[Kst] Lines and Arrows

[Barth Netterfield]

This is pure option #3.

On February 4, 2006 05:38 am, George Staikos wrote:
> Bugs 121068 and 121108 really are the same problem resulting in different
> symptoms.  On a micro-level they look completely different and are fixed
> differently, but in the grand scheme of things I think they're just a
> mis-implementation of the view object interface.  A view object is
> specified by its properties, its paint, and its geometry.  The paint and
> the geometry are entirely separate things, and the paint must be contained
> within the geometry.  The geometry is something used at the program level,
> though in many cases it aligns directly with something at the
> user-interface level.  In the case of plots, labels, and boxes, the
> geometry and the bounds of the object should be identical.  However, in the
> case of lines, arrows, and ellipses, this is not necessarily the case. 
> This does not mean that the geometry is meaningless, nor does it mean that
> the object can paint outside the geometry.  The problem with line (and
> arrow) is that it has "from" and "to" properties, and those properties are
> presently assumed to be two opposite corners of the bounding box.  This
> works for a line of thickness = 1 (or 0), but nothing else.  In other cases
> the line is drawn outside the bounding box.
>
> I think the correct thing is that lines should compute 'from' and 'to'
> based on the geometry of the object, and not store them.  When the user
> tries to adjust a line by setting one of those two properties, it should
> compute the bounding box/geometry required for the line, clip it to the
> parent/view/whatever we want), and then adjust the 'from' or 'to' property
> accordingly.  Basically it should do the opposite of what I did originally,
> which is to draw the largest possible line inside the bounding box.  This
> new idea gives the following properties:
>
> 1) 'from' and 'to' are exactly that and not just "close" for large widths.
> 2) painting is always inside the geometry
> 3) lines can still be manipulated as expected
> 4) lines of all widths work
>
> One interesting side effect: making a line thicker will either make the
> object geometry larger or shorten the line - or both.

Geometry larger.

> How does this apply to 121108?  The manipulation routines need to behave
> accordingly:  lines are on a from-to basis, not a bounding box basis like
> other objects.  they have a bounding box, but it doesn't entirely make
> sense from a UI point of view.  parenting needs to change in different ways
> in this case.
>
> I think this maps up to Barth's options #1 and #3 combined.