Re: cell-array display in class

[Oliver Heimlich]

On 29.04.2015 23:38, Colin Macdonald wrote:
> On 27/04/15 07:04, Oliver Heimlich wrote:
> > It is not so easy. First, you have to decide whether “argn” represents
> > coordinates of a cell-array. You could do this with a regular
> > expression, but can't be 100% sure.
>
> thanks for this `argn`!
>
> Cannot be 100% sure?  Do you have a particular counterexample?  This
> seems to work for me:

I wasn't sure how cell array works internally, see below.

> if (isempty(inputname(1)) && regexp(argn, '^\[\d+,\d+\]$'))
>    % we're in a cell array
>    ...
> else
>    % whatever I did before.
> end
>
> If my input was actually `display([1,2])` then it won't be my
> @class/display that gets called :)

Okay, this kind of false positive can be eliminated from the list.

> (Ok, needs to be a bit more complicated for all possible shapes of cell
> array but that regexp gives idea.)

I looked into the code of cell array:

                  std::ostringstream buf;
                  buf << "[" << i+1 << "," << j+1 << "]";

                  octave_value val = matrix(i,j);

                  val.print_with_name (os, buf.str ());

So, the regexp should be sufficient. Cell array of higher dimensions 
cannot be displayed with content.

> > Second, you want to get access to the following variable (in ov-base.cc)
> > to correctly indent your output. I think there currently is no way to
> > access this value from within a m-file display method (unless you use an
> > oct-file kludge).
> >     int octave_base_value::curr_print_indent_level
>
> Yes, that is unfortunate :(   I'm sure I have other things I should be
> worrying about but perhaps Octave could pass to us the current indent
> level as 2nd argument of display.  Thoughts?
>
> Colin

The method's signature should stay compatible with the core functions, 
which check for nargin==1. Is it a good idea to pass 
current_print_indent_level as a global variable to the display method?

Oliver