Mesh: Subdivide op flushes selection too eagerly #83181

New Issue

Stanislav Blinov · 2020-11-29T13:49:53+01:00

Stanislav Blinov commented

2020-11-29 13:49:53 +01:00

Blender Version
Broken: 2.92 master
Worked: N/A

Short description of error

The Subdivide operator selects geometry that it shouldn't. Consider this example:

It looks as if the op flushes selection a bit too much. Note that the 'Expected' selection is valid selection in edge mode, as the edge rings around the rims are not selected. Those edges should neither participate in the op, nor be selected as the result of it, if vertex selection mode is not enabled.

A better example:

...here in face mode, a lone edge is getting selected by the op.

Exact steps for others to reproduce the error

On the default cube, in face mode select top and bottom faces, inset and bridge them.
In edge mode select vertical edge rings.
Subdivide.

Alternatively, load this simplified example file:

subdivide_selection_issue.blend

...and Subdivide. Resulting selection of the middle face is unwarranted.

**Blender Version** Broken: 2.92 master Worked: N/A **Short description of error** The Subdivide operator selects geometry that it shouldn't. Consider this example: ![image.png](https://archive.blender.org/developer/F9407253/image.png) It looks as if the op flushes selection a bit too much. Note that the 'Expected' selection is valid selection in edge mode, as the edge rings around the rims are **not** selected. Those edges should neither participate in the op, nor be selected as the result of it, if vertex selection mode is not enabled. A better example: ![image.png](https://archive.blender.org/developer/F9415315/image.png) ...here in face mode, a lone edge is getting selected by the op. **Exact steps for others to reproduce the error** 1. On the default cube, in face mode select top and bottom faces, inset and bridge them. 2. In edge mode select vertical edge rings. 3. Subdivide. Alternatively, load this simplified example file: [subdivide_selection_issue.blend](https://archive.blender.org/developer/F9407395/subdivide_selection_issue.blend) ...and Subdivide. Resulting selection of the middle face is unwarranted.

Stanislav Blinov commented

2020-11-29 13:49:53 +01:00

Added subscriber: @Stan_Pancakes

Philipp Oeser commented

2020-11-30 10:58:58 +01:00

Added subscriber: @lichtwerk

Philipp Oeser commented

2020-11-30 10:58:58 +01:00

Think this is a similar problem/situation as {#73771}?
After subdivide, the original selection as well as the added geometry are selected. If this is flushed, you end up with everything selected

2.79b did the same thing, no?

Think this is a similar problem/situation as {#73771}? After subdivide, the original selection as well as the added geometry are selected. If this is flushed, you end up with everything selected 2.79b did the same thing, no? ![image.png](https://archive.blender.org/developer/F9413253/image.png) ![image.png](https://archive.blender.org/developer/F9413263/image.png)

Philipp Oeser commented

2020-11-30 11:01:38 +01:00

![image.png](https://archive.blender.org/developer/F9413272/image.png) ![image.png](https://archive.blender.org/developer/F9413277/image.png)

Philipp Oeser commented

2020-11-30 11:15:49 +01:00

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

Philipp Oeser closed this issue

2020-11-30 11:15:49 +01:00

Philipp Oeser commented

2020-11-30 11:15:49 +01:00

So unless we can really make a point for this being a bug, I think we have to close this.
Like I said, feel free to comment again [we can always reopen], but as it stands now, I dont think this is a bug.

So unless we can really make a point for this being a bug, I think we have to close this. Like I said, feel free to comment again [we can always reopen], but as it stands now, I dont think this is a bug.

Stanislav Blinov commented

2020-11-30 14:05:35 +01:00

My bad, I must've done something wrong while testing on 2.79b. Re-testing indeed shows the same behavior as current.

Still, I maintain that this flush is incorrect. The new geometry in this case is that which is selected in my "Expected" case: the new loops and new faces, and as a result - edges around them. Had this been done in vertex selection mode, then yes, it should've selected the rim as well, as a consequence of vertex selection propagating to edges. But in this case this is done in edge selection mode, yet it seems the op is still flushing from verts.

My bad, I must've done something wrong while testing on 2.79b. Re-testing indeed shows the same behavior as current. Still, I maintain that this flush is incorrect. The new geometry in this case is that which is selected in my "Expected" case: the new loops and new faces, and as a result - edges around them. Had this been done in vertex selection mode, then yes, it should've selected the rim as well, as a consequence of vertex selection propagating to edges. But in this case this is done in edge selection mode, yet it seems the op is still flushing from verts.

Stanislav Blinov commented

2020-11-30 18:06:55 +01:00

Here's a more concrete indication that this is a bug:

In both edge and face mode, the edge between those original selections gets selected as a result of Subdivide. It shouldn't be, in either mode, especially in face mode as that creates selection inconsistent with current mode.

Here's a more concrete indication that this is a bug: ![image.png](https://archive.blender.org/developer/F9415315/image.png) In both edge and face mode, the edge between those original selections gets selected as a result of Subdivide. It shouldn't be, in either mode, especially in face mode as that creates selection inconsistent with current mode.

