (RFM) component. This component layer performs the relational file processing function of translating the code-generated requests into IO file read/write requests. A fourth layer that functions as an IO Controller performs the requested I/O operation designated by such IO file requests that results in reading/writing the relational database files in page increments. The described architecture is characteristic of the INTEREL product developed and marketed by Bull HN Information Systems Inc. For information concerning this product, reference may be made to the publication entitled, "Database Products INTEREL Reference Manual INTEREL Performance Guidelines, Copyright, 1996 by Bull HN Information as Systems Inc., Order No. LZ93 Rev01B.
It was noted that index searches are very common operations in relational databases. They occur when processing SELECT, UPDATE or DELETE statements. Because they occur so frequently, it is an area where performance improvement could result in a substantial benefit for a relational database system such as the above architecture. For example, the above architecture processes the following typical query as follows:
Select accountID from tableFunds where accountBalance>=50000;
With DDL:
create table tableFunds (accountld int, accountBalance numeric(11,0)
create index i1 on tableFunds (accountBalance);
After an examination of this query, it is seen that many accounts may qualify. Because the index is on the accountBalance column and index data is sorted order, the query would be processed in the following manner. First, the index entry for 50,000 must be retrieved, then the subsequent index entries (i.e., all for accountBalances of 50,000 or greater) must be retrieved until the index is exhausted.
The above architecture performs index processing in the following two steps. First, the Codegen component layer calls the RFM component layer to search for a specific index value. This is called a Find Index search which is used to locate the database key (DBK) of a record from an index key value provided by the caller (i.e. user's search request). Once the RFM component layer finds the index entry, it establishes a currency to it. This currency is control information that indicates which fine level index entry corresponds with the search request. This currency information is stored in the RFM component layer's schema structure. The RFM component layer establishes a currency ID for the currency from currency ID information that the code generation component layer sets in an RFM data structure (RFM_XPT) prior to the call.
In the second step, the Codegen component layer calls the RFM component layer to return the next index entry (i.e., Search Next Index entry). The index fine level entries are in sorted order and because the Codegen layer passes in the currency ID from the prior search, the RFM layer can go to the currency information stored in the schema structure and use it to find the next index entry without repeating the index search. After the RFM layer has identified the next index entry, it updates the currency information in the schema. This second step is repeated until query processing is complete. This process is quite time consuming in terms of the overhead expended in invoking/calling lower component layers to perform index processing.
Accordingly, it is a primary object of the present invention to provide a more efficient method and system for improving relational index processing.
SUMMARY OF THE INVENTION
The above objects are achieved in a preferred embodiment of the present invention that can be utilized in a relational database management System (RDMS) that implements the Structured Query Language (SQL) standard. The present invention is a system and method that enhances the index processing performance of a multi-layer relational database manager. According to the teachings of the invention, the code generation component layer of the database manager includes an index processing performance enhancing subroutine designed to execute functions performed by lower component layers substantially faster than if the functions were executed by such lower component layers. The output code generated by the code generation component layer includes calls to the index processing performance enhancing subroutine thereby incorporating such subroutine into the output code.
The subroutine includes logic for establishing the conditions under which the particular subroutine is invoked during the execution of a SQL request. In the preferred embodiment, the logic detects when there is more than one search next index operation is requested for a particular query. On the second search next index request, the subroutine logic examines a fine level index page (CI) from which the prior index entry was retrieved. This CI was obtained by RFM during the initial B Tree index search and resides in the buffer pool. If there have been no changes in index currency and in the fine level index, then the subroutine copies the fine level index entry to the requestor's key buffer along with a database key value. Also, the currency information is updated to point to the next fine level index entry.
When the logic detects the presence of above conditions, this eliminates having to make calls to the lower component layers (i.e., RFM and IO component layers), thus bypassing these layers. This results in substantial increase in performance.
It will be appreciated that not all sequential index searches processed by the output code and that the enhanced index processing subroutine will be able to bypass the lower component layers. That is, it is still necessary to call the layer that performs I/O operations (i.e., IO component layer) while doing a sequential index search if more than one result is required to be processed. This occurs in the case where a result is returned in response to a SELECT statement used in conjunction with a FETCH cursor statement and a subsequent query issues an index search to an identical index in another database (model).
To protect against any possible page (CI) integrity problem, the IO component layer must be called to refresh the fine level index CI pointer once the first or original search request is resumed. But, even for the case where the IO component layer must be called, performance is still greatly enhanced by bypassing the RFM component layer. According to the present invention, at code generation time, the code generation component layer places into the output code generated for a particular query, the control logic for determining when a result has been processed and for setting a result processed flag indicator. Those types of queries that do not return a result or return only a 1 result will not have the code generated for them that sets the result processed flag. Queries that fall into this category are DELETE, UPDATE and SELECT INTO (e.g. SELECT COUNT(*)).
The above objects and advantages of the present invention will be better understood from the following description when taken in conjunction with the accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 is an overall block diagram of a data processing system that utilizes the teachings of the present invention.
FIG. 2 is a block diagram illustrating the multi-layer organization of the relational database manager system (RDMS) of FIG. 1.
FIG. 3a is a block diagram illustrating in greater detail, the major components of the second layer of the RDMS of FIG. 2a.
FIG. 3b is a block diagram illustrating in greater detail, the code generation component of FIG. 2 designed to utilize the teachings of the present invention.
FIG. 3c illustrates the operational relationships between the second and third layers of the RDMS of FIG. 2a.
FIGS. 4a and 4b are flow diagrams used to describe the operation of the preferred embodiment of the present invention.
DESCRIPTION OF THE PREFERRED EMBODIMENT
FIG. 1
FIG. 1 is a block diagram of a conventional data processing system 10 that utilizes the system and method of the present invention. As shown, the system 10 includes a plurality of processing units 12-1 through 12-n which connect to a system bus 14 through their individual physical caches in common with a number of input/output units (IOUs) 16-1 through 16-n and a system control unit (SCU) 22. As shown, each IOU couples to a particular I/O subsystem (i.e., 19-1 through 19-n) which in turn connect to any one of a number of different types of devices both local and remote such as workstation 21-1 via a network 20 or disk mass storage units 21-n as indicated.
The SCU 22 connects to a number of memory units (MUs) 24-1 through 24-n. For the purpose of the present invention, system 10 may be considered conventional in design and may for example utilize a mainframe computer system such as the DPS9000 manufactured by Bull HN Information Systems Inc. which operates under the control of the GCOS8 operating system.
As shown, the system 10 further includes disk storage 21-n that contains the database system that utilizes the teachings of the present invention. It will be appreciated that the software components that comprise the database system including the software components of the present invention may be loaded into the system 10 in a conventional manner (e.g. via CDROM, disk, communications link, etc.). The database system includes a multi-layer relational database management system (RDMS) and a relational database containing established data files. The relational database management system processes all user requests for accessing the files contained on disk storage 21-n. Users initiate such requests via the network 20 by executing transaction processing routines or batch decision support programs via their workstation keyboard and/or via other input devices (e.g. mouse). The system 10 upon receiving an SQL query operates to initiate a search of the relational database files to obtain the data requested by the user.
In system 10, the relational database management system (RDMS) takes the form of the above mentioned INTEREL software which runs under the GCOS8 operating system. As shown, the RDMS contains a SQL Director component layer, a Codegen Executor component layer, a Relational File Manager (RFM) component layer and an IO Random Controller component layer. These component layers are shown in greater detail, in FIG. 2 along with other database related components.
FIG. 2--Multi-layer RDMS Organization
FIG. 2 depicts the major components of the RDMS that utilizes the teachings of the present invention. As shown, these components include the four component layers of RDMS 200 (INTEREL software) discussed above. During normal operation, the different software components of RDMS 200 including the components of the present invention are loaded from disk storage 21-n into memory (e.g. MU 24-1) in a conventional manner.
In greater detail, SQL Director Component layer 202 operatively couples to an SQL adapter 201 that serves as the application's interface to the RDMS 200. The SQL Adapter 201 includes a runtime library that contains runtime routines bound into the application used by an application such as a COBOL-85 program for issuing calls. Each such call results in library sending a query statement to the SQL Director component layer 202.
The SQL Director component layer 202 handles the interface processing between RDMS 200 and a calling program. Thus, it manages the database connection. Layer 202 contains routines which analyze each query statement for determining if the statement is of a type that accesses relational database files and thus is suitable for code generation and caching. Each process utilizes a "local cache" for such storage. The use of "local caches" is discussed in the above referenced INTEREL Reference manual. Additionally, reference may be made to the copending patent application of Donald P. Levine and David A. Egolf, entitled: A Method and Apparatus for Improving the Performance of a Database Management System Through a Central Cache Mechanism, bearing Ser. No. 08/999,248 filed on Dec. 29, 1997 which is assigned to the same assignee as named herein.
