This document covers the following topics:
A field (user-defined variable or database field) which is to be used as an operand in an arithmetic operation must be defined with one of the following formats:
| Format | |
|---|---|
| N | Numeric unpacked |
| P | Packed numeric |
| I | Integer |
| F | Floating point |
| D | Date |
| T | Time |
Note:
For reporting mode: A field which is to be used as an operand in an
arithmetic operation must have been previously defined. A user-defined variable
or database field used as a result field in an arithmetic operation need not
have been previously defined.
All user-defined variables and all database fields defined in a
DEFINE DATA
statement are initialized to the appropriate zero or blank value when the
program is invoked for execution.
Data transfer is performed with a MOVE or
COMPUTE statement.
The following table summarizes the data transfer compatibility of the
formats an operand
may take.
| Sending Field Format | Receiving Field Format | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| N or P | A | U | Bn (n<5) | Bn (n>4) | I | L | C | D | T | F | G | O | |
| N or P | Y | [ 2 ] | [ 14 ] | [ 3 ] | - | Y | - | - | - | Y | Y | - | - |
| A | - | Y | [ 13 ] | [ 1 ] | [ 1 ] | - | - | - | - | - | - | - | - |
| U | - | [ 11 ] | Y | [ 12 ] | [ 12 ] | - | - | - | - | - | - | - | - |
| Bn (n<5) | [ 4 ] | [ 2 ] | [ 14 ] | [ 5 ] | [ 5 ] | Y | - | - | - | Y | Y | - | - |
| Bn (n>4) | - | [ 6 ] | [ 15 ] | [ 5 ] | [ 5 ] | - | - | - | - | - | - | - | - |
| I | Y | [ 2 ] | [ 14 ] | [ 3 ] | - | Y | - | - | - | Y | Y | - | - |
| L | - | [ 9 ] | [ 16 ] | - | - | - | Y | - | - | - | - | - | - |
| C | - | - | - | - | - | - | - | Y | - | - | - | - | - |
| D | Y | [ 9 ] | [ 16 ] | Y | - | Y | - | - | Y | [7] | Y | - | - |
| T | Y | [ 9 ] | [ 16 ] | Y | - | Y | - | - | [ 8 ] | Y | Y | - | - |
| F | Y | [ 9 ] [ 10 ] | [ 10 ] [ 16 ] | [ 3 ] | - | Y | - | - | - | Y | Y | - | - |
| G | - | - | - | - | - | - | - | - | - | - | - | Y | - |
| O | - | - | - | - | - | - | - | - | - | - | - | - | Y |
Where:
| Y | Indicates data transfer compatibility. |
|---|---|
| - | Indicates data transfer incompatibility. |
| [ ] | Numbers in brackets [ ] refer to the corresponding rule for data transfer given below. |
The following rules apply to converting data values:
Alphanumeric to binary:
The value will be moved byte by byte from left to right. The result
may be truncated or padded with trailing blank characters depending on the
length defined and the number of bytes specified.
(N,P,I) and binary (length 1-4) to alphanumeric:
The value will be converted to unpacked form and moved into the
alphanumeric field left justified, that is, leading zeros will be suppressed
and the field will be filled with trailing blank characters. For negative
numeric values, the sign will be converted to the hexadecimal notation
Dx. Any decimal point in the numeric
value will be ignored. All digits before and after the decimal point will be
treated as one integer value.
(N,P,I,F) to binary (1-4 bytes):
The numeric value will be converted to binary (4 bytes). Any decimal
point in the numeric value will be ignored (the digits of the value before and
after the decimal point will be treated as an integer value). The resulting
binary number will be positive or a two's complement of the number depending on
the sign of the value.
Binary (1-4 bytes) to numeric:
The value will be converted and assigned to the numeric value right
justified, that is, with leading zeros. (Binary values of the length 1-3 bytes
are always assumed to have a positive sign. For binary values of 4 bytes, the
leftmost bit determines the sign of the number: 1=negative, 0=positive.) Any
decimal point in the receiving numeric value will be ignored. All digits before
and after the decimal point will be treated as one integer value.
Binary to binary:
The value will be moved from right to left byte by byte. Leading
binary zeros will be inserted into the receiving field.
Binary (>4 bytes) to alphanumeric:
The value will be moved byte by byte from left to right. The result
may be truncated or padded with trailing blanks depending on the length defined
and the number of bytes specified.
