transforming a row vector into a column vector (continued)

[Francois L.]

Hello,

I want to transform the row vector (1-D array) into a column vector.

Having a row vector a:
IDL> a = [1,2,3,4]

If I use the transpose command:
IDL> b = transpose(a)

I get the following:
IDL> print, size(a)
1 4 2 4
IDL> print, size(b)
2 1 4 2 4

With the transpose command b becomes a 2-D array.

How to preserve to one dimension aspect ?

By the way, is there a possibility for knowing what type (row or colum) is a
vector ?

Thanks,

Francois.

[Peter Albert]

Hi Francois,

well, the answer is already in your post :-)

What does the output of size(b) actuall mean?
The first "2" tells us that b is 2-dimensional.
The following "1" tells us that the number of columns (the x-dimension)
is 1.
The following "4" tells us that the number of rows (the y-dimension) is
4.

Voila, that's the definition of a column vector, isn't it? It has one
column and n rows.

IDL does not know different types of vectors, a vector in its
1-dimensional form always is a row vector.

As for your second question, again the answer is in your post: The
"possibility for knowing what type an array is" is the SIZE() command:

case 1 of
(size(array))[0] eq 1:print, "This array is a row vector"
(size(array))[0] eq 2: print, (size(array))[1] eq 1 $
? "This array is a column vector" $
: "This array really is a 2D array"
else: print, "This array has even more dimensions"
endcase

Cheers,

Peter

[Mark Hadfield]

Post by Peter Albert
IDL does not know different types of vectors, a vector in its
1-dimensional form always is a row vector.

I would go further than that. IDL is not a matrix-oriented language
(unlike Matlab, which definitely is) and its fundamental data structures
are the scalar and the array, the latter with 1-7 dimensions. With
arrays it uses a Fortran-like relationship between indexing and storage
(ie inner dimension varies fastest in memory) and it also supports a
handy 1-D indexing scheme that makes use of this relationship.

That's about it. Basically, IDL doesn't know about rows or columns,
vectors or matrices, tensors or quaternions. You *can* simulate these
mathematical structures with arrays, and some built-in IDL functions and
operators will help you, but make sure you check the conventions about
the correspondence between the IDL concepts (arrays, dimensions,
indices) and the mathematical concepts (matrix, row, column).

And read:

http://www.dfanning.com/misc_tips/colrow_major.html

PS: I have used IDL successfully to solve SVD problems, but I had to
read the documentation carefully and check out my understanding with toy
examples.

PS2: To my mind, IDL's agnosticism about matrices is vastly preferable
to Matlab's "everything is a double-precision matrix" stance.