As indicated, the SQL Director component layer 202 operatively couples to the RAM Codegen Executor Component layer 204. The SQL Director component layer 202 also contains routines that generate calls to a cache manager component to see if the code for that statement can be found in the process's local cache. When code for that statement is not found in the local cache, the SQL Director component layer 202 calls the RAM Codegen Executor Component layer 204 to process and "potentially" execute the query statement.
The RAM Codegen Executor layer 204 processes the SQL query. If the code has been generated for a particular query, layer 204 executes such code. When code has not been generated for a particular query, layer 204 optimizes the SQL query, generates code according to the optimized access plan and processes the requested data. The generated output code for the particular query is stored in "local cache" and executed.
As shown, the RAM Codegen Executor Component layer 204 operatively couples to the Relational File Manager component layer 206. During execution, the generated code calls various RFM functions to read or write relational data and/or index information from RFM files. Hence, this layer does not deal with the physical storage of data in a file.
The RFM component layer 206 performs the relational processing for RDMS 200. It receives the read and write requests from layer 204 and then translates them into IO file read and write requests respectively. It processes the file pages (CIs) read by layer 208 to which it operatively couples. Thus, this layer hides the physical storage of data and all other file format details from layer 204.
The IO Random Controller component layer 208 receives the requests from layer 206 and performs the relational file processing of translating the code-generated requests into I/O read/write requests. It processes the database files in page increments (CI). It is oblivious to physical storage of the data on the page. These details are handled by the other components illustrated in FIG. 2a. That is, layer 208 operatively couples to Buffer pools 212 via an Integrity Control Buffer Manager component 210. As known in the art, buffer pools contain buffers having a specific page size (control interval (CI). These buffers are used by the RDMS files. This arrangement is discussed in the above-mentioned INTEREL reference manual.
RAM Codegen Executor Layer 204--FIG. 3a
This figure illustrates in greater detail, specific components utilized according to the teachings of the present invention. As indicated, the layer 204 includes a common RAM Optimizer component 204-2, a code generation component 204-4, an SQL cache memory component 204-6 for storing SQL cache routines and generated code and subroutine library component 204-8. These components are operatively coupled as shown.
As discussed above, optimizer component 204-2 processes the SQL query by determining the appropriate access plan strategy. As a result of such processing, component 204-2 generates a set of EDOT structures that define the operation (SQL query) to execute (e.g. SELECT, UPDATE, INSERT or DELETE), the data to process (e.g. columns), the access method to use (e.g. scan or index or hash) and the restrictions that apply versus the access method to limit the amount of data (or rows) to process. Also, the structures define where the data obtained for the query is to be returned to a user. The path to ODI is used only during EDOT generation for verifying the query for correctness (e.g. does a specified column belong to a specified table). As indicated, the EDOT structures are applied as inputs to Code generation component 204-4. This component generates the required code that is stored in cache storage 204-6.
The cache storage 204-6 operatively couples to a subroutine library 204-8. Library 204-8 contains subroutines for communicating with RFM component layer 206 that operatively couples to buffer manager 210. In accordance with the teachings of the present invention, library 204-8 also includes an index processing enhancing subroutine that allows the bypassing of the RFM component layer 206 and IO component layer 208 as discussed herein. In this case, library 204-8 operatively couples to IO component layer 208 as indicated in FIG. 3a.
As indicated in FIG. 3a, the RFM component layer 206 includes several different types of record and index functions. The index functions include a Search Index function and a Search Next Index function. The Search Index function is used to locate the database key (DBK) of a record from an index key value provided by a caller in response to a user's search request. When the key value in the index satisfying the request contains pointers to multiple records (i.e., a duplicated key), the low DBK in the duplicate list will be returned along with a duplicate count. The caller may request one of the following three types of searches against the provided key value: equal, equal/greater than, or equal/less than. Optionally, this function will call an RFM record function (i.e., RFM retrieve record function) on the caller's behalf to return the record containing the index key value.
The Search Next Index function is used to return next key information to the calling program based upon a currency that has been previously established for the index against which this call is issued. When the current index key value is duplicated, DBKs will be returned from the duplicate list until the list is exhausted. Initial index currency can be established by the search index, store key, delete key and modify key functions. Optionally, this function also will call the RFM retrieve record function on the caller's behalf to return the record associated with the index key value.
Code Generation Component 204-4--FIG. 3b
FIG. 3b illustrates in greater detail, the structure of a portion of component 204-4 according to the present invention. More specifically, FIG. 3b depicts the routines that generate the code to perform index or data file retrievals. These routines include a number of standard routines that correspond to blocks 204-40 through 204-49 and 204-53. Additionally, the routines gen_retr_method function 204-50 and gen_index leaf have been extended. The subroutine gen_retr_method 204-50 generates the code to call the high speed row retrieval subroutine for enhancing data row retrieval performance. This subroutine is described in greater detail, in the second referenced copending patent application. The subroutine 204-50 incorporates into the code, any information that is necessary for carrying out the functions of the bypassed RFM manager component 206.
According to the teachings of the present invention, the subroutine gen_index_leaf 204-54 produces output code that calls the index performance enhancing routine for executing indexing operations under conditions defined by such routine. Appendix IV illustrates this structure in greater detail in generating code for a cursor open (SELECT) command.
Description of Operation
With reference to FIGS. 1 through 3b, the operation of the preferred embodiment of the present invention will now be described with reference to FIGS. 3c, 4a and 4b.
FIG. 3c
FIG. 3c illustrates conceptually, the layered organization of the present invention and more particularly, the organization of the code generation component layer 204 when having processed a particular SQL query. For example, FIG. 3cillustrates the case where layer 204 includes the index performance enhancing subroutine corresponding to subroutine IP prestored in subroutine library 204-8 for implementing certain types of index functions. Also, as indicated in FIG. 3c, library 204-8 further includes the normal RFM interface subroutine for carrying out index functions in a conventional manner by use of the lower RFM component layer 206 routines/procedures stored in RFM library 206-8 and IO component layer 208.
During the processing of an SQL query, the code generation component layer 204 operates to generate output code that includes specific call functions to the index performance enhancing subroutine IP or to the standard RFM interface subroutine. This is indicated in FIG. 3c by the pairs of dotted lines labeled "bypass" and "no bypass" between the output code block and subroutine library 204-8. When a specific operation within a query statement is determined to utilize the index performance enhancing subroutine stored in library 204-8, code generation component layer 204 includes in the output code, a call that references that performance enhancing subroutine thereby bypassing one or more lower component layers.
As indicated in FIG. 3c, the generated output code contains reference calls to subroutine IP for those parts of the SQL query that involves index searching. As discussed above, index searches can occur when processing SELECT, UPDATE or DELETE SQL statements.
FIG. 4a illustrates the prior art manner in which index searches are performed. As indicated, the RFM component layer 206 performs a number of tasks when retrieving an index entry. More specifically, first, the RFM component 206 in the case of an index search or first fetch operation processes a B tree index to locate the fine level index page (CI). Such B tree indexes are created with the RFM component via the standard file open function with index attributes defined. The index fine level entries on such page are in sorted order as discussed herein.
As well known in the art, B tree structures are used by DBMS to allow users to perform queries on large databases or tables using appropriate commands, such as SQL commands. A B tree is an index residing in the database memory having one or more columns, with each column representing another dimension in the index. B trees permit searching for records in a database using one or more keys specified by users via appropriate query commands. Because the keys define a subset of an entire table or records, indexed searching can eliminate the need to search through the entire table of records in order to retrieve a much smaller subset of such records pertinent to the user query.
The present invention takes advantage of the physical layout of the fine level index entries within a fine level page (CI). The index entries are lightly packed, adjacent to one another in sorted order as indicated. Thus, if one knows the offset to a particular fine level index entry (i.e., the reference entry), its adjacent entry is found by incrementing the pointer to the reference entry by the sum of the size of the reference entry plus the DBKEY size. Therefore, when RFM component layer 206 is called to execute the First Search Next operation, it returns a pointer to the reference entry to the enhancing index subroutine I and the subroutine's index logic uses that pointer to "walk through" the fine level index CI taking entries right out from under the control of the RFM component layer 206 as the Search Next requests are made.
During operation, the RFM component layer 206 initializes the currency structure to indicate which index entry the currency processing has started with. If the RFM component layer 206 is performing a Search Next operation, it finds the currency structure in the RFM schema structure. The RFM schema structure stores information about all of the aspects of the database tables to be accessed that must remain unchanged for the execution of the execution plan strategy as well as table and file identification (i.e., names of the tables and files) that can change without making the plan strategy inoperable.
In the case of the Search Next operation, RFM component 206 checks the currency structure's status to determine whether the currency has been invalidated by an UPDATE, INSERT or DELETE operation and if it has, then it returns an error to the requestor. When valid, the RFM component 206 uses the currency structure to determine the index key value to return. It uses this information to identify the page (CI) to make a request to the IO component layer 208. When all the index keys on the fine level CI have been processed by the RFM component 206, it processes the B tree index to determine what CI holds the next fine level key page. If the index entry specifies that it is one of many (duplicate keys), then the duplicate key processing functions of the RFM component layer 206 are invoked.
