I usually scale by -1 in the appropriate dimension for things like this

On Tue, 18 Feb 2020, 14:40 , lar3ry@sasktel.net wrote:

Duh, I did it wrong.
Thanks

On 2/18/2020 1:39 PM, lar3ry@sasktel.net wrote:

On 2/18/2020 12:01 PM, A. Craig West wrote:

The manual specifically warns against negative scales.

https://en.wikibooks.org/wiki/OpenSCAD_User_Manual/Transformations#scale

Note: Do not use negative scale values. Negative scale values appear
to work for previews, but they lead to unpredictable errors when
rendering through CGAL. Use the mirror() function instead.

And indeed, sticking a "mirror" on the front does the right thing.

res = 2;    // Levels per unit height
slope=0.2;  // Units out per vertical unit
size=20;    // Height of font
height=10;  // Height of extruded letters

levels = height * res;

mirror([0,0,1]) {
    for (i=[0:levels-1]) {
        linear_extrude(height=(height-i/res))
            offset((i/res)*slope)
            text("hello", size=size);
    }
}

Although the shape produced is the same, I'd prefer this to rotating the
text, for two reasons:

1)  The result is readable.  Yes, you'd then want to rotate it to put it
on the bottom, but you could also use it on the top or some other
surface, with more obvious transformations.

2)  text() yields a 2D object.  Rotating a 2D object around X or Y, so
that it comes out of the Z=0 plane, seems ... wrong.  It seems like it
yields an object that is in some sense no longer 2D.  Yes, it ends up
back at Z=0, but it seems like an incorrect operation to perform on a 2D
object.

Here's an example of using this pattern on all sides of a cube:

res = 2;    // Levels per unit height
slope=0.15;  // Units out per vertical unit
size=20;    // Height of font
height=10;  // Height of extruded letters

levels = height * res;

epsilon = 0.01;

module engrave(t) {    
    mirror([0,0,1]) {
        for (i=[0:levels-1]) {
            linear_extrude(height=(height-i/res))
                offset((i/res)*slope)
                text(t, size=size, halign="center", valign="center");
        }
    }
}

difference() {
    cube(60, center=true);
    translate([0,0,30+epsilon]) engrave("hello");
    translate([0,30+epsilon,0]) rotate([-90,0,0]) engrave("hello");
    translate([0,-(30+epsilon),0]) rotate([90,0,0]) engrave("hello");
    translate([30+epsilon,0,0]) rotate([0,90,0]) engrave("hello");
    translate([-(30+epsilon),0,0]) rotate([0,-90,0]) engrave("hello");
    translate([0,0,-(30+epsilon)]) rotate([180,0,0]) engrave("hello");
}

Note:  that was very slow to display in preview, and took almost 17
minutes to render.  I haven't tried to figure out exactly why.

JordanBrown wrote

This opinion has come up several times in different threads - and I never
understood why, or found an example, why it shouldn't work. That it is
mentioned in the manual, doesn't prove that it is correct. Look at the
following:

mirror([0,0,1])
cube(1);

scale([1,1,-1])
cube(1);

after F5 the cfg file reads like this:

multmatrix([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, -1, 0], [0, 0, 0, 1]]) {
cube(size = [1, 1, 1], center = false);
}
multmatrix([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, -1, 0], [0, 0, 0, 1]]) {
cube(size = [1, 1, 1], center = false);
}

You can also check the other dimensions.
If negative values for such basic operation like scale were "forbidden" for
some reason, OpenSCAD should give at least a warning - and not rely on a
dubious "best practice" hint in the manual.

--
Sent from: http://forum.openscad.org/

Doesn't scale with -1 turn the object inside out because it flips the
vertices without fixing the winding order that gets reversed by the flip?

On Wed, 19 Feb 2020 at 17:45, NateTG nate-openscadforum@pedantic.org
wrote:

On 2/19/2020 8:36 AM, Parkinbot wrote:

I agree.  I'm only reporting what it says.  Indeed, I'd expect a
negative scale to be exactly equivalent to a mirror, once you turn it
all into a transformation matrix.

The caution was added by an anonymous user in January 2017.  It's
entirely possible that they ran into some rendering problem that was
unrelated, and blamed the wrong piece.

https://en.wikibooks.org/w/index.php?title=OpenSCAD_User_Manual/Transformations&diff=prev&oldid=3173552

Perhaps one of the people who is familiar with the internals can comment.

On 2/19/2020 9:55 AM, nop head wrote:

That was my guess, but I didn't see any evidence of it when I flipped a
cube and looked at Thrown Together.  But I don't really understand that
aspect of the system.

Mirror is likely to have the same effect, though. Unless it explicitly
reverses the winding order, which would only be required depending on
exactly how it is mirrored

On Wed, 19 Feb 2020, 13:48 Jordan Brown, openscad@jordan.maileater.net
wrote: