Copyright 2007, Information Builders. Slide 1 Non-Relational Efficiencies Walter F. Blood Technical Director Information Builders, Inc
Non-Relational Efficiencies Relational Efficiencies Why is Relational Efficiency important? Code optimization for the relational engine Translation of JOINS Aggregation and sorting Record selection Virtual fields Target – reduce answer set returned to minimum to get fastest return Making maximum use of the relational engine Limiting data traffic to a minimum.
Non-Relational Efficiencies Non-Relational Efficiency has similar target Target – get answer set in shortest period of time Make optimum use of WebFOCUS engine functionality What is Non-relational? How do you recognize efficiency?.
Non-Relational Efficiencies Non-relational databases – FOCUS, flatfiles,… Non SQL Intermediate files created during the development of a report Combinations of Non-relational datafiles and relational tables Creating non-relational output – PDF, EXCEL, comma delimited. When Are We Non-Relational?
Efficiency ? Non-Relational Efficiencies Recognizing Efficiency Efficiency ?. Relative Baseline Incremental Vary and Test Measure System CPU IOs WebFOCUS Records Lines
CPU? Non-Relational Efficiencies Recognizing Efficiency From operating system – process based ps -u username time programname
CPU? Non-Relational Efficiencies Recognizing Efficiency From WebFOCUS – fine tuning control -SET &TIME1 = &FOCCPU; WebFOCUS processing -SET &TIME2 = &FOCCPU; -SET &CPUTIME=&TIME2 - &TIME1;
I/O? Non-Relational Efficiencies Recognizing Efficiency From operating system – process based -device oriented iostat
I/O? Non-Relational Efficiencies Recognizing Efficiency From WebFOCUS – fine tuning control -TYPE &READS -TYPE &WRITES -IF &LINES GT 0 THEN GOTO CONTINUE; -IF &RECORDS EQ 0 THEN GOTO EXIT;
ELAPSED TIME ? Non-Relational Efficiencies Recognizing Efficiency From operating system – process based time programname
ELAPSED TIME? Non-Relational Efficiencies Recognizing Efficiency From WebFOCUS – fine tuning control - SET &START = HHMMSS(‘A8’); DEFINE FILE ABC TIME1/A8 WITH FLD=HHMMSS(TIME1);
Non-Relational Efficiencies Web-FOCUS Functionality … in Structures … in Connections … in Expressions … in Selections … in Sorts Where in Web-FOCUS?
Copyright 2007, Information Builders. Slide 13 Non-Relational Efficiencies …In Structures Metadata – Master Files - MFDs Trim out unused fieldsRe-Describe data Index, index, indexMcGyver
Non-Relational Efficiencies …in Structures Trim out unused fields Less parsing Smaller data buffers Remove from master 2. Replace with FILLER in MFD
Non-Relational Efficiencies …in Structures &TEST = Re-Describe data Identify repetitive fields in master file
Non-Relational Efficiencies …in Structures &TEST = Re-Describe data Combine repetitive fields with OCCURS clause &TEST = 128.9
Non-Relational Efficiencies …in Structures Re-Describe data Combine repetitive fields with OCCURS clause
Non-Relational Efficiencies …in Structures Re-Describe data Look for data that has rectype behavior Characterizing data Common position
Non-Relational Efficiencies …in Structures 1. For JOINing – add Internal index on HOLD 2. External index – for static data 3. Multi-dimensional index – REBUILD 4. Create dimension file to act as index - JOIN Index, index, index
Non-Relational Efficiencies …in Structures For homework: FOCALPOINT or techsupport McGyver
Copyright 2007, Information Builders. Slide 21 Non-Relational Efficiencies …In Connections Two Modes of Connecting Nested Loop - JOIN Sort and Merge - MATCH
Copyright 2007, Information Builders. Slide 22 Non-Relational Efficiencies …In Connections Nested Loop AB C
