Re: help with graphics

[John W. Eaton]

On 26-Oct-2011, Michael Goffioul wrote:

| On Wed, Oct 26, 2011 at 8:15 AM, John W. Eaton <address@hidden> wrote:
| > +static bool
| > +all_negative (const MArray<double>& m)
| > +{
| > +  bool retval = true;
| > +  const double *data = m.data ();
| > +  for (octave_idx_type i = 0; i < m.numel (); i++)
| > +    {
| > +      if (data[i] >= 0)
| > +        {
| > +          retval = false;
| > +          break;
| > +        }
| > +    }
| > +  return retval;
| > +}
| 
| Wouldn't this better located in the MArray class, or more generally
| somewhere in liboctave?

Yes, that's why I said my solution was just a temporary hack to test
the idea and that I'd fix it later before committing.

| > @@ -248,6 +270,11 @@
| >       for (int i = 0; i < n; i++)
| >         dest[i] = log10(src[i]);
| >     }
| > +  void do_neg_scale (const double *src, double *dest, int n) const
| > +    {
| > +      for (int i = 0; i < n; i++)
| > +        dest[i] = log10(-src[i]);
| > +    }
| >  };
| >
| >  class scaler
| >
| >
| > The tick marks are displaced so that they are on top of the axis
| > lines.  Not sure yet why that is happening.  I'll take a look at it
| > later.  First I need to get some sleep...
| 
| This is probably because the new scaler flips the axis, so after
| scaling the min and max values are swapped. You might want to try to
| re-flips after log-scaling, like:
| 
|       dest[i] = -log10(-src[i]);

Nice, that seems to fix it.  The scaler class looks like a good
solution to this problem.  Thanks for these pointers.  I would have
screwed things up horribly if I had continued down the path I was on.
So it would probably help a lot if we had some comments in the
graphics code that explained the overall design and the intent/purpose
of some of the variables.  (I know, likewise for other parts of Octave
as well...)

jwe