Sketchup is Inacurrate???
-
I wouldn't say SketchUp is a digital machine. Rather, it runs on one. Obviously floats and doubles have limited accuracy and indeed can cause all sorts of problems, especially in geometry. There are libraries that can work with arbitrary precision though. Checkout http://gmplib.org/ and its quotient class. It has no theoretical limit in accuracy.
Also, even though you can't create an infinitely long edge in SketchUp, Ruby does support the concept of infinity. Try
1/0.to_f -
@alan fraser said:
This thread seems to be getting a little lost. .....he last few pages seem to be little more than bitching about the fact that it's not a NURBS modeller. Absolutely correct....it's not...so what?
Alan, this thread is going exactly where it needs to go. If we keep saying we'll settle for "good enough", nothing will ever be gained to the better.
As much as I appreciate your input I sure wish this thread could turn into something to access Sketchup's many shortcomings, elaborate on them and seek out how it could be done to the better.
Maybe, by chance this could be an inspiration for the developer and new owners of Sketchup.
If we, the users agree on certain issues, then maybe we are being taken serious and the Dev team sees it necessary to ally changes or improvements.
But as I said it is good to see issues from all sides. -
@noelwarr said:
... Obviously floats and doubles have limited accuracy and indeed can cause all sorts of problems, especially in geometry. .....
SU merges endpoints that are 0.0254mm (or less) apart. That is when within the same context.
I'm talking about a separation of about 0.167mm between circumferenses. It's no big deal when things (I mean buildings) get built. Far from that! But it's more than enough to cause errors in the model later on. You just need to know where to find them to fix them. Or much better, how to avoid them in the first place.True construction circles wouldn't make this program that much heavier. They would not even make learning it any more difficult. But they will help quite a bit where it comes to accurate modeling.
Whereas true circles will make SketchUp an entirely different program.
-
@alan fraser said:
A slight little factoid, stemming from my earlier comments about arcs beginning and ending with only half-segments:
I just thought it might be a useful little trick if you needed to butcher or intersect an arc for various reasons, yet still wanted whatever remained to comprise equal segments.
yes alan, i see what you're saying.. but, it still offers nothing in the way of solving anything.. you can't use that midpoint of an arc's segment for any type of measuring and you can't place an extrusion profile there because it's no really on the arc.. the lines drawn between the arc's vertices have nothing to do with the actual arc other than visual reference..
@wo3dan said:
Jeff, you may have misinterpreted my previous post, ("peace" stands).
I still think that the Offset tool does exactly what it is supposed to do, even with SU's arcs, (they are segmented) whether connected to edges or not. Changing the number of segments reproduces a different offset and a different mitter line at the end.I'll go over your example once again (in this thread on page 6 or so).
ok.. here is a different attempt at outlining the issue.. what i've done in this .skp, is manually draw a letter J (step-by-step.. real basic stuff and i know your skill level is way beyond this but i'm just trying to be clear in the example).. then, after i draw it manually, i do the same thing with the follow-me tool then show how the results are different..
the point of the drawing is that it shows two different results.. so either my manually drawn version is wrong -OR- the follow-me version is wrong.. they're not 'both wrong' and they're not 'both right'.. ONE is right - ONE is wrong.. you decide which one that is...
-
@noelwarr said:
I wouldn't say SketchUp is a digital machine. Rather, it runs on one. Obviously floats and doubles have limited accuracy and indeed can cause all sorts of problems, especially in geometry. There are libraries that can work with arbitrary precision though. Checkout http://gmplib.org/ and its quotient class. It has no theoretical limit in accuracy.