Date (D) to time (T):
If date is moved to time, it is converted to time assuming time
00:00:00:0.
Time (T) to date (D):
If time is moved to date, the time information is truncated, leaving
only the date information.
L,D,T,F to A:
The values are converted to display form and are assigned left
justified.
F:
If F is assigned to an alphanumeric or Unicode field which is too
short, the mantissa is reduced accordingly.
Unicode to alphanumeric:
The Unicode value will be converted to alphanumeric character codes
according to the default code page (value of the system variable
*CODEPAGE)
using the International Components for Unicode (ICU) library. The result may be
truncated or padded with trailing blank characters, depending on the length
defined and the number of bytes specified.
Unicode to binary:
The value will be moved code unit by code unit from left to right.
The result may be truncated or padded with trailing blank characters, depending
on the length defined and the number of bytes specified. The length of the
receiving binary field must be even.
Alphanumeric to Unicode:
The alphanumeric value will be converted from the default code page
to a Unicode value using the International Components for Unicode (ICU)
library. The result may be truncated or padded with trailing blank characters,
depending on the length defined and the number of code units specified.
(N,P,I) and binary (length 1-4) to Unicode:
The value will be converted to unpacked form from which an
alphanumeric value will be obtained by suppression of leading zeros. For
negative numeric values, the sign will be converted to the hexadecimal notation
Dx. Any decimal point in the numeric
value will be ignored. All digits before and after the decimal point will be
treated as one integer value. The resulting value will be converted from
alphanumeric to Unicode. The result may be truncated or padded with trailing
blank characters, depending on the length defined and the number of code units
specified.
Binary (>4 bytes) to Unicode:
The value will be moved byte by byte from left to right. The result
may be truncated or padded with trailing blanks, depending on the length
defined and the number of bytes specified. The length of the sending binary
field must be even.
L,D,T,F to U:
The values are converted to an alphanumeric display form. The
resulting value will be converted from alphanumeric to Unicode and assigned
left justified.
If source and target format are identical, the result may be truncated or padded with trailing blank characters (format A and U) or leading binary zeros (format B) depending on the length defined and the number of bytes (format A and B) or code units (format U) specified.
See also Using Dynamic Variables.
The following rules apply to field truncation and rounding:
High-order numeric field truncation is allowed only when the digits to be truncated are leading zeros. Digits following an expressed or implied decimal point may be truncated.
Trailing positions of an alphanumeric field may be truncated.
If the option ROUNDED is specified, the last position of
the result will be rounded up if the first truncated decimal position of the
value being assigned contains a value greater than or equal to 5. For the
result precision of a division, see also Precision of Results of Arithmetic
Operations.
The following table shows the format and length of the result of an arithmetic operation:
| I1 | I2 | I4 | N or P | F4 | F8 | |
|---|---|---|---|---|---|---|
| I1 | I1 | I2 | I4 | P* | F4 | F8 |
| I2 | I2 | I2 | I4 | P* | F4 | F8 |
| I4 | I4 | I4 | I4 | P* | F4 | F8 |
| N or P | P* | P* | P* | P* | F4 | F8 |
| F4 | F4 | F4 | F4 | F4 | F4 | F8 |
| F8 | F8 | F8 | F8 | F8 | F8 | F8 |
On a mainframe computer, format/length F8 is used instead of F4 for improved precision of the results of an arithmetic operation.
P* is determined from the integer length and precision of the operands individually for each operation, as shown under Precision of Results of Arithmetic Operations.
The following decimal integer lengths and possible values are applicable for format I:
| Format/Length | Decimal Integer Length | Possible Values |
|---|---|---|
| I1 | 3 | -128 to 127 |
| I2 | 5 | -32768 to 32767 |
| I4 | 10 | -2147483648 to 2147483647 |
The following topics are covered below:
Floating-point numbers (format F) are represented as a sum of powers of two (as are integer numbers (format I)), whereas unpacked and packed numbers (formats N and P) are represented as a sum of powers of ten.
In unpacked or packed numbers, the position of the decimal point is fixed. In floating-point numbers, however, the position of the decimal point (as the name indicates) is "floating", that is, its position is not fixed, but depends on the actual value.
Floating-point numbers are essential for the computing of trigonometric functions or mathematical functions such as sinus or logarithm.
Due to the nature of floating-point numbers, their precision is limited:
For a variable of format/length F4, the precision is limited to approximately 7 digits.