After the index entry has been chosen, the RFM component 206 updates the currency structure to indicate that the next entry was taken. Finally, the RFM component 206 copies the index entry into the user's buffer and returns.
FIG. 4b--Invention Index Processing Flow
FIG. 4b illustrates the enhanced index processing and more specifically, the manner in which index scan operations are executed according to the teachings of the present invention. As indicated in FIG. 4b, the standard index processing is performed by the code generation layer 204 in conjunction with RFM layer 206 in the manner described.
As indicated in FIG. 4b, the search process is altered to improve index access Search Next performance. The output code calls the index enhancing performance subroutine IP to determine if more than two next index accesses have been processed. On the second Search Next request, the subroutine IP examines the fine level index CI from which the prior index entry was retrieved. If the currency has not changed and the index CI has not changed and if the currency points to a fine level index entry that does not have duplicates, then the subroutine IP copies the requestor's key buffer along with the database key (DBKEY). The subroutine IP updates the currency information to point to the next fine level index entry. Finally, the subroutine IP returns to the caller.
In greater detail, the key aspect of the present invention is for the code generation component layer 204 to have the RFM component layer 206 fetch the first entry as normal so that the fine level index CI is read into the buffer pool and all the currency information will be established. On the first Search Next call, RFM component layer 206 will return a pointer to the index CI and the currency control structure. For the subsequent Search Next calls, the code generation component layer 204 will use the returned pointer and the currency information to obtain the next key.
As indicated in FIG. 4b, if the code generation component layer 204 processes a result following an index access, it calls the IO component layer 208 to refresh the index fine level CI pointer. From the DBKEY stored in the currency control information (idx_tb1 structure), code generation component layer 204 determines the CI that is required. The code generation component layer 204 will update the currency information each time it retrieves from the CI, a key that is not a duplicate. The code generation component layer 204 always tests the currency status. When the currency is obliterated or the CI changes, code generation component layer 204 calls the RFM component layer 206 to handle the request and setup the error status.
When the end of a fine level CI is met, the code generation component layer 204 calls the RFM component layer 206 to do the next Search Next operation so that RFM component layer 206 handles the B tree search for the next CI. When a duplicate is found, the code generation component layer 204 calls the RFM component layer 206 to do the Search Next operation and handle the duplicate processing.
For a more complete understanding of the benefits of the present invention, the following example of a query and its DDL will now be considered:
select firstName, ssNumber from student where lastName=`Jones` and middleinitial=`A`; (and this select is associated with a CURSOR, hence many results may be obtained)
create table student
(lastName char(20), firstName char(20), middleinitial char(1), ssNumber numeric(12));
create index ixStudent on student (lastName, firstName, middleinitial;
And assume that the table includes the following data:
lastName firstName middleInitial ssNumber
James Richard H 512-66-1234
Jones Anne W 512-55-1324
Jones Don H 512-44-1432
Jones Joseph A 512-33-2134
Jones William E 512-22-5566
Justice Don T 512-114144
The index would be a subset of this data, in sorted order (the six rows in this example are sorted).
The processing of the above data involves the following steps:
The RAM Optimizer 204-2 of FIG. 3a processes the query at execution time. The Optimizer 204-2 determines that index access via index ixStudent will provide the most efficient access to the data and generates the set of EDOT structures to define the access method, the operation and the columns to be returned. The type of structures generated are illustrated in an appendix.
Next, code generation component layer 204 is called to generate output code to process the query in accordance with the set of EDOT structures specified by the RAM Optimizer 204-2. Its tasks include the generation of output code to call its subroutine library functions RFM_GET_INDEXE and RFM_IO8_GETNEXT illustrated in FIG. 3c as well as generating output code to evaluate the index entries returned by these functions. The type of output code generated is illustrated in an appendix.
Next, the generated output code is executed. The output code begins index access with a call to its subroutine library function RFM_GET_INDEXE to find the first index entry beginning with `Jones`. This function calls RFM component layer 206 that will search the B tree index until `Jones` is found. This could involve repeated IO component layer calls as the index tree is analyzed to support the search for the fine level index CI that contains the first instance of `Jones`.
Once the fine level index entry for `Jones` is found, the RFM component layer 206 returns the index to the generated output code's key buffer for further analysis by the output code. Next, RFM component layer 206 returns to the subroutine library 204-8 function that will return to the generated output code. The generated output code next verifies that `Jones` was returned and then tests for `A` in the middle initial column of the index entry. Because the first instance of `Jones` does not have `A` as a middle initial, the search continues (assuming that another `Jones` exists).
Next, the generated output code calls its subroutine function RFM_IO8_GETNEXT to retrieve the next index entry for `Jones`. Since the subroutine function RFM_IO8_GETNEXT does not have a pointer to the RFM currency information that is required to "pickpocket" the fine level index CI, it calls the RFM's Search Next function. This function retrieves its currency information and calls the IO component layer 208 for a pointer to the fine level index CI. If the requested CI is in a buffer, IO component layer 208 returns the pointer to the RFM component layer 206. If it is not in a buffer, then IO component layer 208 calls buffer manager 210 of FIG. 2.
The IO component layer 208 returns with a pointer to the fine level index CI that has been read into the buffer pool. RFM component layer 206 finds the next entry and sets up its currency information. It returns to the subroutine function RFM_IO8_GETNEXT with a pointer to the CI and a pointer to its currency information. The function RFM_IO8_GETNEXT then returns to the generated output code that analyzes the returned index key value. Because the second instance of `Jones` does not have `A` as a middle initial, the search continues. The generated output code again calls the subroutine function RFM_IO8_GETNEXT to retrieve the next index entry for `Jones`. Subroutine RFM_IO8_GETNEXT notes that it has all the necessary pointers to find the next fine level index entry in the index CI. Because the prior search did not produce a result, the generated output code did not set the result processed indicator. Therefore, the function RFM_IO8_GETNEXT uses its CI pointer and the RFM currency information to copy the next fine level index entry from the buffer pool to the generated output code key buffer.
The RFM_IO8_GETNEXT subroutine function increments the RFM currency information, setting it to identify the next index entry to be processed. Then, the RFM_IO8_GETNEXT subroutine returns to the generated output code which analyzes the returned index key value. Because the second instance of `Jones` does have `A` as a middle initial, the output code performs a data file fetch sequence to retrieve the index's associated data file using the DBKEY associated with the index as its row identifier. Because this is an acquisition of a result, the output code sets the result processed indicator which means that the next index search will require a call to the IO component layer 208 at a minimum to refresh its pointer to the fine level index CI.
After the result is returned to the caller, the caller might request another FETCH from the open cursor and hence call to see if a second `Jones` entry with middle initial `A` exists. If so, the output code calls the subroutine RFM_IO8_GETNEXT. The RFM_IO8_GETNEXT subroutine will note that all the information required to "pickpocket" exists but the results processed indicator is set. Therefore, the subroutine will find the DBKEY for the last fine level index CI fetched from the currency information structure and from this calculate the CI to request from the IO component layer 208. The IO component layer 208 is called to locate the CI. Once identified, the IO component layer 208 will return with a pointer to the CI.
The RFM_IO8_GETNEXT subroutine uses the CI pointer and the RFM currency information to retrieve the next fine level index entry and copy it to the generated output code key buffer. The RFM_IO8_GETNEXT subroutine then increments the RFM currency information, setting it to identify the next index entry to be processed. The RFM_IO8_GETNEXT subroutine then returns to the generated output code which analyzes the returned index key value. Because the fourth instance of `Jones` does not have an `A` middle initial, the search continues.
The generated output code again calls the RFM_IO8_GETNEXT subroutine to retrieve the next index entry for `Jones`. The subroutine function RFM.sub.--IO8 _GETNEXT notes that it has all the necessary pointers to find the next fine level index entry in the index CI. Because the prior search did not produce a result, the generated output code does not set the result processed indicator. Therefore, subroutine RFM_IO8_GETNEXT uses its CI pointer and the REM currency information to retrieve the next fine level index entry and copy it to the generated output code key buffer. Again, the RFM_IO8_GETNEXT subroutine increments the RFM currency information, setting it to identify the next index entry to be processed.
The RFM_IO8_GETNEXT subroutine then returns to the generated output code which analyzes the return index key value. Because the lastName returned is greater than `Jones`, the generated output code returns the end of set status and processing is complete.
The above example illustrates how the present invention bypasses the RFM and IO component layers resulting in a significant reduction in index search time. The Appendices illustrate examples of output code utilized in the preferred embodiment of the present invention. For example, snippets of the specific code used in generating the call to the index enhancing subroutine are included in the appendices.
While the present invention was described relative to processing SQL statements, it will be obvious to those skilled in the art that the present invention may be used in conjunction with any database oriented systems that seeks to make significant performance gains. This could include all database types including relational database systems as well as "home brew" flat file databases that are used by many application developers. Also, while the preferred embodiment of the present invention operates to invoke the enhanced index processing of the present invention on the second search next index operation, it will be appreciated that as a matter of design choice such processing could occur on the first search next index operation. In the preferred embodiment, it was found that it was simpler to modify the RFM search next logic to capture the required pointers rather than to modify the RFM search index logic to capture the required pointers.