Also, even though you can't create an infinitely long edge in SketchUp, Ruby does support the concept of infinity. Try
1/0.to_fsee, i don't even care about any of that stuff (as in.. i don't even know what you're talking about
)i'm talking about what the user gets as a result from various tools... in the case of sketchup's accuracy, there are inaccuracies due to the way the geometry is being manipulated.. these aren't mathematical errors because "Ruby does support the concept of infinity" (or whatever).. they are programming errors..
if sketchup needs to round a vertex to one-billionth-of-a-hair in order for a face to form.. so be it.. that's fine..
but, that's not the problem here.. -
@unknownuser said:
"DesertRaven"Alan, this thread is going exactly where it needs to go. If we keep saying we'll settle for "good enough", nothing will ever be gained to the better.
Where have I ever indicated that I'd settle for 'good enough'? I have pointed out several times now the shortcomings of the Offset Tool...in both exterior and more especially on interior offsets. I have also mentioned that Follow Me leaves much to be desired. If it didn't, we wouldn't need rubies like Follow Me and keep to perform simple tasks. I've also said...several times...that the arc and circle tools need many more options other than (in the case of the circle) simple Circle from Center. The arc tool is reasonably decent in Layout. I see no reason why it can't operate the same way in SU.
As Wo3Dan points out (and TIG before him), the operation of Offset tool is not necessarily an inaccuracy (which is what this thread is about). It is, in fact entirely logical...it's just not a logic that we find particularly useful most of the time. Again, another mode of operation would be very welcome. I and others have also been long pressing for true construction curves (as opposed to true curves in the geometry) I believe that is also achievable.All these are reasonable and achievable requests for improvement. You, on the other hand seem to be basing almost all of your criticism on the fact that SU does not display true geometry curves. This isn't a matter of not being good enough, it's simply the nature of the beast. SketchUp is a polygon modeller; it's currency is straight lines. You can campaign for true curves from now till eternity...but you won't get them.
-
-
@unknownuser said:
ok.. ..... .. what i've done in this .skp, is manually draw a letter J (step-by-step.. real basic stuff and i know your skill level is way beyond this but i'm just trying to be clear in the example).. then, after i draw it manually, i do the same thing with the follow-me tool then show how the results are different..
....
the point of the drawing is that it shows two different results.. so either my manually drawn version is wrong -OR- the follow-me version is wrong.. they're not 'both wrong' and they're not 'both right'.. ONE is right - ONE is wrong.. you decide which one that is...Thanks for making it so easy for me. Luckily I don't need decide anything. For your manually drawn J isn't according to what 'Offset' is about. The meaning of offset in SU (and in other programs that I know)is offsetting edges, not vertices. You offset to the point of where you offset infinitly short edges 0.0000000000000mm and "beyond" (true curves), but always edges. Now you introduce a totally different meaning of the operation, even unknown in other applications, for all I know.
Doing it manually means that you first decide the number of segments > the rotation angle per segment > construct the inner segmented line ("arc") > offset all segments outwards > extend the outer segments to intersect adjacent segments, all forming the outer "arc" > also intersection of first segment with the long edge > last segment has a perpendicular end (due to no connected edges)
If you fit this one on SU's offset, then it will fit. -
@alan fraser said:
If you draw a default arc then chop a little off each end segment, it's still obviously a 12 segment arc...but with two of those segments shorter than the others. Now use Entity Info to change the resolution to...say...24 segments, then back again to 12.
You find that what you have now is a redrawn 12 segment arc with all its segments now all the same length...but neither end point actually at the stated radius, only the middle ones.
You can play with this to get some interesting insights into the way that SU deals with curves. Afraid I'm not on my work machine, so no pics - but here's a little run-down of an experiment you might like to try.
- Draw an arc with a nice grand curve - a half circle, say, with 12 segments.
- Bisect one of the segments near the middle - chop it right in half at the point where the 'radius' is smallest.
- Select one half of the chopped arc, and choose a small number of segments
- And a large number for the other half.
- Erase the stumps of the line that you used to chop the arc.
- You now have a single arc with very uneven segments.
- Look closely at the bit where the two sections join - eeek.
- See how far you can shrink it with the offset tool before it gets, err, odd.
