Re: Removing disadvantages on struct array and cell struct array?

[Olaf Till]

On Sun, Jan 09, 2011 at 04:27:48PM -0500, John W. Eaton wrote:
Content-Description: message body text
> On  9-Jan-2011, veryhappy wrote:
> 
> | I knew about struct_levels_to_print but for some reason that i can't
> | understand struct arrays only show the name of the fields (they show the
> | full information if you specify a concrete element of the array) :S
> | Thanks for your reply i'll try to make it clearer giving you an example.
> | If execute a script with the following content:
> |   % Struct array example
> |   clear circuit
> |   struct_levels_to_print(10);
> |   component(1)=struct("type",'r',"value",1e4);
> |   component(2)=struct("type",'c',"value",1e-7);
> |   component(3)=struct("type",'r',"value",2.2e3);
> |   component
> |   circuit(1).component=component;
> |   circuit(2).component=component;
> |   circuit
> | I get the next results:
> | component =
> | {
> |   1x3 struct array containing the fields:
> | 
> |     type
> |     value
> | }
> | 
> | circuit =
> | {
> |   1x2 struct array containing the fields:
> | 
> |     component
> | }
> | 
> | As you see i don't get the contents of the struct but using struct arrays i
> | can do things like
> | [circuit(1).component.print_value]={'10K','100nF','2K2'}{:}. I would like to
> | retain that functionality and get something like:
> 
> I think old versions of Octave used to print the struct array
> contents, but then people didn't like that because it tended to spew a
> lot of info.  But I would consider making it possible again, but off
> by default.  For example, with the attached changeset, the development
> version of Octave will do the following for your struct array:
> 
>   octave:10> print_struct_array_contents (true);
>   octave:11> component
>   component =
>   {
>     type =
> 
>     {
>       [1,1] = r
>       [1,2] = c
>       [1,3] = r
>     }
> 
>     value =
> 
>     {
>       [1,1] =  10000
>       [1,2] =  1.0000e-07
>       [1,3] =  2200
>     }
> 
>   }
> 
> But maybe you want the order to be rearranged as follows
> 
>   component =
>   {
>     (1,1) =
>     {
>       type = r
>       value = 10000
>     }
>     (1,2) =
>     {
>       type = c
>       value = 1.000e-07
>     }
>     (1,3) =
>     {
>       type = r
>       value = 2200
>     }
>   }
> 
> or similar?  I'm not sure how to make it clear that the numbers in
> parens correspond the elements of the structure array.  Because of
> that, and because I'm not sure which output format is best, I hesitate
> to check in this change.
> 
> jwe

I found the behavior of former Octave versions very handy and
regretted the change to restricting output information (and suspected
Matlab compatibility as reason); so I also would welcome a change to
reenable something like the old behavior.

I like the first variant of output above better, because I found it
quite practical in the past. And if someone wants the second variant,
he can equivalently achieve it with num2cell().

Just a suggestion to make more clear what the numbers correspond to:
what about leaving off the braces? So it would look like:

  component =
  {
    type =

      [1,1] = r
      [1,2] = c
      [1,3] = r
   ...

This would avoid the misunderstanding with

  component =
  {
    type =

    {
      [1,1] = r
      [1,2] = c
      [1,3] = r
    }
   ...

that the field "type" contains the cell-array {"r", "c", "r"}.

Olaf