Appendicies
I. Glossary
II. Example EDOT output
III. Structure Chart for Code Generation Component Layer Functionality that Generates Code for a Cursor Open (SELECT) Operation
IV. Example of Generated Output Code to Call the Index Processing Subroutine I
V. Code Snippets utilized by the Code Generation Component for Generating the Call to Retrieve the Next Index Entry
VI. Significant Code from the rfm_io8_getNext Subroutine
APPENDIX I
Glossary
.cndot. access method The way in which the RDBMS
accesses relational
data, using hash, indexed, or
scan access.
* access plan- The RDBMS Optimizer's plan for
executing an SQL
statement, including the type
of access to each table,
order of access, whether any
sorts or joins are
performed, and related
information.
.cndot. administrator- An individual who carries out
tasks such as creating
databases and/or monitoring the
use and performance
of those databases.
.cndot. application- A program or set of programs
that performs a
specific task, such as entering
orders and updating
inventory records.
.cndot. application programmer (application developer)- An individual who
participates in the design,
implementation, and/or
testing of applications,
including end-user facilities.
.cndot. attribute- A descriptive characteristic of
an entity, such as
name, order number, item
quantity, and so on. A
characteristic such as UNIQUE
or NOT NULL that
can be applied to a column or
other entity.
.cndot. B Tree An index residing in the
database memory and
database memory and having one
or more columns,
with each column representing
another dimension in
the index.
.cndot. Buffer Manager- The software that manages
buffer pools.
.cndot. buffer pool- Main memory that is reserved
for the buffering
requirements of one or more
tablespaces, tables, or
indexes.
.cndot. Central SQL Cache- Configured memory space in
which code segments
generated by the RDBMS are
stored for reuse if the
same SQL statement is executed
again by any
process that has been
associated with this same
Central SQL Cache. The use of
one or more Central
SQL Caches is optional and
under Administrator
control.
.cndot. CHARACTER data type- An RDBMS data type defined as
fixed-length
character.
.cndot. code generation- The RDBMS process that
generates executable code
for a specific SQL statement
instead of executing
that statement interpretively.
.cndot. code segment- The code generated by the RDBMS
for a specific
SQL statement. Each code
segment is stored in the
Cache for reuse if the same SQL
statement is
executed again by the same
process in the case of
Process-Local SQL Cache, or by
another process in
the case of Central Cache.
.cndot. code_token (node) An item that uniquely refers to
a segment of
generated code.
.cndot. column- The vertical component of a
table. A column
contains information about one
of the attributes of an
entity. The relational
equivalent of a field.
.cndot. concurrency control- The management of file accesses
by processes
operating concurrently, with
the goal of ensuring that
no process interferes with any
other process and that
the integrity of the data
accessed in common is
maintained.
.cndot. concurrency level- The number of processes that
concurrently access the
same data.
.cndot. concurrent access- Two or more processes accessing
the same data table
or partition at the same time.
.cndot. Control Interval (CI)- The physical space increment
used to store data in
RDBMS files, processed as a
unit. Synonymous
with page. In a hash table, a
CI is called a hash
bucket.
.cndot. currency An index currency is the
information required to
uniquely identify a location
within a B Tree index.
The information stored in an
index currency includes
the CI and key entry number
within the CI for each
level of the B Tree. If the key
entry is a duplicate, the
index currency also stores the
duplicate entry
number. This information allows
RFM to retrace the
path down to the particular key
entry that is current.
This also allows RFM to advance
to the next logical
key entry in the B Tree index.
.cndot. data type- A definition of how the data in
a column is stored
and managed. Commonly used data
types include
CHARACTER, NUMERIC, and FLOAT.
.cndot. database- A collection of data that has
meaning to an
organization or to an
individual and that is managed
as a unit.
.cndot. database key (DBK) A database key is a key that
consists of the CI
number and line array for a
row. When a row is
inserted into a table in an
INTEREL/RFM database,
the RFM component assigns a
DBKEY to the row's
index entry. This DBKEY is
appended to the index
entry so that when the entry is
found later by a
search, the DBKEY is then used
to locate the data
row.
.cndot. Database Management System (DBMS)- A set of software that controls
the creation, organization, and
modification of one or
more databases, and provides
access to the stored
data.
.cndot. Data Definition Language (DDL)- Statements used to define the
entities,
attributes, and relationships
that make up a database.
DDL statements in RDBMS include
CREATE
MODEL, CREATE TABLE, CREATE
INDEX, and
others.
.cndot. Data Manipulation Language (DML)- Statements used to access data
and
potentially to change data
content. DML statements
in RDBMS include SELECT and
UPDATE.
.cndot. DATE data type- A special-purpose RDBMS data
type used to store a
four-byte date that can be used
in comparisons or
computations. The date is
formatted in packed
decimal in the row in storage,
and expanded when
read into application memory.
.cndot. DECIMAL data type- An RDBMS data type defined as
fixed-length with
decimal content.
.cndot. EDOT A set of structures resulting
from the processing of a
query by an optimizer. The term
EDOT pertains to
the identification (name) of
the file code where these
structures were once stored
(i.e. the E.file code file).
.cndot. fine level index Equivalent to leaf page. The
leaf level page(s) of a B
Tree index. The level where the
index entries reside.
.cndot. FROM clause- The part of an SQL statement
that specifies the
table(s) from which data is to
be accessed.
.cndot. generated code- A reusable set of machine
instructions that are
produced to execute a
particular SQL statement.
.cndot. gen_time (node) The approximate time that was
required to generate
the associated segment of code.
.cndot. host variable- Storage allocated by a language
processor such as
COBOL-85 for use in passing
variable information to
or from an SQL statement.
.cndot. host-relational files- Files managed by RAM/RFM and
maintained
mass-storage devices attached
to a system. The term
is used to distinguish these
files from relational files
maintained by a Relational Data
Base Computer
linked to a system.
.cndot. IO_XPT This structure is the parameter
passed on every call
to the IO Component Layer. When
performing a
SELECT or FETCH, the file
identifier and CI
number are passed in this
structure. The results of IO
Component Layer operations are
returned in this
structure.
.cndot. index- A set of pointers to data in
relational tables, used to
improve the speed of access to
the data. The index is
maintained in the form of a
balanced B-tree structure.
.cndot. index key- The column(s) whose values are
included in the
index for the table.
.cndot. indexed access- The access mode in which RDBMS
uses one or more
indexes to retrieve the
requested information. In
some cases, RDBMS can retrieve
the requested
information directly from the
index, without
accessing the data table.
.cndot. indicator variable- An indicator used to notify the
application that an
exception condition, such as an
overflow or
truncation error, occurred.
When used to indicate the
presence of the null value, the
indicator is referred to
as a null indicator variable.
.cndot. indicator column- A technique in denormalization
in which an indicator
column in one table indicates
whether an access to a
second table is required. Using
an indicator column
introduces the risk that data
anomalies will occur.
.cndot. INTEGER data type- An RDBMS data type used to
store binary values.
.cndot. integrity- The property of being logically
correct and
consistent. In some cases, this
term also implies
denying access to unauthorized
users and processes.
.cndot. null value- A value designed to be used
when the actual value is
unknown or inappropriate. For
example, using the
null value for a price that has
not yet been defined
avoids the ambiguity of using a
zero value in this
case.
.cndot. null indicator variable- An indicator used to notify the
application that a
column contains the null value.
.cndot. NUMERIC data type- An RDBMS data type defined as
fixed-length and
limited to the numerals 0-9.
.cndot. Operational Directory Interface (ODI)- A data definition directory
used
by RDBMS.
.cndot. Optimizer- The software component of RDBMS
that analyzes
each SQL statement and defines
the access plan that
will produce the optimum
performance for the
execution of that statement.
.cndot. page- The physical space increment
used to store data in
RDBMS files, processed as a
unit. Synonymous
with control interval (CI).
.cndot. partitioned row A row that was originally
inserted into a database
page and was subsequently
updated so that the row
exists in more than a single
page or CI.
.cndot. partitioned tablespace- A tablespace that contains a
single table that is too
large to process efficiently as
one entity. The
tablespace and the table are
separated into partitions
that can be placed on different
mass storage devices.
Each partition can be processed
independently.
.cndot. performance- The ability of applications and
interactive facilities to
meet users' requirements for
speed of response in
interactive applications or
speed of throughput in
production batch and/or
transaction processing
applications.
.cndot. primary index (key)- In some implementations, but
not in RDBMS, an
index (and associated key) that
uniquely identifies
each row of a table.
.cndot. process- An execution unit, which may or
may not be
identical to an application
program. (An application
program may involve multiple
processes.)
.cndot. Process-Local SQL Cache- The configured memory space in
which code
segments generated by RDBMS are
stored for reuse
if the same SQL statement is
executed again by the
same process.