Copyright 2007, Information Builders. Slide 23 Non-Relational Efficiencies …In Connections Sort Merge A B C
Copyright 2007, Information Builders. Slide 24 Non-Relational Efficiencies …In Connections Equality based JOIN Most direct connection Multiple fields Conditional JOIN Less efficient Provides additional functionality Expression controls connection/efficiency JOIN Efficiency Considerations
Copyright 2007, Information Builders. Slide 25 Non-Relational Efficiencies …JOINs
Copyright 2007, Information Builders. Slide 26 Non-Relational Efficiencies …JOINs
Copyright 2007, Information Builders. Slide 27 Non-Relational Efficiencies Connect Based on Range of Values Connection can also be based upon range - LE, LT, GE, GT, NE or FROM /TO EMPLOYEE ID LAST NAME FIRST NAME HIRE DATE CURRENT SALARY … EMPLOYEE TAX YEAR TAX RATE MINIMUM SALARY MAXIMUM SALARY TAX RATE JOIN FILE file AT field TO ALL FILE file AT field AS name WHERE condition
Copyright 2007, Information Builders. Slide 28 Non-Relational Efficiencies Connect Based on Range of Values
Copyright 2007, Information Builders. Slide 29 Non-Relational Efficiencies Connect Based on Expression Connect files that have no apparent connection Employee Salary History New Car Finance Packages JOIN?
Copyright 2007, Information Builders. Slide 30 Non-Relational Efficiencies Connect Based on Expression Employee Salary History New Car Finance Packages JOIN?
Copyright 2007, Information Builders. Slide 31 Non-Relational Efficiencies Connect Based on Expression Employee Salary History New Car Finance Packages JOIN?
Copyright 2007, Information Builders. Slide 32 OLD-OR-NEW – all records from first and second files. This is the default if the AFTER MATCH line is omitted. (The UNION of the sets.) OLD-AND-NEW – only records common to both files. (The INTERSECTION of the sets.) OLD-NOT-NEW – records from the first file with no match in the second file. MATCH Efficiency – variety of outputs Non-Relational Efficiencies MATCH Output
Copyright 2007, Information Builders. Slide 33 NEW-NOT-OLD - records from second file with no match in the first file. OLD-NOR-NEW - non-matching records from both files - records from the first file with no match in the second file, and records from the second file with no match in the first file. OLD – records from the first file with matching records in the second file. NEW – records from the second file with matching records in the first file. Non-Relational Efficiencies MATCH Output
Copyright 2007, Information Builders. Slide 34 Non-Relational Efficiencies Connect Based on Expression 1. What types of files? How large? 2. Where are the fields you are selecting on? 3. Is the connection equality or conditional ? 4. If conditional how complex is the expression ? 5. Where are short paths expected? Host File Xref File JOIN or MATCH
Copyright 2007, Information Builders. Slide 35 Non-Relational Efficiencies …In Expressions DATA IN MATRIX DATA IN DEFINE IF WHERE BY COMPUTE IF TOTAL WHERE TOTAL BY TOTAL
Copyright 2007, Information Builders. Slide 36 Non-Relational Efficiencies …In Expressions For virtual fields, timing is key DEFINE On initial read of data Maximum data volume COMPUTE On exit from matrix Equal or smaller volume Efficiency – evaluate on smaller volume
Copyright 2007, Information Builders. Slide 37 Non-Relational Efficiencies …In Expressions For expressions, not all subroutines are equal Referencing other data DECODE function DBLOOKUP subroutine -READFILE dialogue manager command JOIN with LOOKUP/FIND Changing data CTRAN – 1 character STRREP – string of characters GETTOK/SUBSTR – positional change Efficiency – check possibilities for optimum
