RD - Revision history

Gordon.dye at 02:59, 21 January 2020

2020-01-21T02:59:03Z

← Older revision		Revision as of 02:59, 21 January 2020
Line 10:		Line 10:
	If changing the RD value is necessary, it is important that you change it properly. Using an RD value that is too high could cause unpredictable and invalid results. Using an RD value that is too low could cause Business Rules to overlook a significant difference between two numbers.		If changing the RD value is necessary, it is important that you change it properly. Using an RD value that is too high could cause unpredictable and invalid results. Using an RD value that is too low could cause Business Rules to overlook a significant difference between two numbers.

	The best way to determine the appropriate setting for your needs is to find the largest number, with the most number of decimal positions, that your application will be using in numeric comparisons (remember, RD has no effect on ~~anything other than comparisons~~). Starting from the left-most digit of the number, you should then count 15 digits to the right and draw a line after that digit. (If there are less than 15 digits in your number, draw a line after the last digit.) Now, count the number of digits between the decimal point and your line. Subtract at least one, but preferably two. The number you end up with is the RD setting you should specify.		The best way to determine the appropriate setting for your needs is to find the largest number, with the most number of decimal positions, that your application will be using in numeric comparisons (remember, RD has no effect on actual math performed internally). Starting from the left-most digit of the number, you should then count 15 digits to the right and draw a line after that digit. (If there are less than 15 digits in your number, draw a line after the last digit.) Now, count the number of digits between the decimal point and your line. Subtract at least one, but preferably two. The number you end up with is the RD setting you should specify.

	For example, assume that the following number is the largest that would be used in numeric comparisons for a particular application.		For example, assume that the following number is the largest that would be used in numeric comparisons for a particular application.

Gordon.dye: /* Technical Considerations */

2014-07-05T18:13:55Z

Technical Considerations

← Older revision		Revision as of 18:13, 5 July 2014
Line 129:		Line 129:
	:7.) A useful program for experimenting with extreme numeric values is;		:7.) A useful program for experimenting with extreme numeric values is;

	00010 ~~input fields~~ "5,10,l 20,u": num_var		00010 INPUT FIELDS "5,10,L 20,u": num_var
	00020 ~~print fields~~ "7,10,L 20": num_var		00020 PRINT FIELDS "7,10,L 20": num_var

Gordon.dye: /* Technical Considerations */

2014-07-05T18:12:40Z

Technical Considerations

← Older revision		Revision as of 18:12, 5 July 2014
Line 125:		Line 125:
	STR$(ROUND(NBR,DECIMALS))		STR$(ROUND(NBR,DECIMALS))

	:5.) If it is undesirable to do any rounding of a number, either with the RD specification or with the ROUND function, the next best way to get an accurate comparison is to subtract potentially large numbers that are the result of many calculations and compare the result. By using IF A-B>=.01 instead of IF A>B, the desired variance is specifically tested for.		:6.) If it is undesirable to do any rounding of a number, either with the RD specification or with the ROUND function, the next best way to get an accurate comparison is to subtract potentially large numbers that are the result of many calculations and compare the result. By using IF A-B>=.01 instead of IF A>B, the desired variance is specifically tested for.

			:7.) A useful program for experimenting with extreme numeric values is;

			00010 input fields "5,10,l 20,u": num_var
			00020 print fields "7,10,L 20": num_var


	<noinclude>		<noinclude>
	[[Category:Config]]		[[Category:Config]]
	</noinclude>		</noinclude>

Gordon.dye: /* Comments and Examples */

2014-07-05T18:04:33Z

Comments and Examples

← Older revision		Revision as of 18:04, 5 July 2014
Line 20:		Line 20:
	123,456,789.012345\|67		123,456,789.012345\|67

	The number of digits between the decimal point and the line is 6. Subtract two from that and the result is 4, which would be the appropriate value for your system. ''What this accomplishes is to discard the rightmost two digits of precision for comparison and display purposes. '' So even though BR uses an internal precision of 15, for many additions or subtractions this effectively drops the precision back to 13 significant digits.		The number of digits between the decimal point and the line is 6. Subtract two from that and the result is 4, which would be the appropriate value for your system. ''What this accomplishes is to discard the rightmost two digits of precision for comparison and display purposes. '' So even though BR uses an internal precision of 15, for situations where many additions and/or subtractions are performed this effectively drops the precision back to 13 significant digits.