.cndot. query- Used to refer to any SQL
statement that causes an
access to relational tables,
regardless of whether the
access is followed by a
modification.
.cndot. read-ahead feature- The Buffer Manager's
operational mode in which up
to 16 pages are read into the
buffer pool at one time,
so that the data will be
available when needed. The
read-ahead feature is used in
scan access mode.
.cndot. recovery- The process of restoring
database(s) and/or
application(s) to a known state
after an exception
condition or a process or
system failure.
.cndot. Relational Access Manager (RAM)- The part of the RDBMS software
that
manages relational tables.
.cndot. Relational database- A database that is defined in
terms of entities,
attributes, and relationships,
and that observes the
concepts defined originally by
E.F. Codd.
.cndot. Relational Database Management System (RDBMS)- The INTEREL
Software that manages
relational databases.
.cndot. Relational File Manager (RFM)- The part of the RDBMS software
that
accesses data from relational
files, as directed by the
Relational Access Manager
(RAM).
.cndot. RFM_XPT The RFM_XPT structure is the
parameter that must
be passed on all calls to RFM
functions. It facilitates
the passing of variable
information to the RFM
component layer regarding the
operation to be
performed in a subset of the
structure named the V
block. It also facilitates the
return of information
such as the DBKEY and status in
another subset of
the structure named the R
block. When generated
code's subroutines call the RFM
component layer to
retrieve a row, the subroutines
pass the DBKEY, file
identifier and description of
what columns to fetch
and where to store them (that
is, a pointer to the
generated code's data buffer)
in the RFM_XPT
structure. The subroutines also
examine the
RFM_XPT's R block after the
call to determine the
status of the request.
.cndot. relationship- An association of one or more
entity types with one
or more other entity types.
.cndot. row- The horizontal component of a
table. A row consists
of a sequence of values, one
for each column of the
table.
.cndot. scan access- The access mode in which RDBMS
scans a table
sequentially, row-by-row, to
retrieve the requested
information.
.cndot. schema structure It is used by the RFM and is
the root of the RFM
logical definition of a related
group of files, all of the
information that the RFM uses
to access a physical
file is in some way related in
this structure.
.cndot. search condition- The specification of how to
identify the desired data
in the rows and columns
retrieved when processing
an SQL statement.
.cndot. secondary index- In some RDMBS implementations,
any index that is
not a primary index. In RDBMS,
most indexes are
treated as secondary indexes;
cluster indexes and
system-generated indexes
created because a column
is defined with the UNIQUE
attribute are the
exceptions to this rule. RDBMS
enforces the
uniqueness of the columns
included in these indexes.
.cndot. SELECT- An SQL statement that defines
the information to be
selected from one or more
tables. Also, a clause that
includes the SELECT verb in an
SQL statement.
.cndot. single-column index (key)- An index and key made up of
only one column.
.cndot. single-table tablespace- An unpartitioned tablespace
that contains one table.
The tablespace is a single
physical file. Contrast this
with multi-table tablespace and
with partitioned
tablespace.
.cndot. SMALLINT data type- An RDBMS data type used to
store binary values,
using less space than is
required for the INTEGER
data type.
.cndot. SQL- Originally an acronym for
Structured Query
Language. Now the name of the
language most
commonly used to access
relational databases.
.cndot. SQL Cache- Configured memory space in
which code segments
generated by RDBMS are stored
for reuse if the same
SQL statement is executed
again. There are two
levels of SQL Cache,
Process-Local SQL Cache and
Central SQL Cache.
.cndot. SQLCA The SQL Communications Area
(SQLCA) is a fixed-
length data structure which
allows the application
program to provide the specific
database options and
information to the database
content manager. It also
allows the application program
to interrogate the
execution results of an
embedded SQL statement.
.cndot. SQLDA The SQL Descriptor Area (SQLDA)
is a variable
length data structure what when
used in an OPEN or
FETCH statement, it provides
the information about
the host variables to INTEREL
software.
.cndot. table- The relational component that
corresponds to a file in
non-relational technology.
Tables have two forms:
hash and non-hash.
.cndot. tablespace- Physical space that is created
explicitly to contain
one or more tables in a
relational database. If a
tablespace is not created
explicitly, the table is stored
in space allocated by the
system. Space allocated by
the system is not called a
tablespace.
.cndot. TIME data type- A special-purpose RDBMS data
type used to store a
three-byte time that can be
used in comparisons or
computations. The time is
formatted as packed
decimal in the row in storage,
and expanded when
read into application memory.
.cndot. TIMESTAMP data type- A special-purpose RDBMS data
type used to store a
10-byte time stamp that can be
used in comparisons
or computations. The timestamp
is formatted as
packed decimal in the row in
storage, and expanded
when read into application
memory.
.cndot. Transaction Processing Routine (TPR)- An application program that
processes transactions under
control of a transaction
monitor. In the context of this
publication, the
transaction monitor is TP8.
.cndot. tuning- The process of adjusting
database definitions,
application implementations,
and/or system
parameters to improve
performance.
.cndot. UNIQUE key- An index key or a hash key made
up of one or more
columns in which no duplicate
values are allowed.
.cndot. unpartitioned tablespace- A tablespace that contains
either a single table or
multiple tables. The latter
case is called a multi-table
tablespace. In both cases, the
tablespace is a single
physical file. Contrast this
with partitioned
tablespace.
.cndot. user- An individual who accesses
RDBMS databases by
means of an application or
interactive facility.
.cndot. VARCHAR data type- An RDBMS data type defined as
variable-length
character.
.cndot. WHERE clause- The part of an SQL statement
that defines the
conditions (restrictions) for
selecting data before
grouping (if grouping is
specified). These
restrictions are referred to as
search conditions.
APPENDIX II
1. Optimizer Output-EDOT Example:
For the query:
select firstName, ssNumber from student
where lastName = `Jones` and middleInitial =
`A`;
the RAM Optimizer produces the following EDOT:
##STR1##
TC <- - - - - - - - - - - - - - - - - - - - - - - - - 145322006112
TYPE = 1
NON_AGGR_PRESENT = YES
DUPLICATE_GROUP_ITEMS = 0
SELECT$ = 145576006112
EC = 145374006112
METHOD = 150070006112
RESTRICT = 147752006112
AGGR_COUNT = 0
GROUP_SIZE = 0
LEAF_COUNT = 3
LEAF_BASE = 145714006112
VR_SIZE = 8912896
VIRTUAL_REC = 000000006000
PARAMETER = 000000006000
*** BEGIN TC.SELECT$ ***
STRUCT <- - - - - - - - - - - - - - - - - - - - - 145576006112
Offset from TC = 000000000254
TYPE = 3
MAX_SIZE = 0
COUNT = 2
SEL = 145610006112
