Sorry to jump in, I was reading your discussion.
I will vote for shape.
Order and rank have other meanings in linear mathematics
that may confuse some readers.
Rank (or sometimes essential rank) is also used
to denote the number of independent vectors in a matrix. In a 2D pdl
(matrix) this
can never exceed the smallest dimension. That means a 2D pdl 2 rows by
3 columns
can not have a rank higher than 2.
As for order. Order is used as proposed here, but also in matrix
decomposition.
For example the order of a singular (i.e. non invertable) matrix is
zero no matter
how many rows and columns it has.
Hope it helps. Louis
--
Please note my email adres has changed to [email protected]
Citeren Bryan Jurish <[email protected]>:
On 2012-01-05 14:36, David Mertens wrote:
Yes, rank. Come to think of it, though, I was thinking "rank" would be a
good word in place of "number of dimensions," as in a rank-two piddle,
rather than a two-dimensional piddle.
The issue arises with describing data. For example, a point in three
dimensional space is described by a one dimensional piddle with three
elements. As you can see, the word "dimension" is used in two ways, one
to describe the extent of space, the other to describe the shape of the
piddle. Using the word "rank," we would say that a point in three
dimensional space is described by a rank-one piddle with three elements.
A potential revised lexicon, then, would be this:
dims -> shape
ndims -> rank
... I find this use of "rank" confusing, since "rank" (and "order" and
"degree") seems to me more like words for what we've been calling the
_size_ of a given dimension (e.g. Rank_i($p) == $p->dim($i)). IIRC, the
usual theoretical term for the number of components of a vector space
(i.e. the "3" in "3-D space") is in fact "dimension" (e.g. Dimension($p)
== $p->ndims), but that's pretty much in direct contradiction with the
usual use of "dimension" in the PDL docs, where it refers to (the
projection of) an individual component of the vector space. Compare
http://en.wikipedia.org/wiki/Rank_%28linear_algebra%29 and
http://en.wikipedia.org/wiki/Dimension_%28linear_algebra%29
In order of specificity, we would talk about a...
3x5 piddle,
rank-two piddle
piddle
I would prefer the current "2-D piddle" to "rank-two piddle", for the
abovementioned reasons. I agree it would be nice to have a noun for
this property, so we could assert things about "all piddles of $THINGY
= 2" instead of "all N-D piddles with N >= 2" or "all (N>=2)-D
piddles", which are both a bit kludgy. Maybe "dimension-rank" ? Or
maybe it's possible to distinguish between "the Dimension of a piddle"
(ndims) and "the i^th dimension of a piddle" (dim($i)) ... although
that's cutting it pretty fine...
One final, very deep annoyance that I have is with the term "thread." It
should be "autoloop." But that should wait until after 2.4.10 at the
very least.
I agree, "thread" is misleading. Although it doesn't bother me much
(anymore), I would also vote to replace it with "autoloop" at some
unspecified point in the future.
marmosets,
Bryan
On Jan 5, 2012 4:48 AM, "Matthew Kenworthy"
<[email protected] <mailto:[email protected]>> wrote:
The "shape" proposal is an *excellent* idea - it gets my vote!
On Thu, Jan 5, 2012 at 5:17 AM, Craig DeForest
<[email protected] <mailto:[email protected]>> wrote:
Hrm. As long as we're on terminology, how do you describe a PDL
with dim list [3,5]?
We've/I've been calling it a 2-D PDL with dim sizes 3 and 5 (as
in "dim 0 has size 3, and dim 1 has size 5"), or alternatively a
3x5-PDL. Its first row would be called a 3-PDL or a 1-D PDL
with size 3.
My preference is "3x5 PDL" as "2d 3x5 PDL" is a bit redundant.
I'd say no to "3-PDL" - I'd prefer "1-D PDL with size 3".
"A PDL with shape 3 by 5" sounds good to me!
--
Bryan Jurish "There is *always* one more bug."
[email protected] -Lubarsky's Law of Cybernetic Entomology
_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
_______________________________________________
Perldl mailing list
[email protected]
http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
