Document Number:     J4/02-0126

July 7, 2002 

 

Page    1  of   3

Subject: 

SET statement problems - indexes

Author: 

Don Schricker

References:

1. 

Base Document:  FDIS, April 2002

a.

8.4.1.2, Subscripts, general rules 1 and 2

b.

13.16.36, OCCURS clause, general rule 2

c.

14.8.35, SET statement, general rules 2 and 3

DISCUSSION:

Wim Ebbinkhuijsen reported the following problems with reference 1c:  

In GR 2.a.1 there is no reference to the situation where the evaluation of the arithmetic

expression does not result into a valid subscript.

In GR 2.a.3 there is no reference to the situation where the occurrence number of the table

element referenced by index-name-2 is too large to be a valid occurrence number for the table

referenced by index-name-1.

On the other hand, GR's 3.a and 3.b do cover these situations.

I think that GR 2 on page 89, 8.4.1.2 Subscripts, covers this. But I don't think that the GR's 2

and 3 are consistent

Initial response from Don Schricker:  

No exception would be raised by the SET statement here, but if you use the invalid index as a

subscript without using relative indexing, 8.4.1.2 specifies the exception condition that would

result.

 You are correct that GR 2.a.3 should specify an exception because the SET statement

could try to set the index to a value outside the range that can be physically stored in the index.

Note:  I see other holes, too.  Like SET does not allow an index to be set to zero or to even one

greater than the max.

Direction from Don Schricker for 02-0126 after giving the subject due consideration:

There are several problems with reference 1c:

the rules allow some situations that should result in runtime exceptions

the rules prohibit some situations that should be allowed.

I think that reference 1b, the OCCURS clause, provides the direction for what should be allowed in an

index as a result of the execution of a SET statement.  That is, any value that fits within the range that

an implementor has specified can be contained in the index.  Of course, when an index is used as a

subscript, it must reference, possibly with relative indexing, a valid occurrence of the table.

Rule 3b of the SET statement is too restrictive, because it requires the SET statement to set the index to

a valid occurrence number.  Admittedly, the rule has to be reworded entirely because it relies on the

subscript being valid to describe the effect of the statement.  When the index points to occurrence zero,

Document Number:     J4/02-0126

July 7, 2002 

 

Page    2  of   3

or the occurrence that is one larger than the actual size of the table, the description of the effect of the

SET statement needs to be changed.  There is a similar problem with SET statement GR 2a1.

On the other hand, SET statement GR 2a3b could set the value of an index outside the range of

allowable values, and there is nothing in the SET statement prohibiting this.  Perhaps nothing needs to

be said in the SET statement, because this situation is covered by GR2 of the OCCURS clause, but

then there needs to be a reference here to the OCCURS clause.

One problem with the rules as structured is that the SET statement refers to the rules for subscripting

just to avoid saying that the arithmetic expression must result in an integer or else the EC-BOUND-

SUBSCRIPT exception is set to exist.  This could be repeated and avoid reference to other restrictions

that apply when subscripting, but not when setting the index.

I have reworded general rules 2 and 3 of the SET statement in the changes needed section. This is how

they appear in the draft so you can compare the two versions:

2) The following occurs for each recurrence of index-name-1 or identifier-1. Each time, the value of the sending

operand is used as it was at the beginning of the execution of the statement. Item identification of the data item

referenced by identifier-1 is done immediately before the value of that data item is changed:

 

 

a) If index-name-1 is specified:

1. If arithmetic-expression-1 is specified, a subscript is determined by evaluating arithmetic-expression-1

as described in 8.4.1.2, Subscripts. If the evaluation of that subscript is successful, index-name-1 is set to

a value causing it to refer to the table element that corresponds in occurrence number to that subscript;

otherwise the execution of the SET statement is unsuccessful and the content of the receiving operand is

unchanged.

2. If identifier-2 is specified, the content of the data item referenced by identifier-2 is placed in the index

referenced by index-name-1 unchanged.

3. If index-name-2 is specified:

a. If index-name-2 is associated with the same table as index-name-1, the content of the index

referenced by index-name-2 is placed in the index referenced by index-name-1 unchanged.

b. Otherwise, index-name-1 is set to a value causing it to refer to the table element that corresponds in

occurrence number to the table element associated with index-name-2.

b) If identifier-1 references an index data item, the content of index-name-2 or of the data item referenced by

identifier-2 is placed in the index referenced by index-name-1 unchanged.

c) If identifier-1 references a numeric data item, that item is set to the occurrence number of the table element

referenced by index-name-2.