SEL = 145420006112
SEL = 145535006112
TT <- - - - - - - - - - - - - - - - - - - - - - - - - 145610006112
Offset from TC = 000000000266
TYPE = 5
MODE = 000000000
TWA_DATA_TYPE = 000000
STAR_NULL = YES
TWA_EIS = 000000000000000000000000000000000000
RWA_EIS = 000000000000000000000000000000000000
START_ELEMENT = 000
END_ELEMENT = 000
LENGTH = 1
SCALE = 2
OBJECT = 000000006000
ECTT = 145432006112
EC = 145374006112
FR = 000000006000
AL = 000000006000
SELECTOR <- - - - - - - - - - - - - - - - - - - 145420006112
Offset from TC = 000000000076
TYPE = 4
AL_FOR_TT_LIST = 000000000000
OBJECT = 145432006112
LENGTH = 9
NAME = FIRSTNAME
SELECTOR <- - - - - - - - - - - - - - - - - - - 145536006112
Offset from TC = 000000000214
TYPE = 4
AL_FOR_TT_LIST = 000000000000
OBJECT = 145546006112
LENGTH = 8
NAME = SSNUMBER
*** BEGIN TC.EC ***
EC <- - - - - - - - - - - - - - - - - - - - - - - - - 145374006112
Offset from TC = 000000000052
TYPE = 2
SELECTED = YES
TT = 145774006112
TCEC = 000000006000
TC = 145322006112
RC = 535723006112
IA_PTR = 402702006112
SDOT_ID = 000000000001100001000000000100011001
000011000110000011000100011101101011
CALC_COUNT = 0
BUFF_SIZE = 0
BUFF = 402504006112
FILE_TYPE = 5
VOLUME = 100
DBKEY = 0
AREA_ID_OFFSET = 0
AL = 000000006000
TT <- - - - - - - - - - - - - - - - - - - - - - - - - 145774006112
Offset from TC = 000000000452
TYPE = 5
MODE = 001000000
TWA_DATA_TYPE = 000000
INDEX_ACC = YES
RESTRICT = YES
NULL_ALLOWED = YES
TWA_EIS = 011000000000000000000000000000000001
RWA_EIS = 001000000000000000000000000000000001
START_ELEMENT = 000
END_ELEMENT = 000
LENGTH = 1
SCALE = 0
OBJECT = 000000006000
ECTT = 145666006112
EC = 145374006112
FR = 536023006112
AL = 000000006000
TT <- - - - - - - - - - - - - - - - - - - - - - - - - 145666006112
Offset from TC = 000000000344
TYPE = 5
MODE = 001000000
TWA_DATA_TYPE = 000000
INDEX_ACC = YES
RESTRICT = YES
NULL_ALLOWED = YES
TWA_EIS = 011000000000000000000000000000010100
RWA_EIS = 001000000000000000000000000000010100
START_ELEMENT = 000
END_ELEMENT = 000
LENGTH = 20
SCALE = 0
OBJECT = 000000006000
ECTT = 145610006112
EC = 145374006112
FR = 535767006112
AL = 000000006000
TT <- - - - - - - - - - - - - - - - - - - - - - - - - 145610006112
Offset from TC = 000000000266
TYPE = 5
MODE = 000000000
TWA_DATA_TYPE = 000000
STAR_NULL = YES
TWA_EIS = 000000000000000000000000000000000000
RWA_EIS = 000000000000000000000000000000000000
START_ELEMENT = 000
END_ELEMENT = 000
LENGTH = 1
SCALE = 2
OBJECT = 000000006000
ECTT = 145432006112
EC = 145374006112
FR = 000000006000
AL = 000000006000
TT <- - - - - - - - - - - - - - - - - - - - - - - - - 145432006112
Offset from TC = 000000000110
TYPE = 5
MODE = 001000000
TWA_DATA_TYPE = 000000
INDEX_ACC = YES
NULL_ALLOWED = YES
TWA_EIS = 011000000000000000000000000000010100
RWA_EIS = 001000000000000000000000000000010100
START_ELEMENT = 000
END_ELEMENT = 000
LENGTH = 20
SCALE = 0
OBJECT = 000000006000
ECTT = 145546006112
EC = 145374006112
FR = 536005006112
AL = 000000006000
TT <- - - - - - - - - - - - - - - - - - - - - - - - - 145546006112
Offset from TC = 000000000224
TYPE = 5
MODE = 010100000
TWA_DATA_TYPE = 000101
NULL_ALLOWED = YES
TWA_EIS = 011000000000000000001010000000001101
RWA_EIS = 001000000000000000000010000000001101
START_ELEMENT = 000
END_ELEMENT = 000
LENGTH = 13
SCALE = 0
OBJECT = 000000006000
ECTT = 150174006112
EC = 145374006112
FR = 536041006112
AL = 000000006000
TT <- - - - - - - - - - - - - - - - - - - - - - - - - 150174006112
Offset from TC = 000000002652
TYPE = 5
MODE = 001000000
TWA_DATA_TYPE = 000000
INDEX_ACC = YES
MULTI_FIELD_KEY = YES
FOR_RFM_KEY_FIELD = YES
TWA_EIS = 001000000000000000000000000000101001
RWA_EIS = 001000000000000000000010000000101001
START_ELEMENT = 000
END_ELEMENT = 000
LENGTH = 41
SCALE = 0
OBJECT = 000000006000
ECTT = 000000006000
EC = 145374006112
FR = 536065006112
AL = 000000006000
FE <- - - - - - - - - - - - - - - - - - - - - - - - - 536057006112
TYPE = 6
FILE_TYPE = 1
DATA_RAP_NUM = 1249
INDX_RAP_NUM = 0
OCCUPIED_PAGES = 7
LOAD_MODE = YES
PCTFREE = 25
PCTGROW = 50
PCTTHRESHOLD = 20
CISIZE = 512
TE <- - - - - - - - - - - - - - - - - - - - - - - - - 535723006112
TYPE = 2
TOTAL_SIZE = 138
CREATE_USERID = DSEDWARDS
NUMBER = 583
DATA_RAP = 1249
FD_COUNT = 4
KE_COUNT = 1
ROW_SIZE = 54
KEY_SIZE = 41
PAGE_COUNT = 0
NAME = STUDENT
ATTACHED = YES
WRITE_ALLOC = YES
FILE_ALLOCATED = YES
RFM_OPENED = YES
HASH_ERA_TE = YES
CI_SIZE = 0
LINES_PER_CI = 32
LOAD_LIMIT = 0
SPACE_CTL = 0
SCHEMA_PTR = 000000006000
DATA_FC = A+
INDX_FC = B+
NEXT_TE = 000000006000
FD_PTR = 535767006112
KE_PTR = 536065006112
DATA_FE = 536057006112
INDX_FE = 536105006112
MEM_BLOCK_PTR = 000000006000
MODEL_PTR = 476734006112
RECORD_COUNT = 100
STATS_ENTRY_NO = -1
DATA_ROW_SIZE = 48
TS_PTR = 000000006000
CI_SIZE_RAMINDEX = 512
CI_SIZE_CLUSTER = 0
FE <- - - - - - - - - - - - - - - - - - - - - - - - - 536105006112
TYPE = 6
FILE_TYPE = 2
DATA_RAP_NUM = 1249
INDX_RAP_NUM = 1250
OCCUPIED_PAGES = 0
LOAD_MODE = YES
PCTFREE = 25
PCTGROW = 50
PCTTHRESHOLD = 20
CISIZE = 512
FD <- - - - - - - - - - - - - - - - - - - - - - - - - 535767006112
TYPE = 3
LENGTH = 20
NAME = LASTNAME
IN_USE_FOR_COMM = YES
PRESENT_IN_ROW = YES
DATA_TYPE = 3
NULL_TYPE = 0
OFFSET = 0
NUMBER = 1
NEXT_FD = 536005006112
KD_PTR = 536113006112
HASH_KD_PTR = 000000006000
FD <- - - - - - - - - - - - - - - - - - - - - - - - - 536005006112
TYPE = 3
LENGTH = 20
NAME = FIRSTNAME
IN_USE_FOR_COMM = YES
PRESENT_IN_ROW = YES
DATA_TYPE = 3
NULL_TYPE = 0
OFFSET = 20
NUMBER = 2
NEXT_FD = 536023006112
KD_PTR = 536123006112
HASH_KD_PTR = 000000006000
FD <- - - - - - - - - - - - - - - - - - - - - - - - - 536023006112
TYPE = 3
LENGTH = 1
NAME = MIDDLEINITIAL
IN_USE_FOR_COMM = YES
PRESENT_IN_ROW = YES
DATA_TYPE = 3
NULL_TYPE = 0
OFFSET = 40
NUMBER = 3
NEXT_FD = 536041006112
KD_PTR = 536133006112
HASH_KD_PTR = 000000006000
FD <- - - - - - - - - - - - - - - - - - - - - - - - - 536041006112
TYPE = 3
LENGTH = 7
NAME = SSNUMBER
IN_USE_FOR_COMM = YES
PRESENT_IN_ROW = YES
ODD_NUMERIC_LENGTH = YES
DATA_TYPE = 4
NULL_TYPE = 0
OFFSET = 41
NUMBER = 4
NEXT_FD = 000000006000
KD_PTR = 000000006000
HASH_KD_PTR = 000000006000
KE <- - - - - - - - - - - - - - - - - - - - - - - - - 536065006112
TYPE = 4
KD_COUNT = 3
NAME = IXSTUDENT
DUPES_ALLOWED = YES
ASCEND_SEQ = YES
INDX_NUM = 557
LENGTH = 41
NEXT_KE = 000000006000
INDX_FE = 536105006112
KD_PTR = 536113006112
KEY_COUNT = 0
INDEX_RANGE_PTR = 000000000000
KD <- - - - - - - - - - - - - - - - - - - - - - - - - 536113006112
TYPE = 5
KD_TYPE = 0
OFFSET = 0
FD_PTR = 535767006112
NEXT_KD = 536123006112
KE_PTR = 536065006112
FD_KD_NEXT = 000000006000
KD <- - - - - - - - - - - - - - - - - - - - - - - - - 536123006112
TYPE = 5
KD_TYPE = 0
OFFSET = 20
FD_PTR = 536005006112
NEXT_KD = 536133006112
KE_PTR = 536065006112
FD_KD_NEXT = 000000006000
KD <- - - - - - - - - - - - - - - - - - - - - - - - - 536133006112
TYPE = 5
KD_TYPE = 0
OFFSET = 40
FD_PTR = 536023006112
NEXT_KD = 000000006000
KE_PTR = 536065006112
FD_KD_NEXT = 000000006000
*** BEGIN TC.RESTRICT ***
NODE <- - - - - - - - - - - - - - - - - - - - - - - 147752006112
Offset from TC = 000000002430
TYPE = 9
OBJECT = 147762006112
OP = 19
LEFT = 146022006112
RIGHT = 145714006112
TREE LEAF = 146022006112
TREE LEAF = 145714006112