For a variable of format/length F8, the precision is limited to approximately 15 digits.
Values which have more significant digits cannot be represented exactly as a floating-point number. No matter how many additional digits there are before or after the decimal point, a floating-point number can cover only the leading 7 or 15 digits respectively.
An integer value can only be represented exactly in a variable of format/length F4 if its absolute value does not exceed 2 23 -1.
When an alphanumeric, unpacked numeric or packed numeric value is converted to floating-point format (for example, in an assignment operation), the representation has to be changed, that is, a sum of powers of ten has to be converted to a sum of powers of two.
Consequently, only numbers that are representable as a finite sum of powers of two can be represented exactly; all other numbers can only be represented approximately.
This number has an exact floating-point representation:
1.25 = 20 + 2-2
This number is a periodic floating-point number without an exact representation:
1.2 = 20 + 2-3 + 2-4 + 2-7 + 2-8 + 2-11 + 2-12 + ...
Thus, the conversion of alphanumeric, unpacked numeric or packed numeric values to floating-point values, and vice versa, can introduce small errors.
Because of different hardware architecture, the representation of floating-point numbers varies according to platforms. This explains why the same application, when run on different platforms, may return slightly different results when floating-point arithmetics are involved. The respective representation also determines the range of possible values for floating-point variables, which is (approximately)
±1.17 * 10-38 to ±3.40 * 1038 for F4 variables,
±2.22 * 10-308 to ±1.79 * 10308 for F8 variables.
Note:
The representation used by your pocket calculator may also be
different from the one used by your computer - which explains why results for
the same computation may differ.
With formats D (date) and T (time), only addition, subtraction, multiplication and division are allowed. Multiplication and division are allowed on intermediate results of additions and subtractions only.
Date/time values can be added to/subtracted from one another; or integer values (no decimal digits) can be added to/subtracted from date/time values. Such integer values can be contained in fields of formats N, P, I, D, or T.
The intermediate results of such an addition or subtraction may be used as a multiplicand or dividend in a subsequent operation.
An integer value added to/subtracted from a date value is assumed to be in days. An integer value added to/subtracted from a time value is assumed to be in tenths of seconds.
For arithmetic operations with date and time, certain restrictions apply, which are due to the Natural's internal handling of arithmetic operations with date and time, as explained below.
Internally, Natural handles an arithmetic operation with date/time variables as follows:
COMPUTE
result-field = operand1
+/- operand2
|
The above statement is resolved as:
intermediate-result =
operand1 +/-
operand2
result-field =
intermediate-result
That is, in a first step Natural computes the result of the addition/subtraction, and in a second step assigns this result to the result field.
More complex arithmetic operations are resolved following the same pattern:
COMPUTE
result-field = operand1
+/- operand2 +/- operand3
+/- operand4
|
The above statement is resolved as:
intermediate-result1 =
operand1 +/-
operand2
intermediate-result2 =
intermediate-result1 +/-
operand3
intermediate-result3 =
intermediate-result2 +/-
operand4
result-field =
intermediate-result3
The resolution of multiplication and division operations is similar to the resolution for addition and subtraction.
The internal format of such an intermediate result depends on the formats of the operands, as shown in the tables below.
The following table shows the format of the intermediate result of an
addition (intermediate-result =
operand1 +
operand2):
Format of operand1
|
Format of operand2
|
Format of
intermediate-result
|
|---|---|---|
| D | D | Di |
| D | T | T |
| D | Di, Ti, N, P, I | D |
| T | D, T, Di, Ti, N, P, I | T |
| Di, Ti, N, P, I | D | D |
| Di, Ti, N, P, I | T | T |
| Di, N, P, I | Di | Di |
| Ti, N, P, I | Ti | Ti |
| Di | Ti, N, P, I | Di |
| Ti | Di, N, P, I | Ti |
The following table shows the format of the intermediate result of a
subtraction (intermediate-result =
operand1 -
operand2):
Format of operand1
|
Format of operand2
|
Format of
intermediate-result
|
|---|---|---|
| D | D | Di |
| D | T | Ti |
| D | Di, Ti, N, P, I | D |
| T | D, T | Ti |
| T | Di, Ti, N, P, I | T |
| Di, N, P, I | D | Di |
| Di, N, P, I | T | Ti |
| Di | Di, Ti, N, P, I | Di |
| Ti | D, T, Di, Ti, N, P, I | Ti |
| N, P, I | Di, Ti | P12 |
The following table shows the format of the intermediate result of a
multiplication (intermediate-result =
operand1 *
operand2) or division
(intermediate-result =
operand1 /
operand2):
Format of operand1
|
Format of operand2
|
Format of
intermediate-result
|
|---|---|---|
| D | D, Di, Ti, N, P, I | Di |
| D | T | Ti |
| T | D, T, Di, Ti, N, P, I | Ti |
| Di | T | Ti |
| Di | D, Di, Ti, N, P, I | Di |
| Ti | D | Di |
| Ti | Di, T, Ti, N, P, I | Ti |
| N, P, I | D, Di | Di |
| N, P, I | T, Ti | Ti |
Di is a value in internal date format; Ti is a value in internal time format; such values can be used in further arithmetic date/time operations, but they cannot be assigned to a result field of format D (see the assignment table below).
