Draw infinite lines?

thomthom

Given a 3d vector, does anyone have any idea of drawing an infinite line using the view.draw* methods?

fredo6

There is no way to do it directly (you would have to make advanced computation about the vie camera to make sure that the line looks 'infinite', with update for any change of view and type of perspective. So, unless you really want the line to appear as embedded within the model (i.e. being hidden by objects of the mdoel when behind, etc...), you should use view.draw2d instead with clipping on the viewport boundaries.

Fredo

TIG

Or temporarily set the model.bounds +1000m and then make the 'line' ~ that long, so you still see it - it doesn't really need to be 'infinite' - does it

thomthom

I just want to draw these infinite guide lines that SU's tools uses some times, like the rotation or protractor tool does.
So I'm trying to find a way for it to be in 3D so one can see the guidelines in 3d space. Otherwise I'd just use the draw2d methods.

fredo6

@thomthom said:

I just want to draw these infinite guide lines that SU's tools uses some times, like the rotation or protractor tool does.
So I'm trying to find a way for it to be in 3D so one can see the guidelines in 3d space. Otherwise I'd just use the draw2d methods.

Have a look at what is done in FredoScale protractor tool (for instance for the rotation tool). If this is what you need I can tell you then where is the corresponding code.

Fredo

thomthom

The Rotation (Free) tool?
Looks to be drawn in the view plane, 2d. ?

I see you've also remade SU's protractor, I guess this gizmo is something that could be make into a reusable code snippet. I've just reinvented it myself.

fredo6

@thomthom said:

The Rotation (Free) tool?
Looks to be drawn in the view plane, 2d. ?

Yes it is 2D (but in the plan of rotation), so the direction is 3D

@thomthom said:

I see you've also remade SU's protractor, I guess this gizmo is something that could be make into a reusable code snippet. I've just reinvented it myself.

I am afraid many scripters will end up rewriting their own ruby code for mimiquing the standrad SU tools (unless SU publishes some classes).

fredo

TIG

If you want 3D guides why not just use 3D guides [clines] made within the draw command and then erase them when they are no longer needed or on deactivate ? I use this in 2D tools with cpoints etc to let you snap to them on a line you have 'drawn' in temporary graphics but is not really there otherwise; they are collected into an array of entities and they are erased afterwards

thomthom

I want them for illustrative purposes, but I don't want my inputpoints to snap to them.
Too bad the API lacks this feature.

fredo6

@thomthom said:

Given a 3d vector, does anyone have any idea of drawing an infinite line using the view.draw* methods?

Tom,

Attached is a small code snippet (extracted and adapted from LibFredo6) which computes the 2D line simulating an infinite line in 3D.

Say you have 2 points in 3D space: pt1_3d and pt2_3d.

Based on the current view, you can compute the line in 2D representation
pt1, pt2 = infinite_line(view, pt1_3d, pt2_3d)

Then you just have to draw it
view.draw2d GL_LINE_STRIP, pt1, pt2

Of course, because the view can be altered by zooming and orbiting, you need

either to do the computation within the draw method of the Tool
compute it once and then only on each resume events

Fredo

Not a script - only code inside

Dan Rathbun

You have the first point ( pt1 ) a Geom::Point3d object.
How about using Geom::Point3d.offset like so: (not tested)
vec is some vector you set


len = model.bounds.diagonal * 2
pt2 = pt1.offset( vec, len )
xytl=view.corner(0) # topleft  an [x,y] Array
xybr=view.corner(3) # botright an [x,y] Array
xyp2=view.screen_coords(pt2) # a Point3d (ignore z value)
until not(xyp2.x.between?(xytl.x,xybr.x) and xyp2.y.between?(xytl.y,xybr.y))
  len += len  # double the length
  pt2 = pt1.offset( vec, len ) # get a new pt2
  xyp2=view.screen_coords(pt2) # get it's new coords
end
view.draw_polyline( pt1, pt2 )

@unknownuser said:

Of course, because the view can be altered by zooming and orbiting, you need

either to do the computation within the draw method of the Tool

compute it once and then only on each resume events
Same goes for my example. Recompute pt2 onResume.

thomthom

What if you are able to calculate the vanishing point?

fredo6

@thomthom said:

What if you are able to calculate the vanishing point?

This is a good idea.
This will allow also to use view.draw_line method and have the line be fully embedded in the 3D model (i.e. not visible when it crosses solids).

Of course, one need to compute this vanishing point based on the perspective view. In principle, offsetting by a distance of 1 million miles should make it!

Fredo

thomthom

Might still be an issue, you need to use the Tool's getExtent method to return a Boundingbox big enough. But if you do that you cause the OpenGL clipping.

I've been trying to search for how it's generally done, infinite lines, and I think that they are drawn with special instructions in order to create such projection...

Not sure though... but this infinite line thing is bugging me!

thomthom

@unknownuser said:

Of course, one need to compute this vanishing point based on the perspective view. In principle, offsetting by a distance of 1 million miles should make it!

And... what if the view is parallel...?

fredo6

@thomthom said:

@unknownuser said:

Of course, one need to compute this vanishing point based on the perspective view. In principle, offsetting by a distance of 1 million miles should make it!

And... what if the view is parallel...?

I mean if you have a point and a direction, I guess that by offsetting the point by a huge distance will make the line looks infinite. This should be independent of the view.

Fredo

PS: do you wish to create a physical line or just draw it in a view?

TIG

Two 'infinite' parallel lines that are in fact only a few km long will appear to have a common end-point in the far distance... simply because both of their end-points will use the same screen-pixel - so they'll look like they spring from a common point BUT are actually parallel and will not.
Isn't it pointless to worry about this infinity-ness, provided that they 'appear' infinitely long ?

thomthom

@tig said:

Isn't it pointless to worry about this infinity-ness, provided that they 'appear' infinitely long ?

Yes, that is what I try to do. Draw a line that appear to be infinite - that i ends in the SU horizon. Like the protractor tool does when you rotate. Or like an Guide Line does.

I was thinking that if you could calculate where the projection meet, you could then use that as a measurement of how long the line should be, provided you could work out what the screen point represented in world coordinates.