*** BEGIN TC.LEAF_BASE ***
LEAF <- - - - - - - - - - - - - - - - - - - - - - - 145714006112
Offset from TC = 000000000372
TYPE = 7
PRIME_ACCESS = YES
ACCESS = 1
INDX = 1
COUNT = 2
NEXT = 146022006112
ARG = 0
OPERATOR = 0
OBJECT = 145666006112
ARG = 0
OPERATOR = 145642006112
OBJECT =
ARG = 2
OPERATOR = 10
OBJECT = 145704006112
SCARD <- - - - - - - - - - - - - - - - - - - - - - 145642006112
Offset from TC = 000000000320
TYPE = 6
MODE = 000000010
LENGTH = 5
SCALE = 0
EIS = 000000000000000000000000000000000000
OBJECT = Jones
SCARD <- - - - - - - - - - - - - - - - - - - - - - 145704006112
Offset from TC = 000000000362
TYPE = 6
MODE = 000000010
LENGTH = 1
SCALE = 0
EIS = 000000000000000000000000000000000000
OBJECT =
LEAF <- - - - - - - - - - - - - - - - - - - - - - - 146022006112
Offset from TC = 000000000500
TYPE = 7
PRIME_ACCESS = YES
ACCESS = 1
INDX = 2
COUNT = 2
NEXT = 150146006112
ARG = 0
OPERATOR = 0
OBJECT = 145774006112
ARG = 1
OPERATOR = 0
OBJECT = 145746006112
ARG = 2
OPERATOR = 10
OBJECT = 146012006112
SCARD <- - - - - - - - - - - - - - - - - - - - - - 145746006112
Offset from TC = 000000000424
TYPE = 6
MODE = 001000000
LENGTH = 1
SCALE = 0
EIS = 000000000000000000000000000000000000
OBJECT = A
SCARD <- - - - - - - - - - - - - - - - - - - - - - 146012006112
Offset from TC = 000000000470
TYPE = 6
MODE = 000000010
LENGTH = 1
SCALE = 0
EIS = 000000000000000000000000000000000000
OBJECT =
LEAF <- - - - - - - - - - - - - - - - - - - - - - - 150146006112
Offset from TC = 000000002624
TYPE = 7
ACCESS = 0
INDX = 3
COUNT = 2
NEXT = 000000006000
ARG = 0
OPERATOR = 0
OBJECT = 150174006112
ARG = 1
OPERATOR = 0
OBJECT = 150240006112
ARG = 2
OPERATOR = 10
OBJECT = 150164006112
SCARD <- - - - - - - - - - - - - - - - - - - - - - 150240006112
Offset from TC = 000000002716
TYPE = 6
MODE = 000000010
LENGTH = 1
SCALE = 0
EIS = 000000000000000000000000000000000000
OBJECT = Jones
SCARD <- - - - - - - - - - - - - - - - - - - - - - 150164006112
Offset from TC = 000000002642
TYPE = 6
MODE = 000000010
LENGTH = 1
SCALE = 0
EIS = 000000000000000000000000000000000000
OBJECT =
METHOD <- - - - - - - - - - - - - - - - - - - - 150070006112
Offset from TC = 000000002546
TYPE = 10
REQUIRED = YES
CONJUNCTIVE_APPROACH = YES
SIMPLE_RESTRICT = YES
INDEX_ONLY_QUALIFY = YES
HOLD_SC = 0
SYSTAB_ROW = 0
IO_TYPE = 1
LEFT = 000000006000
RIGHT = 000000006000
EC = 145374006112
FATHER = 145322006112
CODE_PTR =
PAGE_EST =
ENTITY_EST =
CONTROL = 150146006112
PRIME = 146022006112
SECOND = 000000006000
CURRENT = 000000006000
PRIME_KEY_TABLE_PTR = 000000006000
RA_PTR = 000000006000
CODE_GEN_PTR = 000000006000
KEY_VALUE_PTR = 000000006000
TABLE_NAME_LIST_PTR = 000000006000
TABLE_NAME_LIST_SIZE = 0
KEY_TT_PTR = 150174006112
INDEX_PTR 536065006112
APPENDIX III
1. Structure Chart for Code Generation Component Layer Functionality
that Generates Code for Cursor Open (SELECT) Operation
##STR2##
##STR3##
##STR4##
note 1: If method.control points to a node, gen_index_node is called. Otherwise it must point to a leaf in which case gen_index_leaf is called.
note 2: The gen_rfm _getni routine is called when a range of values is being retrieved (e.g., WHERE lastName BETWEEN `Jones` AND `Lemke`) or when an equal restriction is applied against a non-unique index (e.g., WHERE lastName=`Jones`).
note 3: gen_rfm_getni generates a call to the subroutine library function rfm_io8_getNext if the bypassRFM configuration option is set on (the default state). Otherwise, if off, it generates a call to the routine that was used previously, rfm_getnext_index.
note 4: When afhgrfrn.c option bypassRFM is enabled (default) the new code that generates the call to the library subroutine rfm_io8_getrecord is made. (This does not show the "old" path).
APPENDIX IV
1. Example of Code Generated to Call the Index Retrieval Subroutine
* Initialize the RFM_XPT structure
000366 107122223003 LDX3 0107122, DU
000367 001506443000 SXL3 01506
000370 001542450000 STZ 01542
000371 001543450000 STZ 01543
000372 001544450000 STZ 01544
000373 001545450000 STZ 01545
000374 006000236007 LDQ 06000, DL
000375 001564756000 STQ 01564
000376 001565756000 STQ 01565
000377 001242236000 LDQ 01242 RFM Schema Pointer
000400 001466756000 STQ 01466
*************************************
* Get RFM index entry for record with EC at 502216006112
*************************************
000401 001572635400 EPPR5 01572
000402 000002630404 EPPR0 02, IC
000403 346130710100 TRA RFM_GET_INDEXE-BIAS, , P3
000404 343662710100 TRA RAM_RFM_ERROR-BIAS, , P3
000405 000000116007 CMPQ 00, DL
000406 000767604000 TMI 0767 No More Records
000407 001244756000 STQ 01244 Data Base Key
000410 000426710000 TRA 0426
000411 001432236000 LDQ 01432
000412 001252756000 STQ 01252
000413 001436236000 LDQ 01436
000414 001250756000 STQ 01250
000415 001434236000 LDQ 01434 IXNUM for Retrieval
000416 001254756000 STQ 01254
*************************************
* Get next RFM index entry for record with EC at 502216006112
*************************************
000417 001602635400 EPPR5 01602
000420 000002630404 EPPR0 02, IC
000421 346260710100 TRA RFM_IO8_GETNEXT-BIAS, , P3 (New
Subroutine Library function)
000422 343662710100 TRA RAM_RFM_ERROR-BIAS, , P3
000423 000000116007 CMPQ 00, DL
000424 000406604051 TMI 0406, I No More Records
000425 001244756000 STQ 01244 Data Base Key
* Compare saved key buffer to key returned by RFM
000426 000002227003 LDX7 02, DU
000427 000012336007 LCQ 012, DL
000430 000040106440 CMPC (,RL, ,), (,RL, ,)
000431 001306000017 ADSC9 01306, 0, x7
000432 001460000017 ADSC9 01460, 0, x7
000433 000411600000 TZE 0411 Get next Record
Appendix V
1. Code Snippets Utilized by Gen_index_leaf Function for Generating the Call to Retrieve the Next Index Entry
Snippet 1-- gen_index_leaf
The following code sets the transfer offset based on the state of the bypassRFM option then builds the call to the subroutine library.
if (bypassRFM)
calofs=ptr_to_long((void *) rfm_io8_getNext);
else
calofs=ptr_to_long((void *) rfm_getnext_indexe);
calofs &=0077777000000;
. . .
if((sts=putcode(calofs .vertline. 0300000710100, RELOC_NONE)) <0)
return (sts); /* TRA getix,,p3 */
APPENDIX VI
1. Significant Code from the frm_io8_getNext Subrouting I
code snippet 1 - index state tests.
if rfm_xpt.r.key_data_prt is NULL then Search Next has not been called.
ldp p7,1,,p5 p7 = rfm_xpt rfmiwn6m
ldq rfm_ky,,p7 if rfm_xpt.r.key_data_ptr rfmiwn6m
cmpq g8null.dl is null rfmiwn6m
tze rfmnby ..then skip.. rfmiwn6m
rfm8bb null rfmiwn6m
ldp p0,rfm_ix,,p7 if rfmiwn6m
szn idx_dp,,p0 last entry was a dup rfmiwn6m
tmi rfmnby ..then skip.. rfmiwn6m
ldq idx_cs,,p0 if the currency rfmiwn6m
cmpq ccancl,dl was cancelled rfmiwn6m
tze rfmnby ..then skip.. rfmiwn6m
rfm8bc null
ldq idx_f1,,p0 fetch the fine level index rfmiwn6m
qrl 18 level for idx_level index rfmiwn6m
qls 1 multiply by 2 for array index. rfmiwn6m
stq idxlvl-bias,,p4 save for later reference. rfmiwn6m
lxl x0,idx_ky,q1,p0 if rfmiwn6m
adlx x0,1,du all indexes in this ci rfmiwn6m
cmpx x0,idx_ky,q1,p0 have been processed, rfmiwn6m
trc rfmnby ..then skip (end of index or ci).. rfmiwn6m
rfm8bd null rfmiwn6m
als 0 if no io8 call mandated, rfmiwn6m
tze rfm8go ..then skip.. rfmiwn6m
code snippet 2 - IO8 Call to guarantee the fine level index CI
state.