In complex arithmetic operations in which an intermediate result of internal format Di or Ti is used as operand in a further addition/subtraction/multiplication/division, its format is assumed to be D or T respectively.
The following table shows which intermediate results can internally be
assigned to which result fields (result-field
= intermediate-result).
Format of result-field
|
Format of
intermediate-result
|
Assignment possible |
|---|---|---|
| D | D, T | yes |
| D | Di, Ti, N, P, I | no |
| T | D, T, Di, Ti, N, P, I | yes |
| N, P, I | D, T, Di, Ti, N, P, I | yes |
A result field of format D or T must not contain a negative value.
COMPUTE DATE1 (D) = DATE2 (D) + DATE3 (D) COMPUTE DATE1 (D) = DATE2 (D) - DATE3 (D)
These operations are not possible, because the intermediate result of the addition/subtraction would be format Di, and a value of format Di cannot be assigned to a result field of format D.
COMPUTE DATE1 (D) = TIME2 (T) - TIME3 (T) COMPUTE DATE1 (D) = DATE2 (D) - TIME3 (T)
These operations are not possible, because the intermediate result of the addition/subtraction would be format Ti, and a value of format Ti cannot be assigned to a result field of format D.
COMPUTE DATE1 (D) = DATE2 (D) - DATE3 (D) + TIME3 (T)
This operation is possible. First, DATE3 is subtracted from
DATE2, giving an intermediate result of format Di; then, this
intermediate result is added to TIME3, giving an intermediate
result of format T; finally, this second intermediate result is assigned to the
result field DATE1.
COMPUTE DATE1 (D) = DATE2 (D) + DATE3 (D) * 2 COMPUTE TIME1 (T) = TIME2 (T) - TIME3 (T) / 3
These operations are not possible, because the attempted multiplication/division is performed with date/time fields and not with intermediate results.
COMPUTE DATE1 (D) = DATE2 (D) + (DATE3(D) - DATE4 (D)) * 2
This operation is possible. First, DATE4 is subtracted from
DATE3 giving an intermediate result of format Di; then, this
intermediate result is multiplied by two giving an intermediate result of
format Di; this intermediate result is added to DATE2 giving an
intermediate result of format D; finally, this third intermediate result is
assigned to the result field DATE1.
If a format T value is assigned to a format D field, you must ensure that the time value contains a valid date component.
When doing arithmetic operations, the choice of field formats has considerable impact on performance:
For business arithmetic, only fields of format I (integer) should be used, if possible.
For scientific arithmetic, fields of format F (floating point) should be used, if possible.
In expressions where formats are mixed between numeric (N, P) and floating point (F), a conversion to floating point format is performed. This conversion results in considerable CPU load. Therefore it is recommended to avoid mixed format expressions in arithmetic operations.
| Operation | Digits Before Decimal Point | Digits After Decimal Point |
|---|---|---|
| Addition/Subtraction | Fi + 1 or Si + 1 (whichever is greater) | Fd or Sd (whichever is greater) |
| Multiplication | Fi + Si + 2 | Fd + Sd (maximum 7) |
| Division | Fi + Sd | (see below) |
| Exponentiation | 15 - Fd (See Exception below) | Fd |
| Square Root | Fi | Fd |
- where:
| F | First operand |
|---|---|
| S | Second operand |
| R | Result |
| i | Digits before decimal point |
| d | Digits after decimal point |
If the exponent has one or more digits after the decimal point, the exponentiation is internally carried out in floating point format and the result will also have floating point format. See Arithmetic Operations with Floating-Point Numbers for further information.
