Re: Problem array subscripting

[Henry Chapman <chapman9(at)llnl.gov>]

David Fanning wrote:
> > IDL> result = array[x, y, index, 0]
> 
> The trouble with this trick is that I don't understand
> how it works. It's fine for production code, don't get
> me wrong, where a little smoke and mirrors can even be
> elegant. I just don't want to get up in front of an IDL
> programming class and have to explain it. Do you have
> any theories?

Well, here's what's written in "Building IDL applications" on pp 62,63:

---- [begin quote from building.pdf] ---

"Extra" Dimensions

When creating arrays, IDL eliminates all size 1, or "degenerate",
trailing
dimensions. Thus, the statements

A = INTARR(10, 1)
HELP, A

print the following:

A INT = Array(10)

This removal of superfluous dimensions is usually convenient, but it can
cause problems when attempting to write fully general procedures and
functions. Therefore, IDL allows you to specify "extra" dimensions for
an array as long as the extra dimensions are all zero. For example,
consider a vector defined as follows:

ARR = INDGEN(10)

The following are all valid references to the sixth element of ARR:

X = ARR[5]
X = ARR[5, 0]
X = ARR[5, 0, 0, *, 0]

Thus, the automatic removal of degenerate trailing dimensions does not
cause problems for routines that attempt to access the resulting array.

------- [end quote] ---

So, IDL says that it's all Kosher to put in the extra dimension.

But, I guess, that doesn't quite answer all of it, since the above
doesn't address the behaviour of subscripting.  Let me quote some more
from building.pdf (p 68):

---- [start quote]

When combining two subscript arrays, each element of the first array is
combined with the corresponding element of the other subscript array.
The two
subscript arrays must have the same number of elements. The resulting
subscript
array has the same number of elements as its constituents. For example,
the
expression A[[1, 3], [5, 9]] yields the elements A[1,5] and A[3,9].

---- [end quote]

That is, when you try and do

result = array[x, y]

IDL wants to return result[0] = array[x[0], y[0]], ....
result[i] = array[x[i], y[i]], ...

And this may well be what you want.  So how does one switch between
these behaviours?  There is a hint on p. 67 of the same good book:

-- [start quote]

When combining an array subscript with a subscript range, the result is
an array of subscripts constructed by combining each element of the
subscript array with each member of the subscript range. Combining an
n-element array with an m-element subscript range yields an nm-element
subscript. Each dimension of the result is equal to the number of
elements in the corresponding subscript array or range.

For example, the expression A[[1, 3, 5], 7:9] is a nine-element, 3 ´x 3
array composed of the following elements:

-- [end quote]


So, I think my trick is consistant with IDL's documentation, although it
might not have been what IDL had in mind.  Here is my summary of array
subscripting:

There are two behaviours you may want when you try result = array[x,
y,..., z], and x, y, ..., z are 1-d arrays:

1.  return a 1-d array with result[i] = array[x[i], y[i],..., z[i]]
2.  return a 2-d array with result[i, j,..., k] = array[x[i],
y[j],...,z[k]]

IDL does (1), which requires x, y,...,z all have the same length. 
However, IDL does (2) if one of the dimensions is a subscript range. 
Since a single number is a subscript range, it will do (2) in that
case.  And since 0, written in as an extra dimension is also a subscript
range, IDL will use behaviour (2)

Henry.

-- 
Henry Chapman                          mailto:chapman9@llnl.gov
Information Science & Technology Program    phone:(925)423-1580
Lawrence Livermore National Lab               fax:(925)423-1488
L-395, 7000 East Ave., Livermore CA 94550