a few minor changes proposal

[John W. Eaton]

On 25-Nov-2009, Jaroslav Hajek wrote:

| hi all,
| 
| I propose the following changes:
| 1. Deprecate cellidx. This seems to be a leftover from the old control
| functions, that was moved to general probably because it was used elsewhere.
| Alas, it is not used anywhere now. It even talks about signals, not strings,
| and the name is not descriptive enough to suggest that it is a string
| operation (generally those are prefixed by str...). The same goal is
| achievable using ismember, and more efficiently because the cellidx
| implementation just uses a double loop.
| 
| 2. change the behaviour of issquare (x) to a simple ndims (x) == 2 && rows
| (x) == columns (x). Currently, issquare ([]) is false, which I think is damn
| confusing and unexpected.
| 
| 3. Move issymmetric and isdefinite to linear-algebra. Contrary to its name
| and documentation, issymmetric actually checks for hermitianness, so create
| also ishermitian to clean up. Also, I think the default tolerance should be
| zero (check for exact symmetry), boiling down to the simple test: all ((x ==
| x.')(:)). And finally, like for the issquare method, these should be made
| consistent for 0x0 matrix.
| 
| patches attached.
| 
| any objections/comments?

These changes look OK to me.

Thanks,

jwe