	To understand the significance of the RD specification, it is important to understand that adding or subtracting a number with other numbers many times has a predictable loss of precision. (Any language that uses binary floating point numbers will lose precision in this manner.)		To understand the significance of the RD specification, it is important to understand that adding or subtracting a number with other numbers many times has a predictable loss of precision. (Any language that uses binary floating point numbers will lose precision in this manner.)
Line 104:		Line 104:


	* The default value of 6 will allow a value of 9,999,999,999.999999 to be ~~used~~.		* The default value of 6 will allow a value of 999,999,999.999999 to be output, input and displayed. But if you were to add a large quantity of numbers in the 100,000,000 range that also had fractional amounts, you could get an inaccuracy at the fifth and sixth decimal positions.
	* Changing the value to ~~8 will allow for~~ a ~~value~~ of 99,~~999~~,~~999.99999999 to be used~~.

	====Syntax====		====Syntax====

Gordon.dye: /* Comments and Examples */

2014-07-05T17:45:27Z

Comments and Examples

← Older revision		Revision as of 17:45, 5 July 2014
Line 4:		Line 4:

	====Comments and Examples====		====Comments and Examples====
	Business Rules floating point numbers are precise up to the first 15 digits, whether the digits occur before or after the decimal point. The value of the RD specification indicates how many decimal digits should be considered meaningful in numeric comparisons. It has no effect on internal ~~calculations that are not involved in numeric calculations~~.		Business Rules floating point numbers are precise up to the first 15 digits, whether the digits occur before or after the decimal point. The value of the RD specification indicates how many decimal digits should be considered meaningful in numeric comparisons. And when formatting a numeric value for output it rounds at the RD position. It has no effect on internal mathematical processing.

	For most applications, the RD default should be used. The RD specification is only needed when the data includes very large numbers (100 million or greater) or very small numbers (less than .00001).		For most applications, the RD default should be used. The RD specification is only needed when the data includes very large numbers (100 million or greater) or very small numbers (less than .00001).
Line 20:		Line 20:
	123,456,789.012345\|67		123,456,789.012345\|67

	The number of digits between the decimal point and the line is 6. Subtract two from that and the result is 4, which would be the appropriate value for your system.		The number of digits between the decimal point and the line is 6. Subtract two from that and the result is 4, which would be the appropriate value for your system. ''What this accomplishes is to discard the rightmost two digits of precision for comparison and display purposes. '' So even though BR uses an internal precision of 15, for many additions or subtractions this effectively drops the precision back to 13 significant digits.

	To understand the significance of the RD specification, it is important to understand that adding or subtracting a number with other numbers many times has a predictable loss of precision. (Any language that uses binary floating point numbers will lose precision in this manner.)		To understand the significance of the RD specification, it is important to understand that adding or subtracting a number with other numbers many times has a predictable loss of precision. (Any language that uses binary floating point numbers will lose precision in this manner.)
Line 51:		Line 51:


	The following table shows the various levels of precision supported by RD settings:		The following table shows the various levels of addition/subtraction precision supported by RD settings:

