[QUANTITY] My Summary of the "arguments"/transform discussion/thread

[David Berry]

Brian,

> Well, I guess "Ed" means me too :) Here's what I have observed (summary)
> of the discussion of multi-dimensional quantities/transforms:

I agree with your summary, except for the points noted below.

> b. Multi-dimensional quantities have axes. Axes are described as a set of
> values with associated units, accuracy.

I am happy with the first sentence, but I think the second sentence comes
into the area where our models disagree. In your model, axes are indeed
described as a set of values, and so you choose to store them in a nested
argument Quantity (the natural choice for storing a set of numerical
values). But in our model, the axes are described purely by an algorithm
(i.e. a mapping into a specified frame) and so we choose a FrameSet
rather than a Quantity. If I understand things correctly, your model
supplies a set of numerical values to the client saying "here are the
axis values which you can use (i.e. the pixel centres)", whereas out model
says to the client "tell me what pixel coordinates you are interested in
and I shall tell you the WCS position" (or vice-versa). The point is that
our model imposes no restrictions on what positions the client may be
interested in - it can have the pixel centres, or the pixel corners or
anywhere else.

If I have understood things correctly, I think this is an important point
because it determines whether we can align two Quantities at a sub-pixel
level.

> h. Values in the quantity or on the axes can either be actual numbers, or
> generated by an algorithm. And, I believe, we agree that this algorithm is
> likely to be described by a frameset, or something similar, that operates
> on the underlying "list index" which is a simple index of the values in
> the quantity/axis.

Same points as above really. I think we disagree about what the FrameSet
"operates on". In your model it operates on the list indices to create a
fixed set of axis values, whereas in our model it operates on list indices
(fractional or integer) supplied by the client. Also, I'd add that the
axis information should be able to go the other way. That is, given an
arbitrary set of axis values, find the corresponding list indices (which
may be fractional).

Otherwise, we are in agreement.

David


----------------------------------------------------------------------
Dr David S. Berry    (dsb at ast.man.ac.uk)

STARLINK project		 |	Centre for Astrophysics
(http://www.starlink.ac.uk/)	 |	University of Central Lancashire
Rutherford Appleton Laboratory	 |	PRESTON
DIDCOT				 |	United Kingdom
United Kingdom			 |	PR1 2HE
OX11 0QX