Setting curves to another spline type makes them shorter #86086

New Issue

Daniel Bystedt · 2021-02-28T15:41:34+01:00

Daniel Bystedt commented

2021-02-28 15:41:34 +01:00

System Information
Operating system: Windows-10-10.0.18362-SP0 64 Bits
Graphics card: GeForce GTX 1080/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 460.79

Blender Version
Broken: version: 2.92.0, branch: master, commit date: 2021-02-24 16:25, hash: 02948a2cab
Worked: (newest version of Blender that worked as expected)

Short description of error
I discussed this with Campbell during a meeting 2021-02-23. He suggested that I report it as a bug.

When setting curve spline type from nurbs to bezier, the curve becomes shorter. It also becomes shorter when I convert itfrom bezier to nurbs.

It is worth to note that when converting from bezier to nurbs, the curve can get back the length by changing the following settings in
object data properties > active spline

Screen grab of issue

convert curve spline makes it shorter.mp4

Exact steps for others to reproduce the error

Add curve > path
Go into curve edit mode
Subdivide the curve 2 times in order to get enough vertices for changing spline type to bezier
Change spline type from Nurbs to Bezier - observe that the curve becomes shorter.

Change spline type from Bezier to Nurbs - observe that the curve becomes shorter

**System Information** Operating system: Windows-10-10.0.18362-SP0 64 Bits Graphics card: GeForce GTX 1080/PCIe/SSE2 NVIDIA Corporation 4.5.0 NVIDIA 460.79 **Blender Version** Broken: version: 2.92.0, branch: master, commit date: 2021-02-24 16:25, hash: `02948a2cab` Worked: (newest version of Blender that worked as expected) **Short description of error** I discussed this with Campbell during a meeting 2021-02-23. He suggested that I report it as a bug. When setting curve spline type **from nurbs to bezier**, the curve becomes shorter. It also becomes shorter when I convert it**from bezier to nurbs**. It is worth to note that when converting from bezier to nurbs, the curve can get back the length by changing the following settings in object data properties > active spline ![image.png](https://archive.blender.org/developer/F9852332/image.png) --- Screen grab of issue [convert curve spline makes it shorter.mp4](https://archive.blender.org/developer/F9852336/convert_curve_spline_makes_it_shorter.mp4) **Exact steps for others to reproduce the error** - Add curve > path - Go into curve edit mode - Subdivide the curve 2 times in order to get enough vertices for changing spline type to bezier - Change spline type from Nurbs to Bezier - observe that the curve becomes shorter. # Change spline type from Bezier to Nurbs - observe that the curve becomes shorter

Daniel Bystedt commented

2021-02-28 15:41:34 +01:00

Added subscriber: @DanielBystedt

Falk David commented

2021-03-01 09:43:38 +01:00

Added subscriber: @filedescriptor

Falk David commented

2021-03-01 09:43:38 +01:00

Changed status from 'Needs Triage' to: 'Confirmed'

Falk David commented

2021-03-01 09:43:38 +01:00

Can confirm this on the 2.92 release and the latest 2.93.0 Alpha, branch: master, commit date: 2021-03-01 06:12, hash: bfc70a6a95.

Can confirm this on the 2.92 release and the latest 2.93.0 Alpha, branch: master, commit date: 2021-03-01 06:12, hash: `bfc70a6a95`.

Bastien Montagne commented

2021-03-01 15:42:49 +01:00

Added subscriber: @mont29

Bastien Montagne commented

2021-03-01 15:42:49 +01:00

Would consider that a known issue rather than a bug, this is just the conversion 'algorithm' being dumb and super-basic, instead of trying to be smart...

Note that you will always lose some data during such conversion anyway, since interpolation methods are quiet different.

Would consider that a known issue rather than a bug, this is just the conversion 'algorithm' being dumb and super-basic, instead of trying to be smart... Note that you will always lose some data during such conversion anyway, since interpolation methods are quiet different.

Campbell Barton commented

2021-08-06 09:36:53 +02:00

Added subscriber: @ideasman42

Campbell Barton commented

2021-08-06 09:36:53 +02:00

Setting as paper-cut since this should be straightforward.

The desirable behavior is to keep the same number of knots (control points) for each curve type instead of treating the handles of a bezier curve as knots.

Setting as paper-cut since this should be straightforward. The desirable behavior is to keep the same number of knots (control points) for each curve type instead of treating the handles of a bezier curve as knots.