	{\| border=1		{\| border=1
Line 57:		Line 57:
	!# Digits		!# Digits
	!Precision		!Precision
	\|-
	~~\|RD 13~~
	~~\|3.12~~
	~~\|###.############~~
	\|-		\|-
	\|RD 12		\|RD 12
	\|4.11		\|1.12
	\|#~~,##~~#.###########		\|#.############
	\|-		\|-
	\|RD 11		\|RD 11
	\|5.10		\|2.11
	\|##~~,##~~#.##########		\|##.###########
	\|-		\|-
	\|RD 10		\|RD 10
	\|6.10		\|3.10
	\|###~~,##~~#.#########		\|###.##########
	\|-		\|-
	\|RD 9		\|RD 9
	\|7.9		\|4.9
	\|#,###,#~~##.~~########		\|#,###.#########
	\|-		\|-
	\|RD 8		\|RD 8
	\|8.8		\|5.8
	\|~~##,#~~##,###.########		\|##,###.########
	\|-		\|-
	\|RD 7		\|RD 7
	\|9.7		\|6.7
	\|~~###,~~###,###.#######		\|###,###.#######
	\|-		\|-
	\|RD 6		\|RD 6
	\|10.6		\|7.6
	\|#,###,###,###.######		\|#,###,###.######
			\|-
			\|RD 5
			\|8.5
			\|##,###,###.#####
			\|-
			\|RD 4
			\|9.4
			\|###,###,###.####
			\|-
			\|RD 3
			\|10.3
			\|#,###,###,###.###
			\|-
			\|RD 2
			\|11.2
			\|##,###,###,###.##
	\|}		\|}

Gordon.dye: /* Comments and Examples */

2014-07-05T16:03:48Z

Comments and Examples

← Older revision		Revision as of 16:03, 5 July 2014
Line 51:		Line 51:


	The following table shows the various levels of ~~prevision~~ supported by RD settings:		The following table shows the various levels of precision supported by RD settings:

	{\| border=1		{\| border=1

Laura at 16:25, 8 April 2013

2013-04-08T16:25:46Z

← Older revision		Revision as of 16:25, 8 April 2013
Line 1:		Line 1:
	The '''RD''' [[BRconfig.sys]] specification sets the tolerance level for assuming two values are equal. The value of the RD specification affects operation of the [[AIDX]], [[DIDX]], [[INT]] X, [[IP]] X and [[STR$]] [[~~internal functions~~]], the [[SORT]] utility's [[RECORD]] specification, numeric comparisons used in [[IF]] and [[FOR]] [[statements]], and the printing of a number in default format (PRINT X).		The '''RD''' [[BRconfig.sys]] specification sets the tolerance level for assuming two values are equal. The value of the RD specification affects operation of the [[AIDX]], [[DIDX]], [[INT]] X, [[IP]] X and [[STR$]] [[Internal Functions]], the [[SORT]] utility's [[RECORD]] specification, numeric comparisons used in [[IF]] and [[FOR]] [[statements]], and the printing of a number in default format (PRINT X).

	(Not available in BR! Version [[4.03]]jx or before)		(Not available in BR! Version [[4.03]]jx or before)

Laura at 02:05, 4 April 2013

2013-04-04T02:05:38Z

← Older revision		Revision as of 02:05, 4 April 2013
Line 97:		Line 97:
	====Syntax====		====Syntax====
	[[Image:RD.png\|300px]]		[[Image:RD.png\|300px]]

			====Parameter====
			"Integer" is a number from zero to fifteen that represents the number of decimal places to be used in numeric comparisons.

	====Start-up Default====		====Start-up Default====
	RD 6		RD 6

	~~====Parameter====~~
	~~"Integer" is a number from zero to fifteen that represents the number of decimal places to be used in numeric comparisons.~~

	====Technical Considerations====		====Technical Considerations====
Line 108:		Line 108:
	:2.) Using RD values larger than 8 will most likely result in unexpected rounding as RD 8 only support values up to 9 million.		:2.) Using RD values larger than 8 will most likely result in unexpected rounding as RD 8 only support values up to 9 million.
	:3.) When numbers larger than 99,999,999.99 are used in comparisons, the RD value should be set to three or four.		:3.) When numbers larger than 99,999,999.99 are used in comparisons, the RD value should be set to three or four.
	:4.) In special situations where changing the RD value is not desirable, the ROUND function may be used to stop rounding errors from accumulating. See ~~the~~ Functions ~~chapter~~ for more information.		:4.) In special situations where changing the RD value is not desirable, the ROUND function may be used to stop rounding errors from accumulating. See [[Functions]] for more information.
	:5.) The RD specification affects the STR$ function since the number of significant decimals is not specified by STR$. If the number to be converted is large and the number is the result of many computations, one of the following two techniques should be used to truncate the number and accomplish the required results instead. The CNVRT$ function may be used with a PIC format or N specification or the STR$ function may be used in conjunction with the ROUND function. The following examples show each of these techniques:		:5.) The RD specification affects the STR$ function since the number of significant decimals is not specified by STR$. If the number to be converted is large and the number is the result of many computations, one of the following two techniques should be used to truncate the number and accomplish the required results instead. The CNVRT$ function may be used with a PIC format or N specification or the STR$ function may be used in conjunction with the ROUND function. The following examples show each of these techniques:

Line 114:		Line 114:
	STR$(ROUND(NBR,DECIMALS))		STR$(ROUND(NBR,DECIMALS))

	:5.) If it is undesirable to do any rounding of a number, either with the RD specification or with the ROUND function, the next best way to get an accurate comparison is to subtract potentially large numbers that are the result of many calculations and compare the result. By using IF A-B>=.01 instead of IF A>B, the desired variance is specifically tested for.		:5.) If it is undesirable to do any rounding of a number, either with the RD specification or with the ROUND function, the next best way to get an accurate comparison is to subtract potentially large numbers that are the result of many calculations and compare the result. By using IF A-B>=.01 instead of IF A>B, the desired variance is specifically tested for.

	<noinclude>		<noinclude>
	[[Category:Config]]		[[Category:Config]]
	</noinclude>		</noinclude>

Laura: /* Syntax */

2013-04-04T02:03:13Z

Syntax

← Older revision		Revision as of 02:03, 4 April 2013
Line 96:		Line 96:

	====Syntax====		====Syntax====
	[[Image:~~HELP0021~~.~~jpg~~]]		[[Image:RD.png\|300px]]

	====Start-up Default====		====Start-up Default====

Laura: /* Comments and Examples */

2013-04-04T01:59:16Z

Comments and Examples

← Older revision		Revision as of 01:59, 4 April 2013
Line 29:		Line 29:
	\| RD \|\| Loop Count \|\| Variance		\| RD \|\| Loop Count \|\| Variance
	\|-		\|-
	\| 6 \|\| 94\|\| 0.00000051		\| 6 \|\| 94 \|\| 0.00000051
	\|-		\|-
	\| 5 \|\| 932 \|\| 0.00000501		\| 5 \|\| 932 \|\| 0.00000501
Line 35:		Line 35:
	\| 4\|\| 9321 \|\| 0.00005001		\| 4\|\| 9321 \|\| 0.00005001
	\|-		\|-
	\| 3\|\| over 10001 \|\|		\| 3 \|\| over 10001 \|\|
	\|}		\|}

	While the above worst case is an unlikely occurrence with real data, it does highlight the potential for problems if more than 1000 numbers are added or subtracted and the resulting figure is in the 100 million or larger range. Note that the difference in a number printed on a report is still at least 1000 times too small to show up as a penny rounding error.		While the above worst case is an unlikely occurrence with real data, it does highlight the potential for problems if more than 1000 numbers are added or subtracted and the resulting figure is in the 100 million or larger range. Note that the difference in a number printed on a report is still at least 1000 times too small to show up as a penny rounding error.

	;In contrast to the worst case, a more "typical case" illustration could add a very large number (99,000,000) to a randomly selected small number between .01 and .99. When RD is set to 6, it would take 1930 computations of this type to cause a rounding error. With an RD of 5, it would take more than 10,000 such computations to cause a rounding error:		In contrast to the worst case, a more "typical case" illustration could add a very large number (99,000,000) to a randomly selected small number between .01 and .99. When RD is set to 6, it would take 1930 computations of this type to cause a rounding error. With an RD of 5, it would take more than 10,000 such computations to cause a rounding error:

	~~[[Image:HELP0020~~.~~jpg]]<br>~~		{\| border=1
			\| RD \|\| Loop Count \|\| Variance
			\|-
			\| 6 \|\| 1930 \|\| -0.00000051
			\|-
			\| 5 \|\| over 10001 \|\|
			\|}