The precision of the result of a division depends whether a result field is available or not:
If a result field is available, the precision is: Fd or Rd (whichever is greater) *.
If no result field is available, the precision is: Fd or Sd (whichever is greater) *.
* If the ROUNDED option is used, the precision
of the result is internally increased by one digit before the result is
actually rounded.
A result field is available (or assumed to be available) in a
COMPUTE and DIVIDE statement, and in a logical
condition in which the division is placed after the comparison operator (for
example: IF #A = #B / #C THEN ...).
A result field is not (or assumed to be
not) available in a logical
condition in which the division is placed before the comparison operator (for
example: IF #B / #C = #A THEN ...).
If both dividend and divisor are of integer format and at least one of
them is a variable, the division result is always of integer format (regardless
of the precision of the result field and of whether the ROUNDED
option is used or not).
The precision of arithmetic expressions, for example: #A / (#B *
#C) + #D * (#E - #F + #G), is derived by evaluating the results of the
arithmetic operations in their processing order. For further information on
arithmetic expressions, see arithmetic-expression
in the COMPUTE
statement description.
In an addition, subtraction, multiplication or division, an error occurs if the total number of digits (before and after the decimal point) of the result is greater than 31.
In an exponentiation, an error occurs in any of the following situations:
if the base is of packed format and either the result has over 16 digits or any intermediate result has over 15 digits;
if the base is of floating-point format and the result is greater than approximately 7 * 1075.
Generally, the following rules apply:
All scalar operations may be applied to array elements which consist of a single occurrence.
If a variable is defined with a constant value (for example,
#FIELD (I2) CONSTANT <8>), the value will be assigned to the
variable at compilation, and the variable will be treated as a constant. This
means that if such a variable is used in an array index, the dimension
concerned has a definite number of occurrences.
If an assignment/comparison operation involves two arrays with a different number of dimensions, the "missing" dimension in the array with fewer dimensions is assumed to be (1:1).
Example: If #ARRAY1 (1:2) is assigned to #ARRAY2
(1:2,1:2), #ARRAY1 is assumed to be #ARRAY1
(1:1,1:2).
The following topics are covered below:
The first, second and third dimensions of an array are defined as follows:
| Number of Dimensions | Properties |
|---|---|
| 3 | #a3(3rd dim., 2nd dim., 1st dim.) |
| 2 | #a2(2nd dim., 1st dim.) |
| 1 | #a1(1st dim.) |
If an array range is assigned to another array range, the assignment is performed element by element.
Example:
DEFINE DATA LOCAL 1 #ARRAY(I4/1:5) INIT <10,20,30,40,50> END-DEFINE * MOVE #ARRAY(2:4) TO #ARRAY(3:5) /* is identical to /* MOVE #ARRAY(2) TO #ARRAY(3) /* MOVE #ARRAY(3) TO #ARRAY(4) /* MOVE #ARRAY(4) TO #ARRAY(5) /* /* #ARRAY contains 10,20,20,20,20
If a single occurrence is assigned to an array range, each element of the range is filled with the value of the single occurrence. (For a mathematical function, each element of the range is filled with the result of the function.)
Before an assignment operation is executed, the individual dimensions of the arrays involved are compared with one another to check if they meet one of the conditions listed below. The dimensions are compared independently of one another; that is, the 1st dimension of the one array is compared with the 1st dimension of the other array, the 2nd dimension of the one array is compared with the 2nd dimension of the other array, and the 3rd dimension of the one array is compared with the 3rd dimension of the other array.
The assignment of values from one array to another is only allowed under one of the following conditions:
The number of occurrences is the same for both dimensions compared.
The number of occurrences is indefinite for both dimensions compared.
The dimension that is assigned to another dimension consists of a single occurrence.
The following program shows which array assignment operations are possible.