Similarly, you can show that if you change the segment count to even up the segment length after chopping a curve mid-segment, you'll still have 'off-radius' end points - and if you explode the arc and measure, you'll find that the end segment is not quite the same length as all the others
Just a rather contrived illustration that the lines are being offset, not the arc's 'parameters', of course - I offer it more as a curiosity than a judgement of SU.
I do think, though, that Jeff made a good point about this being a problem of expectations. We may know that it is the lines that get offset, but once you decide to call something an 'arc' or a 'circle', it is not surprising that users expect them to exhibit the behaviour associated with those categories of objects.
The "line offset" tool is a very good, and accurate, "line offset" tool - what is really being requested is a different tool, not a more accurate one.
-
And Jeff, even the 'Follow Me' tool is consistant when applied on a series of connected coplanar edges. It's just a different, now 3D way of creating an offset. If you delete the third dimension's geometry, it will fit on what the offset does.
b.t.w. You could do your J with the arc tool in a few seconds.
The arc tool acceps (during the operation) changing the number of segments (s), the chord length ((units)) and the bulge ((units)) or the radius ((units)r). (Units) is optional, needed when not according to current settings.
So what IS the problem.
You want a tool that lets you "offset" a series of edges and "arcs" (in one go) where edges and "arcs" act differently in the same operation. This requires a different tool that spits the entities by property. It could be integrated but IS a special, not according to offset's rules, operation. I remember your previous example with the first "curve"-stud parallel to the last "edge"-stud and a consistant shell width of the bottom plate. Special cases! -
@wo3dan said:
For your manually drawn J isn't according to what 'Offset' is about. The meaning of offset in SU (and in other programs that I know)is offsetting edges, not vertices.
gerrit
it sounds as if you've learned geometry from sketchup... instead of in school (or where_ever) then bringing that knowledge into sketchup..(on a side note, this is a point where i really (really) wish jbacus or another su official* would comment on the topic.. like i've been implying, he knows exactly what i'm talking about.. he absolutely has to.. much of the thread is simply trying to point out that there is an error occurring and i think if someone more official than me gave a statement, then more people (everybody?) would be less likely to defend the app and re_go through some of the example files with a fresh mind to see and understand the error themselves..) (*or some other definitive source where this part of the discussion can be cleared up)
regarding "The meaning of offset in SU is offsetting edges, not vertices."..... i'm no programer but there are some in this thread (tt?) that could clear this up.. when moving geometry around etc, i'm willing to bet at the root, sketchup actually never moves edges around.. it moves vertices.. if it moves an line, it's in fact moving the two end points of that line..
if you draw an edge, you click on the start and endpoints of that line.. it's ALL about the vertices.. when sketchup offsets a square, it boils down to moving the cornerpoints around.. (maybe not the case with guidelines which are infinitive? but even then, points and not lines are being used to define it's placement)
and, believe it or not, the math behind measuring/transforming/etc the points defining an arc are different than the points defining straight edges.. in many ways, sketchup recognizes the real math of an arc (bullet list a few pages back outlining some of these ways).. it's not totally dumb and it doesn't always treat the points defining an arc in the same manner it treats the points defining a collection of edges.. but, sometimes, it does.. which is a pretty major error..
-
Jeff, I'll get back to you tomorrow. Disagreeing with you.
-
Gerrit is correct, Jeff...you haven't given SU a fair crack of the whip. In your hand-drawn version you have the final radial section running horizontally along the red axis. Yet in the prepared path for Follow Me, the final segment of the path is still at a slight angle. You can't possibly expect Follow Me to produce a squared-off final cross section under those circumstances.
