Align method not supported?

R2005 Vlisp:

I am working on a program that takes a list of triangles that have been drawn in various ucs's and am using a transformation matrix to move them to the WCS with their corners at z zero. This part works fine.
Then I use the transformation matrix again to try to move them to their relative position with the other triangles. This works fairly well except when I calculate the base corner position from the last triangle for the next triangle, the accuracy is only within 0.0001 or so. You don't notice until you get 1000 triangles down the list, then it becomes apparent that the triangles overlap 0.01 to 0.08".
Is there some way to "align" my triangles using 2 points?
I check the methods of my 3D polyline triangles and don't see an align method although there seems to be a lot of other choices:
Methods supported:
; AppendVertex (1)
; ArrayPolar (3)
; ArrayRectangular (6)
; Copy ()
; Delete ()
; Explode ()
; GetBoundingBox (2)
; GetExtensionDictionary ()
; GetXData (3)
; Highlight (1)
; IntersectWith (2)
; Mirror (2)
; Mirror3D (3)
; Move (2)
; Rotate (2)
; Rotate3D (3)
; ScaleEntity (2)
; SetXData (2)
; TransformBy (1)
; Update ()

Thanks

Bill

 

Hi Bill,

I noticed the same thing in transformations, I think it's due to sin-cos
slight inconsistancies? Although I tend to doubt that, but it's the only
conclusion I can draw. Perhaps someone else knows why. Anyways, once
they're flat can't you simply move from the corner of one tri to the corner
of the other tri. now you have one point aligned, then determine the angle
between the next aligning corners and rotate them together?

Devin

 

Hi Devin,

Yes the angles seem to have a descrepancy.

I think too that the triangles may be changing slightly in geometry also when they are transformed. Maybe because of this. Not sure.
Yes I'll find a work around. Just checking my options.
They seem to transform fine to 0,0. So that would be my next step to "rotate/move/align" somehow.

Thanks

Bill

 

I just learned something new about the align command when you put it in a loop.

It errors out and gives a message that says you cannot nest it more than 4 deep!

Bill

 

Maybe you can add an error checking section that "reads" the triangle (size,
angles, etc) both before transforms and after transforms with a fine degree
of granularity so as to detect if and when they get modified (beyond your
allowable tolerances) - even if just during the debug process - until you're
sure you're getting reliable results (or maybe you're already doing that)

???just a thought

interesting threads you two have been generating, I've enjoyed lurking and
learning - or trying to - I also would like to understand matrix math better

Keep up the good work

Mark

when they are transformed. Maybe because of this. Not sure.

 

Maybe you can add an error checking section that "reads" the triangle (size,
angles, etc) both before transforms and after transforms with a fine degree
of granularity so as to detect if and when they get modified (beyond your
allowable tolerances) - even if just during the debug process - until you're
sure you're getting reliable results (or maybe you're already doing that)

???just a thought
<<<
Mark.
Thanks for the input.
Accuracy of the transformed triangles would be key to accurate flat pattern. But I have already generated some excitement with what I can show them here so far with this basic program.

Been a fun project all in all.

Bill

 

Read up on using external programs in the Developer's Guide.

--
R. Robert Bell


I just learned something new about the align command when you put it in a
loop.

It errors out and gives a message that says you cannot nest it more than 4
deep!

Bill

 

I did a check on two of the legs and two of the angles for each triangle. Once at origin, once at first transformation and once at final placement. No changes in angles or length of legs were detected.
So it must be a placement issue.
Next I'll pursue Roberts suggestion and research external commands.

Bill

 

I'm having a lottle trouble finding any info on the topic.

Can you be more topic specific?
I looked in both the Autolisp and ActiveX developers guides.

Thx

Bill

 

AutoLISP Reference, Externally Defined Commands, Align

--
R. Robert Bell


I'm having a lottle trouble finding any info on the topic.

Can you be more topic specific?
I looked in both the Autolisp and ActiveX developers guides.

Thx

Bill

 

Found it, thanks!

Bill

 

Interestingly enough,
After 132 times through the loop, the align command goes dead. After that, even if I type in "align" at the command prompt, It prompts me to select objects, so I select the triangle and hit enter and nothing else happens, It returns to the command prompt.

