[OctDev] FW: Cell array support

[Julian A. de M. Ph.D <ju...@ma...>]

I've been watching patches & usage tips float past it seems forever now.
Do y'all not fancy we could compile these @SF in a way that's easier to
use than troving through the octave.org mailing list archives?

Another unsolicited, obvious, back-seat remark from
joolean


-----Original Message-----
From: oct...@be...
[mailto:oct...@be...]On Behalf Of Jianming
Sent: Thursday, May 17, 2001 6:47 AM
To: Paul Kienzle; octave-sources
Cc: John W. Eaton
Subject: Re: Cell array support


Hi,
    Attached are the new patches for cell array support with octave
(2.1.34). After experimenting with various implementations, I've taken mo=
st
of your advices regarding the following:
    1. Using () to access cell elements should not narrow down to the bas=
e
type automatically. This will reduce confusion. I.E, narrowing conversion=
 is
removed.
    2. Using ASSIGNANYOP in op-cell.cc

    Attached also is a sample session with the patch. Some comments on th=
e
(incomplete) patch:

    1. Some features with Paul mentioned in the previous email are missin=
g:
        a) {} producing multi-valued output
        b) {} receive multi-valued output,
        c) splice cell arrays into an argument list =E0 la Matlab:
    2. As shown in the sample session, assigning cell elements to an
undefined indexed variable will give an error. To overcome this, cell(0,0=
)
must be used first to "declare" the variable as a cell. There is no probl=
em
assigning cell elements to an undefined variable in it's entirety though.
    3. Increasing the size of a cell array will not give unassigned eleme=
nts
as empty matrices as Matlab does. Instead, the undefined elements will ha=
ve
undefined type until assigned a value.
    4. Removing elements from a cell is done using {} instead of []. This=
 is
simpler to implement. In fact, there is no need for any code for this. If=
 we
want to use [], I believe we have to delibrately check the type of the le=
ft
hand side first. I don't really like this hack.

    Any comments, improvements are welcomed. Thanks Paul for your comment=
s
in the previous emails. I hope this gets into 2.1.35 together with some o=
f
your patches for cell functions.

Regards,
Jianming