In situations like that (or in more practical circumstances like a pipe bending through 90 degrees before disappearing into a wall) you have to add a final 'follow-through' segment to get Follow Me to terminate properly. In the case of the pipe and wall, that would be a segment perpendicular to the wall (which a 90 degree arc doesn't give you). In the case of the J, it's a final segment running along the Green axis. The length doesn't matter...because after you've extruded, you just Push/Pull it back to the intended end point. If you do it that way, the two versions seem to match exactly. I tried it...they do. -
@wo3dan said:
So what IS the problem.
for absolute starters.. i can't even do this in sketchup:
and that's a super basic example.. there are potentially hundreds of other things that need to happen after a step like this but this step breaks any possibility of doing much more afterwards.. it's crap geometry..
-
@alan fraser said:
Gerrit is correct, Jeff...you haven't given SU a fair crack of the whip. In your hand-drawn version you have the final radial section running horizontally along the red axis. Yet in the prepared path for Follow Me, the final segment of the path is still at a slight angle. You can't possibly expect Follow Me to produce a squared-off final cross section under those circumstances.
ok alan.. to turn the tables a bit..
let's see you draw that J using the follow-me tool.. you can arrange your profile anyway you wish. and you can draw the arc (and/or it's segments) in any way you wish..
now look at the results.. measurements matter here. (i mean, that's basically the entire premise of the discussion- accuracy)
your idea of a 'fix' does nothing other than correcting one area an putting error in a different place.
-
@unknownuser said:
...
gerrit
it sounds as if you've learned geometry from sketchup... instead of in school (or where_ever) then bringing that knowledge into sketchup..
......
Jeff, I've been selfemployed for about 26 years, created many constructions (drawings+calculations), mostly buildings (from small up to some up to 50m50m8m / factory+office etc, total package) with and without architects, and also steel constructions. Some big projects for one single person...., me. I'm not bragging about this. Point is, in those early days there was no sketchup, nor drafting programs for us poor souls. I had to solve many geometrical challenges with nothing but a calculator and what I know/had learned. Maybe that's why I can appreciate so much how SU works. Clean yet powerful. It performs how I think in 3D, but much faster. -
@wo3dan said:
And Jeff, even the 'Follow Me' tool is consistant when applied on a series of connected coplanar edges. It's just a different, now 3D way of creating an offset. If you delete the third dimension's geometry, it will fit on what the offset does.
youre exactly right.. ive been trying to say that over and over again in this thread..
the offset tool is used (by me at least) in the discussion because it's the least confusing to talk about.. but the same exact error that occurs in the offset tool occurs in the follow me tool (only it's now multiplied the amount of bad offsets)... and then- all those plugins i listed a few pages back do the same_exact_thing.. the offset error is like a virus and it infiltrates many other areas of the app..
sketchup does not handle arcs properly.. end of story.
-
@wo3dan said:
Point is, in those early days there was no sketchup, nor drafting programs for us poor souls. I had to solve many geometrical challenges with nothing but a calculator and what I know/had learned.
then dust off that old compass and drafting table.. draw the J.. measure it.. draw it in sketchup using any of the problem tools ive mentions.. it's not going to happen..
-
oh. and hey.. with those last jpgs i posted, delete the profile shape.. take followme out of it..
select the surface and offset inwardly the same distance as the corner arc's radius..what the heck is that? and if you must, use the smart offset plugin since it's smarter..
what the heck is that? -
@unknownuser said:
...and that's a super basic example.. there are potentially hundreds of other things that need to happen after a step like this but this step breaks any possibility of doing much more afterwards.. it's crap geometry..
Jeff, it's an example still according to "offset" rules. It may not look nice and needs fixing. But it's a cosistant way of offsetting an SU arc. What you are looking for is a special case as mentioned before: split edges and "arcs" in the selection prior to operating the offset, ignore the "arc's" first segment's direction, Offset edges perpendicular at the ends, draw the inner (or outer) "arc" with radius R-offset (resp. with R+offset) between offsetted (perpendicular) last edges's ends, to fill in the gaps.
(now it's time to Zzzzzzzzzzzzz / see you tomorrow)
Advertisement