DEFINE DATA LOCAL 1 A1 (N1/1:8) 1 B1 (N1/1:8) 1 A2 (N1/1:8,1:8) 1 B2 (N1/1:8,1:8) 1 A3 (N1/1:8,1:8,1:8) 1 I (I2) INIT <4> 1 J (I2) INIT <8> 1 K (I2) CONST <8> END-DEFINE * COMPUTE A1(1:3) = B1(6:8) /* allowed COMPUTE A1(1:I) = B1(1:I) /* allowed COMPUTE A1(*) = B1(1:8) /* allowed COMPUTE A1(2:3) = B1(I:I+1) /* allowed COMPUTE A1(1) = B1(I) /* allowed COMPUTE A1(1:I) = B1(3) /* allowed COMPUTE A1(I:J) = B1(I+2) /* allowed COMPUTE A1(1:I) = B1(5:J) /* allowed COMPUTE A1(1:I) = B1(2) /* allowed COMPUTE A1(1:2) = B1(1:J) /* NOT ALLOWED (NAT0631) COMPUTE A1(*) = B1(1:J) /* NOT ALLOWED (NAT0631) COMPUTE A1(*) = B1(1:K) /* allowed COMPUTE A1(1:J) = B1(1:K) /* NOT ALLOWED (NAT0631) * COMPUTE A1(*) = B2(1,*) /* allowed COMPUTE A1(1:3) = B2(1,I:I+2) /* allowed COMPUTE A1(1:3) = B2(1:3,1) /* NOT ALLOWED (NAT0631) * COMPUTE A2(1,1:3) = B1(6:8) /* allowed COMPUTE A2(*,1:I) = B1(5:J) /* allowed COMPUTE A2(*,1) = B1(*) /* NOT ALLOWED (NAT0631) COMPUTE A2(1:I,1) = B1(1:J) /* NOT ALLOWED (NAT0631) COMPUTE A2(1:I,1:J) = B1(1:J) /* allowed * COMPUTE A2(1,I) = B2(1,1) /* allowed COMPUTE A2(1:I,1) = B2(1:I,2) /* allowed COMPUTE A2(1:2,1:8) = B2(I:I+1,*) /* allowed * COMPUTE A3(1,1,1:I) = B1(1) /* allowed COMPUTE A3(1,1,1:J) = B1(*) /* NOT ALLOWED (NAT0631) COMPUTE A3(1,1,1:I) = B1(1:I) /* allowed COMPUTE A3(1,1:2,1:I) = B2(1,1:I) /* allowed COMPUTE A3(1,1,1:I) = B2(1:2,1:I) /* NOT ALLOWED (NAT0631) END
Generally, the following applies: if arrays with multiple dimensions are compared, the individual dimensions are handled independently of one another; that is, the 1st dimension of the one array is compared with the 1st dimension of the other array, the 2nd dimension of the one array is compared with the 2nd dimension of the other array, and the 3rd dimension of the one array is compared with the 3rd dimension of the other array.
The comparison of two array dimensions is only allowed under one of the following conditions:
The array dimensions compared with one another have the same number of occurrences.
The array dimensions compared with one another have an indefinite number of occurrences.
All array dimensions of one of the arrays involved are single occurrences.
The following program shows which array comparison operations are possible:
DEFINE DATA LOCAL 1 A3 (N1/1:8,1:8,1:8) 1 A2 (N1/1:8,1:8) 1 A1 (N1/1:8) 1 I (I2) INIT <4> 1 J (I2) INIT <8> 1 K (I2) CONST <8> END-DEFINE * IF A2(1,1) = A1(1) THEN IGNORE END-IF /* allowed IF A2(1,1) = A1(I) THEN IGNORE END-IF /* allowed IF A2(1,*) = A1(1) THEN IGNORE END-IF /* allowed IF A2(1,*) = A1(I) THEN IGNORE END-IF /* allowed IF A2(1,*) = A1(*) THEN IGNORE END-IF /* allowed IF A2(1,*) = A1(I -3:I+4) THEN IGNORE END-IF /* allowed IF A2(1,5:J) = A1(1:I) THEN IGNORE END-IF /* allowed IF A2(1,*) = A1(1:I) THEN IGNORE END-IF /* NOT ALLOWED(NAT0629) IF A2(1,*) = A1(1:K) THEN IGNORE END-IF /* allowed * IF A2(1,1) = A2(1,1) THEN IGNORE END-IF /* allowed IF A2(1,1) = A2(1,I) THEN IGNORE END-IF /* allowed IF A2(1,*) = A2(1,1:8) THEN IGNORE END-IF /* allowed IF A2(1,*) = A2(I,I -3:I+4) THEN IGNORE END-IF /* allowed IF A2(1,1:I) = A2(1,I+1:J) THEN IGNORE END-IF /* allowed IF A2(1,1:I) = A2(1,I:I+1) THEN IGNORE END-IF /* NOT ALLOWED(NAT0629) IF A2(*,1) = A2(1,*) THEN IGNORE END-IF /* NOT ALLOWED(NAT0629) IF A2(1,1:I) = A1(2,1:K) THEN IGNORE END-IF /* NOT ALLOWED(NAT0629) * IF A3(1,1,*) = A2(1,*) THEN IGNORE END-IF /* allowed IF A3(1,1,*) = A2(1,I -3:I+4) THEN IGNORE END-IF /* allowed IF A3(1,*,I:J) = A2(*,1:I+1) THEN IGNORE END-IF /* allowed IF A3(1,*,I:J) = A2(*,I:J) THEN IGNORE END-IF /* allowed END