rfm8be null rfmiwn6m
* rfmiwn6m
* s7918d call io8 to get another copy of the fine level index ci.
rfmiwn6m
* rfmiwn6m
lda idx_rp,,p0 a = index rap number rfmiwn6m
ana -1,dl a 18-35 = index rap number rfmiwn6m
ldq idx_dk,q1,p0 q = dbkey. rfmiwn6m
ldp p7,rfm_sp,,p7 p7 = schema pointer rfmiwn6m
ldp p1,filecb,,p7 p1 = io_xpt pointer rfmiwn6m
* rfmiwn6m
* set file info io_xpt rfmiwn6m
ldp p0,prtptr,,p7 p0 = partition pointer table ptr. rfmiwn6m
sbla lo_ptn,,p0 a = rap index (ref getdbk_set_ptrs) rfmiwn6m
als 1 a = index into partition ptr table. rfmiwn6m
ldp p6,ptn_ar,al,p0 p6 = rfm_file_control pointer. rfmiwn6m
* rfmiwn6m
* calc the ci number and line number (ref get_dbk) rfmiwn6m
* q = dbkey. rfmiwn6m
adlq dbk_ci,,p6 inc by adjustment constant. rfmiwn6m
div lpci,,p6 div by lines per ci. rfmiwn6m
stq ci_num,,p1 save ci number for fetch in io_xpt. rfmiwn6m
* rfmiwn6m
* set the filecode_id rfmiwn6m
lxl x7,fc_id,,p6 copy the rfmiwn6m
sxl x7,ifc_id,,p1 filecode id. rfmiwn6m
* rfmiwn6m
* call io8 via io_xpt.enter.read_ci passing io_xpt in the a reg. rfmiwn6m
rfm8cl null rfmiwn6m
lda filecb,,p7 a = ptr to io_xpt rfmiwn6m
sta io_ptr-bias,,p4 rfmiwn6m
ldp p1,readci,,p1 p1 = readci io8 function pointer rfmiwn6m
ldx x0,stfsiz,du x0 = stack size. rfmiwn6m
eppr0 *+3 p0 = return. rfmiwn6m
tra 0,,p1 call io8 rfmiwn6m
zero 1,0 passing 1 arg. rfmiwn6m
code snippet 3 - Test CI control words for change.
rfm8go null rfmiwn6m
* rfmiwn6m
lda rfm_ky,,p7 a = ptr to key. rfmiwn6m
ana =o7777,d1 strip all but segid (start of ci) rfmiwn6m
sta temp-bias,,p4 then save. rfmiwn6m
ldp p6,temp-bias,,p4 p6 = pointer to ci. rfmiwn6m
ldp p0,2,,p5 p0 = area to store ci header info. rfmiwn6m
rfm8b1 null rfmiwn6m
ldaq 0,,p6 if ci header word 0 and 1 changed rfmiwn6m
cmpaq 0,,p0 ..then skip.. rfmiwn6m
tnz rfmnby rfmiwn6m
ldaq 2,,p6 if ci header word 2 or 3 changed rfmiwn6m
cmpaq 2,,p0 ..then skip.. rfmiwn6m
tnz rfmnby rfmiwn6m
code snippet 4 - Duplicate entry test.
rfm8b2 null rfmiwn6m
ldp p0,rfm_ix,,p7 p0 = idx_tbl pointer rfmiwn6m
lxl x0,idx_k1,,p0 x0 = size of key (includes rap) rfmiwn6m
ldp p6,rfm_ky,,p7 set p6 to next index entry rfmiwn6m
a9bd 1,x0,p6 inc by size of dbkey + key rfmiwn6m
rem where 1 in y field = 4 bytes. rfmiwn6m
rfm8tb null
mlr (ar,,,x0),(ar) copy the dbkey rfmiwn6m
adsc9 0,,4,p6 to dbkr. rfmiwn6m
adsc9 dbkr-bias,,4,p4 the dup flag is a bit 0 of dbkey. rfmiwn6m
szn dbkr-bias,,p4 if set rfmiwn6m
tmi rfmnby ..then skip.. rfmiwn6m
code snippet 5 - Fetch index entry from the fine level CI and copy
it to the generated code's key buffer.
rfm8db null rfmiwn6m
stp p6,rfm_ky,,p7 save pointer to this dbkey. rfmiwn6m
* rfmiwn6m
* return next key to codegen then do accounting. rfmiwn6m
ldp p5,rfm_kr,,p7 p5 = key return pointer rfmiwn6m
mlr (ar,r1),(ar,r1) move key contents rfmiwn6m
adsc9 0,,x0,p6 including rap if partitioned rfmiwn6m
adsc9 0,,x0,p5 to key buffer. rfmiwn6m
* rfmiwn6m
rfm8ok null rfmiwn6m
lda idxlvl-bias,,p4 a = index into idx_level rfmiwn6m
aos idx_ky,a1,p0 inc the key currency key id. rfmiwn6m
ldq dbkr-bias,,p4 set the current rfmiwn6m
stq idx_dp,,p0 dbkey in the idx structure. rfmiwn6m
tra i8nret take the normal return. rfmiwn6m
code snippet 6 - RFM call.
* rfmiwn6m
* call rfm because an exception was detected. rfmiwn6m
* rfmiwn6m
rfmnby null call rfm rfmiwn6m
lda 1,,p5 ar = ptr to xpt rfmiwn6m
ldx0 stfsiz,du x0 = size of current fram rfmiwn6m
ldp1 ixnptr p1 = addr of rfm routine rfmiwn6m
eppr0 *+3 p0 = return from rfm rfmiwn6m
tra 0,,p1 go to rfm rfmiwn6m
zero 1,0 1 argument, 0 attrib's rfmiwn6m
stq savq-bias,,p4 save return indicator rfmiwn6m
ldp5 savp5-bias,,p4 reload p5 rfmiwn6m
* check for errors from rfm rfmiwn6m
lda savq-bias,,p4 was it alt return rfmiwn6m
tnz rfm8er yes - skip rfmiwn6m
code snippet 7 - Save the CI header control words.
* rfmiwn6m
* save the ci header info. rfmiwn6m
* the ci is not locked. therefore, it can change out from under rfmiwn6m
* get_next calls, to detect change, this information will be rfmiwn6m
* used to validate that the ci referenced in the next get_next rfmiwn6m
* call is the same one that was accessed on this call. Otherwise,
rfmiwn6m
* the ci would have to be locked (and unlocked). rfmiwn6m
* rfmiwn6m
rfm8nr null rfmiwn6m
ldp p7,1,,p5 p7 = rfm_xpt rfmiwn6m
lda rfm_ky,,p7 if rfm_xpt.r.key_data_ptr rfmiwn6m
cmpa g8null,dl is null rfmiwn6m
tze i8nret ..then skip.. rfmiwn6m
ana =o7777,dl strip all but segid (start of ci) rfmiwn6m
sta temp-bias,,p4 then save. rfmiwn6m
ldp p6,temp-bias,,p4 p6 = pointer to ci. rfmiwn6m
ldp p7,2,,p5 p7 = area to store ci header info. rfmiwn6m
ldaq 0,,p6 copy ci header info rfmiwn6m
staq 0,,p7 for later reference. rfmiwn6m
ldaq 2,,p6 rfmiwn6m
staq 2,,p7 rfmiwn6m
While in accordance with the provisions and statutes there has been illustrated and described the best form of the invention, certain changes may be made without departing from the spirit of the invention as set forth in the appended claims and that in some cases, certain features of the invention may be used to advantage without a corresponding use of other features.
| «Previous |
Next» |
| Method and data processing system for detecting patterns in SQL to allow optimized use of multi-column indexes |
Method and system for using dynamically generated code to perform record management layer functions in a relational database manager |
|
- Inventors
Edwards, David S.; Kneisel, Todd;
- Assignee
Bull HN Information Systems Inc. (Billerica, MA)
- Published
Mar-5-2002
- Current US Classes:
707/10
707/101
707/2
707/4
709/215
711/100
711/216
714/11
- Application #
408986
- International Classes
G06F 017/30
- Field of Search
707/1 707/2 707/3 707/4 707/5 707/10 707/101 707/102 707/103 709/201 709/215 709/216 711/4 711/100 711/112 711/150 711/171 711/200 711/202 711/216 711/217 714/11 705/35
- Examiner
Alam; Hosain T.
- Agent
Driscoll; Faith F., Solakian; John S.
- US Patent References:
5283894
5408652
5544355
5555388
5659727
5706495
5717919
5724569
5794228
5794229
5812996
5822749
5873099
5918225
5956706
6073129
6115703
6175835
6192370
6202070
6233569
6240422
6240428
6289334
|
