Creating and accessing an IML List of Lists

[rbettinger]

MATLAB has structure syntax like

a.b( index ).c = d ;

where index is an integer.

How do I create an IML list equivalent to the MATLAB statement? This will be a hierarchical heterogeneous list, I think.

Given the first instance, e.g.,

a.b(1).c = d ;

if index is incremented, how do I tell IML to create the second instance,

a.b(2).c = e;

so that an array of sublists is created?

[Rick_SAS]

Please post a SAS/IML program that shows what you are trying to accomplish. Use the IML syntax, not the syntax from a different language.

As a general rule, use ListAddItem to add a new item to an existing list. If the item you add is a list, you can use ListAddItem to add items to that list, thus emulating a hierarchical structure.

[IanWakeling]

Hi Rick,

Can you give an example of how you can use ListAddItem to extend a sublist?  I can make it work at the top level, but not lower down in the hierarchy.

[Rick_SAS]

Please show what you have tried.

Keep in mind that MATLAB data structures are more general than lists. The main purpose of the SAS/IML lists is to pack objects of various types into a single object that can be passed to and from modules. You might be able to coerce lists to emulate other data structures, but you will have to be clever. 

[IanWakeling]

Here is an example.  If I create a list with 3 items where the second item is itself a list - like this:

L = [[1:3], [i(2),i(3)], [4:6]];

Then I might want to add i(4) to the sublist.   I have tried syntax like:

Call ListAddItem(L$2, i(4));

but this, and using L[2] instead of L$2, both leave L unmodified. If by being clever you mean something like:

  L$2 = [ L$2 || [i(4)] ];

then this works but essentially I have just manually reconstructed L$2.  It feels like there should be a subroutine called ListAddSubItem to do what I want, but it is not there.

[rbettinger]

I thank all of you for your interest in my question and for your suggestions.

I am trying to create a SAS/IML list with the following structure:

a$1$c = 'item 1'

a$2$c = 'item 2'

a$3$c = 'item 3'

b$1$c = 'item 1'

b$2$c = 'item 2'

b$3$c = 'item 3'

b$4$c = 'item 4'

...

After studying Dr. Wicklin's example, I think that this code does the task:

-----------

proc iml ; package load listutil ;
L = [] ;
do i = 1 to 2 ;
   call ListAddItem( L, [ #type='input', i, #item='item' + char(i) ] ) ;
end ;
do i = 1 to 3 ;
   call ListAddItem( L, [ #type='output', i, #item='item' + char(i) ] ) ;
end ;  
call struct( L ) ;
quit ;

----------

How can I produce the same result using the list syntax "$" instead of "ListAddItem" ?

Ross

[Rick_SAS · Accepted solution]

Not sure what people are asking, but maybe study this program and see if it helps to answer your questions. At least it provides a concrete example that we can discuss.

proc iml;
package load ListUtil;

/* create list of lists. 
   Specify L$1$1, L$1$2, L$2$1, and L$2$2 */
L = ListCreate(3);
L$1 = ListCreate(2);
L$1$1 = 'L11';
L$1$2 = 12;
L$2 = ListCreate(2);
L$2$1 = 'L21';
L$2$2 = 22;

/* Now add a third item to L$2 and access it as L$2$3 */
newVal = 23;
L2 = ListGetSubItem(L, 2, 'm');     /* get the sublist you want to expand */
call ListAddItem(L2, newVal);       /* expand it */
call ListSetSubItem(L, 2, L2, 'm'); /* set it back into the list */
call struct(L);

check = L$2$3;     /* make sure the new item exists */
print check;

[IanWakeling]

Thank you Rick, I see how to do it now.  So following your example I could make my own version of ListAddSubItem as follows:

proc iml;

start ListAddSubItem(A, Pos, Val);
  if type(A) ^= 'L' then return;
  B = ListGetSubItem(A, Pos, 'm');
  if type(B) = 'L' then call ListAddItem(B, Val);
  call ListSetSubItem(A, Pos, B, 'm');
finish;

package load listutil;

L = [ 2, [5, 6], 9 ];
call ListAddSubItem(L, 2, 7);
call struct(L);
quit;

[Rick_SAS]

> How can I produce the same result using the list syntax "$" instead of "ListAddItem" ?

You can't, unless you use concatenation as Ian showed earlier. The '$' syntax refers to an item that already exists. The ListAddItem function creates a new item in the list. 

Think of the $ operator as analogous to subscripts for matrices. If you define
m = 11:13;

you cannot reference m[4] because that element does not exist. If you want to add a new item to the m matrix, you need to use concatenation or "allocate and copy":

m1 = m || {14};
or

m2 = j(1, ncol(m)+1, .);
m2[1:3] = m;
m2[4] = 14;