When you compare two array ranges, note that the following two expressions lead to different results:
#ARRAY1(*) NOT EQUAL #ARRAY2(*) NOT #ARRAY1(*) = #ARRAY2(*)
Example:
IF #ARRAY1(1:2) NOT EQUAL #ARRAY2(1:2)
This is equivalent to:
IF (#ARRAY1(1) NOT EQUAL #ARRAY2(1)) AND (#ARRAY1(2) NOT EQUAL #ARRAY2(2))
Condition A is therefore true if the first occurrence of
#ARRAY1 does not equal the first occurrence of
#ARRAY2 and the second occurrence of #ARRAY1
does not equal the second occurrence of #ARRAY2.
IF NOT #ARRAY1(1:2) = #ARRAY2(1:2)
This is equivalent to:
IF NOT (#ARRAY1(1)= #ARRAY2(1) AND #ARRAY1(2) = #ARRAY2(2))
This in turn is equivalent to:
IF (#ARRAY1(1) NOT EQUAL #ARRAY2(1)) OR (#ARRAY1(2) NOT EQUAL #ARRAY2(2))
Condition B is therefore true if either the first occurrence
of #ARRAY1 does not equal the first occurrence of
#ARRAY2 or the second occurrence of #ARRAY1
does not equal the second occurrence of #ARRAY2.
A general rule about arithmetic operations with arrays is that the number of occurrences of the corresponding dimensions must be equal.
The following illustrates this rule:
#c(2:3,2:4) := #a(3:4,1:3) + #b(3:5)
In other words:
| Array | Dimension Number | Number of Occurrences | Range |
|---|---|---|---|
| #c | 2nd | 2 | 2:3 |
| #c | 1st | 3 | 2:4 |
| #a | 2nd | 2 | 3:4 |
| #a | 1st | 3 | 1:3 |
| #b | 1st | 3 | 3:5 |
The operation is performed element by element.
Note:
An arithmetic operation of a different number of dimensions is
allowed.
For the example above, the following operations are executed:
#c(2,2) := #a(3,1) + #b(3) #c(2,3) := #a(3,2) + #b(4) #c(2,4) := #a(3,3) + #b(5) #c(3,2) := #a(4,1) + #b(3) #c(3,3) := #a(4,2) + #b(4) #c(3,4) := #a(4,3) + #b(5)
Below is a list of examples of how array ranges may be used in the
following ways in arithmetic operations (in COMPUTE,
ADD or
MULTIPLY
statements). In examples 1-4, the number of occurrences of the corresponding
dimensions must be equal.
range + range = range.
The addition is performed element by element.
range * range = range.
The multiplication is performed element by element.
scalar + range = range.
The scalar is added to each element of the range.
range * scalar = range.
Each element of the range is multiplied by the scalar.
range + scalar = scalar.
Each element of the range is added to the scalar and the result is assigned to the scalar.
scalar * range = scalar2.
The scalar is multiplied by each element of the array and the result is assigned to scalar2.
Since intermediate results will be generated for arithmetic operations as shown in the above examples, the result of overlapping index ranges is computed element by element in an intermediate result array and finally the intermediate result array is assigned to the result field.
Example:
DEFINE DATA LOCAL 1 #ARRAY(I4/1:5) INIT <10,20,30,40,50> END-DEFINE #ARRAY(3:5) := #ARRAY(2:4) + 1 /* A temporary array for the /* intermediate result values is /* generated implicitly: #temp(1:3). /* The following operations are /* performed internally: /* #temp(1) := #ARRAY(2) + 1 /* #temp(2) := #ARRAY(3) + 1 /* #temp(3) := #ARRAY(4) + 1 /* #ARRAY(3:5) := #temp(1:3) /* /* #ARRAY contains 10,20,21,31,41