Copyright 2007, Information Builders. Slide 38 Non-Relational Efficiencies …In Expressions Working with more than one record instance at a time Writing to multiple files simultaneously
Copyright 2007, Information Builders. Slide 39 Non-Relational Efficiencies …In Expressions Working with more than one record instance at a time Available in COMPUTE, DEFINE, WHERE References the field value in the previous record Works with real and virtual fields THE FUNCTION
Copyright 2007, Information Builders. Slide 40 Non-Relational Efficiencies …In Expressions In COMPUTE
Copyright 2007, Information Builders. Slide 41 Non-Relational Efficiencies …In Expressions Same as DEFINE Operates on matrix Processes same or fewer records than DEFINE In COMPUTE
Copyright 2007, Information Builders. Slide 42 Non-Relational Efficiencies …In Expressions In WHERE LAST Considerations: The order the data is processed is critical You may need to presort data to a hold file Not optimizable to relational databases DEFINES work with WHERE COMPUTES work with WHERE TOTAL In DEFINEIn COMPUTE
Non-Relational Efficiencies …In Expressions Writing to multiple files simultaneously To create a log of specific data values read Which customers placed orders? To create of record of calculated values Will I go out of stock on any these orders? To create additional output What page number will this item be on? PUTDDREC
Non-Relational Efficiencies …In Expressions PUTDDREC syntax: PUTDDREC(ddname, dd_len, record_string, record_len, outfield) ddname - ddname assigned by filedef to output file dd_len - length of the ddname record_string - string of characters to write to file or field containing that string record_len- length of the string to be included outfield - return code Output file must be filedef’d DEFINE/COMPUTE field format always I1 Handles open, write and close
Non-Relational Efficiencies …In Expressions FILEDEF PUTDD1 DISK putdd1.dat TABLE FILE EMPLOYEE PRINT EMP_ID CURR_JOBCODE AS 'JOB' CURR_SAL COMPUTE SALA/A12 = EDIT(CURR_SAL); NOPRINT COMPUTE EMP1/A50= LAST_NAME|FIRST_NAME|EMP_ID|CURR_JOBCODE|SALA ; NOPRINT COMPUTE OUT1/I1 = PUTDDREC('PUTDD1',6, EMP1, 50, OUT1); BY LAST_NAME BY FIRST_NAME END PUTDDREC in Action
Non-Relational Efficiencies …In Expressions LAST_NAME FIRST_NAME EMP_ID JOB CURR_SAL OUT BANNING JOHN A17 $29, BLACKWOOD ROSEMARIE B04 $21, CROSS BARBARA A17 $27, GREENSPAN MARY A07 $9, IRVING JOAN A15 $26, JONES DIANE B03 $18, MCCOY JOHN B02 $18, MCKNIGHT ROGER B02 $16, ROMANS ANTHONY B04 $21, SMITH MARY B14 $13, RICHARD A01 $9, STEVENS ALFRED A07 $11, Report Created
Non-Relational Efficiencies …In Expressions BANNING JOHN A BLACKWOOD ROSEMARIE B CROSS BARBARA A GREENSPAN MARY A IRVING JOAN A JONES DIANE B MCCOY JOHN B MCKNIGHT ROGER B ROMANS ANTHONY B SMITH MARY B SMITH RICHARD A STEVENS ALFRED A Sequential File Created
Non-Relational Efficiencies …In Expressions PUTDDREC Considerations: Create fixed data file to fit a specific Master file Create comma-delimited data file for loading System and User &variables are available &MDYY, &FOCCPU, &FOCUSER, etc Write control at any point or multiple points Write to multiple PUTDDREC files in a request
Non-Relational Efficiencies …in Selections WHERE more expensive than IF IF limited to comparison to values – match array WHERE allows “expression operator expression” WHERE Decomposed IFs removed WHEREs converted to DEFINE/IF combination DEFINE expressions converted to polishes
Non-Relational Efficiencies …in Selections Selection placement Within the request Before matrix After matrix – TOTAL test Within the data Indexed or alternative views in TABLE Type of comparison Selection considerations
Copyright 2007, Information Builders. Slide 51 Non-Relational Efficiencies Questions?