3) The following occurs for each recurrence of index-name-3. Each time, the value of arithmetic-expression-2 is

used as it was at the beginning of the execution of the statement::

a) If arithmetic-expression-2 does not evaluate to an integer, the EC-BOUND-SUBSCRIPT exception

condition is set to exist, the execution of the SET statement is unsuccessful, and the content of the receiving

operand is unchanged.

b) If the occurrence number represented by the content of index-name-3 incremented (UP BY) or

decremented (DOWN BY) the result of the evaluation of arithmetic-expression-2 is a valid subscript for the

table associated with index-name-3 according to the constraints described in 8.4.1.2, Subscripts, index-

name-3 is set to a value causing it to refer to the table element that corresponds in occurrence number to that

subscript; otherwise the EC-BOUND-SUBSCRIPT exception condition is set to exist, the execution of the SET

statement is unsuccessful, and the content of the receiving operand is unchanged.

Document Number:     J4/02-0126

July 7, 2002 

 

Page    3  of   3

Changes needed in standard:

1.  Change SET statement, general rules 2 and 3 to read as follows:  

[DAS:  I'm not sure whether we need the parts that I've highlighted with yellow.  They help to clarify what

has changed, but they may not be needed in the specificatin.]]

"2) The following occurs for each recurrence of index-name-1 or identifier-1. Each time, the value of the sending

operand is used as it was at the beginning of the execution of the statement. Item identification of the data item

referenced by identifier-1 is done immediately before the value of that data item is changed:

 

 

a) If index-name-1 is specified:

1. If arithmetic-expression-1 is specified, a value is determined by evaluating arithmetic-expression-1.  If

the evaluation of arithmetic-expression-1 does not result in an integer, the EC-BOUND-SUBSCRIPT

exception condition is set to exist, the execution of the SET statement is unsuccessful and the content of

the receiving operand is unchanged.  Further constraints on the value of the expression are specified in

general rule 2 of 13.16.36, OCCURS clause.  If the evaluation of the expression is successful, index-

name-1 is set to a value causing it to refer to the table element that corresponds in occurrence number to

that expression, even if that value is outside of the range of valid subscripts for the table.

2. If identifier-2 is specified, the content of the data item referenced by identifier-2 is placed in the index

referenced by index-name-1 unchanged.

3. If index-name-2 is specified:

a. If index-name-2 is associated with the same table as index-name-1, the content of the index

referenced by index-name-2 is placed in the index referenced by index-name-1 unchanged.

b. Otherwise, a value is used that corresponds to the occurrence number of the table element

associated with index-name-2.   Constraints on that value are specified in general rule 2 of 13.16.36,

OCCURS clause.  If these constraints are satisfied, index-name-1 is set to refer to the table element

that corresponds in occurrence number to that value, even if that value is outside of the range of valid

subscripts for the table.

b) If identifier-1 references an index data item, the content of index-name-2 or of the data item referenced by

identifier-2 is placed in the index referenced by index-name-1 unchanged.

c) If identifier-1 references a numeric data item, that item is set to the occurrence number of the table element

referenced by index-name-2.

3) The following occurs for each recurrence of index-name-3. Each time, the value of arithmetic-expression-2 is

used as it was at the beginning of the execution of the statement::

a) If arithmetic-expression-2 does not evaluate to an integer, the EC-BOUND-SUBSCRIPT exception

condition is set to exist, the execution of the SET statement is unsuccessful, and the content of the receiving

operand is unchanged.

b) If the occurrence number represented by the content of index-name-3 incremented (UP BY) or

decremented (DOWN BY) the result of the evaluation of arithmetic-expression-2 is a valid index for the table

associated with index-name-3 according to the constraints specified in general rule 2 of 13.16.13, OCCURS

clause, index-name-3 is set to a value causing it to refer to the table element that corresponds to that

occurrence number, even if that occurrence number is outside of the range of valid subscripts for the table.

"

Open Issues:

1.  See note in proposal item 1 about the portions in yellow.