So much for align.

Bill

 

Found the answer in this NG.

You have to feed the align function a selection set.

Don't ask me why.


Bill

 

Devin,

I did find very small differences in the triangle leg lengths after transformations.
I re-read and measured the distance between the corner points of each triangle after each transformation.

Command: (rtos ob1xlen 2 22) ;at origin
"4.203056214440523"

Command: (rtos ob1xlen2 2 22) ;at 0,0,0
"4.203056214440524"

Command: (rtos ob1xlen3 2 22) ;at new placement
"4.203056214440545"

Command: (rtos ob1ylen 2 22) ;at origin
"1.190832206495775"

Command: (rtos ob1ylen2 2 22) ;at 0,0,0
"1.190832206495771"

Command: (rtos ob1ylen3 2 22) ;at new location
"1.190832206495798"

I know these numbers are very, very small, but it makes me wonder if this is why I cannot place them without a bit of mis-alignment?

Bill

 

Bill,

That's right. The why is because the align function expects a selection set. It
deletes the ss after each pass of the command.

On the other hand, if you feed it two objects, it creates a selection set. And when
the number of selection sets exceeds 128... dead-in-the-water.

Joe Burke

 

On the other hand, if you feed it two objects, it creates a selection set. And when
the number of selection sets exceeds 128... dead-in-the-water.
<<<

Kinda what I was thinking...
And it does it even when fed ONE ename.
selection set is the way to go.

Thanks

Bill

 

hi, my 2 cents advise

get datas from 3dpoly => ( (triangle_1_vertex_A, triangle_1_vertex_B,
triangle_1_vertex_C) (triangle_2_vertex_A, triangle_2_vertex_B,
triangle_2_vertex_C)........)
some 2D mathematics (1 translation and 1 rotation) )
drawn new entities.

Bruno Toniutti.

 

Bill,

Looking at the code I would have to say that most likely the error is due to
either non-planar or non-perpendicular points used in creating the matrix.
It probably occures in the 3pxa_matrix function. The function takes and
creates an origin, x, y and z direction points then converts the direction
points to unit vectors. I bet that the direction points are slightly off.
In order to create a more accurate matrix the points need to be precisely
perpendicular to each other on their respective planes from the origin.

When I get some free time I'll look into it. Perhaps you'll find a fix
sooner. Just leave the following portion of the code alone and focus on the
code above it that calcs the direction points...

(setq
xo (car org)
yo (cadr org)
zo (caddr org)
xx (car -x)
yx (cadr -x)
zx (caddr -x)
xy (car -y)
yy (cadr -y)
zy (caddr -y)
xz (car -z)
yz (cadr -z)
zz (caddr -z)
numx (sqrt (+ (sqr (- xx xo)) (sqr (- yx yo)) (sqr (- zx zo))))
numy (sqrt (+ (sqr (- xy xo)) (sqr (- yy yo)) (sqr (- zy zo))))
numz (sqrt (+ (sqr (- xz xo)) (sqr (- yz yo)) (sqr (- zz zo))))
v (vector (list 0.0 0.0 0.0) origin)
ucs_matrix
(list
(list (/ (- xx xo) numx) (/ (- xy xo) numy) (/ (- xz xo) numz) (car
v))
(list (/ (- yx yo) numx) (/ (- yy yo) numy) (/ (- yz yo) numz) (cadr
v))
(list (/ (- zx zo) numx) (/ (- zy zo) numy) (/ (- zz zo) numz) (caddr
v))
(list 0.0 0.0 0.0 1.0)
)
ucs_matrix_array (vlax-tmatrix ucs_matrix)
)
ucs_matrix

 

some 2D mathematics (1 translation and 1 rotation) )
drawn new entities.<<

Do you have the code that shows how to do this?


Bill

 

When I get some free time I'll look into it. <<<

Great, let me know if you find a fix.

I'll be exausting my options in the mean time.
code above it that calcs the direction points...<<<

The code above looks like the axial rotate part itself....hmmm.

Thanks

Bill