Kilo Charlie Alpha commented

2021-08-28 05:39:19 +02:00

Added subscriber: @KiloCharlieAlpha

Kilo Charlie Alpha commented

2021-08-28 05:39:19 +02:00

Hi, new dev here. May I have a go at this? Also, any deadlines to complete this?

Hans Goudey commented

2021-08-28 06:22:00 +02:00

Added subscriber: @HooglyBoogly

Hans Goudey commented

2021-08-28 06:22:00 +02:00

Hi @KiloCharlieAlpha, the curve code is in a bit of a state of flux currently with the work on curves in geometry nodes for 3.0.
We have introduced a new data structure with support for attributes, clearer code, and better performance.

In geometry nodes there is a "Set Spline Type" node with an implementation the 6 type conversions.
If you're interested, you could look at changing the behavior in the node. Then as we transition edit mode curves to use
that data structure (still a loose plan at this point), we would get the improvements there too.

That said, I'm not sure I'd call this behavior a bug-- there are two possible goals here, keep the shape of the input curve, and keep the length.
AFAIK, the "dumb" algorithm that shortens the curve also keeps the shape quite nicely. Still worth some investigation though!

Hi @KiloCharlieAlpha, the curve code is in a bit of a state of flux currently with the work on curves in geometry nodes for 3.0. We have introduced a new data structure with support for attributes, clearer code, and better performance. In geometry nodes there is a "Set Spline Type" node with an implementation the 6 type conversions. If you're interested, you could look at changing the behavior in the node. Then as we transition edit mode curves to use that data structure (still a loose plan at this point), we would get the improvements there too. That said, I'm not sure I'd call this behavior a bug-- there are two possible goals here, keep the shape of the input curve, and keep the length. AFAIK, the "dumb" algorithm that shortens the curve also keeps the shape quite nicely. Still worth some investigation though!

Kilo Charlie Alpha commented

2021-08-30 11:29:00 +02:00

I tried converting a simple Bezier with 2 BezTriples to NURBS. It doesnt generate any knots because the curve is invalid:

In BKE_nurb_check_valid_u, it checks if the number of ctrl pts is less than the order, that looks correct to me, but why does it only use the ctrl pts in the 2 BezTriples? Should it use the 2nd handle of the first ctrl pt, and the first handle of the second ctrl pt to make it 4 ctrl pts so that the validity check passes? But this still fails the next check, where 4th order Beziers must have at least 5 points, but isn't 4 points sufficient?

I have only worked with Beziers and B-splines, not NURBS, and even the Beziers I work with are 4 ctrl pts per segment, so these triples are confusing to me as well.

I tried converting a simple Bezier with 2 BezTriples to NURBS. It doesnt generate any knots because the curve is invalid: - In BKE_nurb_check_valid_u, it checks if the number of ctrl pts is less than the order, that looks correct to me, but why does it only use the ctrl pts in the 2 BezTriples? Should it use the 2nd handle of the first ctrl pt, and the first handle of the second ctrl pt to make it 4 ctrl pts so that the validity check passes? But this still fails the next check, where 4th order Beziers must have at least 5 points, but isn't 4 points sufficient? I have only worked with Beziers and B-splines, not NURBS, and even the Beziers I work with are 4 ctrl pts per segment, so these triples are confusing to me as well.

Kilo Charlie Alpha commented

2021-10-25 15:29:39 +02:00

Removed subscriber: @KiloCharlieAlpha

Yuchen Wen commented

2021-11-28 16:44:39 +01:00

Added subscriber: @ksar

Yuchen Wen commented

2021-11-28 19:12:50 +01:00

Hi, I think maybe I can have a try.

Hans Goudey commented

2021-11-29 01:55:08 +01:00

I'm removing the "Good First Issue' tag from this one since I'm not convinced it's actually a bug (as in, the current algorithm is working as expected, it's not some easy mistake to fix). The conversions can be improved, but I would like that to be done for the set spline type node, since the edit mode code will be replaced with that version at some point.

Yuchen Wen commented

2021-11-29 07:55:42 +01:00

Yes, since the interpolation methods are different, different performances are expected. However, should the backward conversion keep a consistent shape?

Piotr Makal commented

2021-12-11 20:13:21 +01:00

Added subscriber: @pmakal

Piotr Makal commented

2021-12-11 20:13:21 +01:00

@HooglyBoogly if you have a minute please take a look at: D13546

@HooglyBoogly if you have a minute please take a look at: [D13546](https://archive.blender.org/developer/D13546)

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Setting curves to another spline type makes them shorter #86086

Change spline type from Bezier to Nurbs - observe that the curve becomes shorter