Stanislav Blinov commented

2020-12-02 11:47:56 +01:00

@lichtwerk Any thoughts wrt. above?

Stanislav Blinov commented

2020-12-12 00:59:51 +01:00

Should I open a new one?..

Philipp Oeser commented

2020-12-12 07:53:12 +01:00

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

Philipp Oeser reopened this issue

2020-12-12 07:53:12 +01:00

Philipp Oeser commented

2020-12-12 07:53:12 +01:00

Need to have a look again...

888 commented

2020-12-17 07:55:49 +01:00

Added subscriber: @wiwa

888 commented

2020-12-17 07:55:49 +01:00

This comment was removed by @wiwa

*This comment was removed by @wiwa*

888 commented

2020-12-17 07:56:50 +01:00

This comment was removed by @wiwa

*This comment was removed by @wiwa*

Sybren A. Stüvel commented

2020-12-17 10:39:08 +01:00

Added subscriber: @dr.sybren

Sybren A. Stüvel commented

2020-12-17 10:39:19 +01:00

Removed subscribers: @dr.sybren, @wiwa

Philipp Oeser commented

2020-12-17 11:08:15 +01:00

@Stan_Pancakes : Still think this is expected behavior tbh.

We have https://docs.blender.org/api/current/bmesh.html#keeping-a-correct-state which tells us each faces vertices should get selected on the operation and flushed.
Both participating verts of the edge in question are selected in your example (and thus drawn selected).
In a way, your second example does not really differ from your first example?

This (drawing indirectly participating edges selected) doesnt happen on raw face selection (because I think flushing only takes places when changing to e.g. vertex or edge selection mode), so I would still think this is all expected actually.
Why again would you think the second example is different from the first?

@Stan_Pancakes : Still think this is expected behavior tbh. We have https://docs.blender.org/api/current/bmesh.html#keeping-a-correct-state which tells us each faces vertices should get selected on the operation and flushed. Both participating verts of the edge in question are selected in your example (and thus drawn selected). In a way, your second example does not really differ from your first example? This (drawing indirectly participating edges selected) doesnt happen on raw face selection (because I think flushing only takes places when changing to e.g. vertex or edge selection mode), so I would still think this is all expected actually. Why again would you think the second example is different from the first?

Stanislav Blinov commented

2020-12-17 11:53:35 +01:00

It's not different, it's the same thing just better illustraded. That edge is getting selected regardless of mode you're in. For face mode, that's not even a correct selection state - you can't achieve such selection through normal means (and should not be able to).

I haven't looked at the code yet, but from what's going on I gather that the op simply selects all the resulting verts and flushes from them. That's not correct behavior. If you're in edge selection mode, flushing should be done from edges to verts and to faces, but not to verts and then back to edges. I.e. something like this (quick mockup, not thoroughly tested):

import bpy
import bmesh

def subdivide_mockup(context):
    mesh = context.object.data
    bm = bmesh.from_edit_mesh(mesh)
    
    edges = [e for e in bm.edges if e.select]
    
    bpy.ops.mesh.select_all(action = 'DESELECT')
    
    # other options omitted for brevity
    result = bmesh.ops.subdivide_edges(bm, edges = edges, cuts = 1, use_grid_fill = True)
    
    for e in filter(lambda e: isinstance(e, bmesh.types.BMEdge), result['geom_inner']):
        e.select = True
        # also select new faces since that's what Subdivide appears to want to do
        for f in e.link_faces:
            f.select = True
            
    for g in result['geom_split']:
        g.select = True
    
    bm.select_flush_mode()
    bmesh.update_edit_mesh(mesh)
    
subdivide_mockup(bpy.context)

It's not different, it's the same thing just better illustraded. That edge is getting selected regardless of mode you're in. For face mode, that's not even a correct selection state - you can't achieve such selection through normal means (and should not be able to). I haven't looked at the code yet, but from what's going on I gather that the op simply selects all the resulting verts and flushes from them. That's not correct behavior. If you're in edge selection mode, flushing should be done from edges to verts and to faces, but not to verts and then back to edges. I.e. something like this (quick mockup, not thoroughly tested): ``` import bpy import bmesh def subdivide_mockup(context): mesh = context.object.data bm = bmesh.from_edit_mesh(mesh) edges = [e for e in bm.edges if e.select] bpy.ops.mesh.select_all(action = 'DESELECT') # other options omitted for brevity result = bmesh.ops.subdivide_edges(bm, edges = edges, cuts = 1, use_grid_fill = True) for e in filter(lambda e: isinstance(e, bmesh.types.BMEdge), result['geom_inner']): e.select = True # also select new faces since that's what Subdivide appears to want to do for f in e.link_faces: f.select = True for g in result['geom_split']: g.select = True bm.select_flush_mode() bmesh.update_edit_mesh(mesh) subdivide_mockup(bpy.context) ```

Philipp Oeser commented

2020-12-22 15:57:45 +01:00

Changed status from 'Needs Triage' to: 'Needs Developer To Reproduce'

Philipp Oeser commented

2020-12-22 15:57:45 +01:00

Added subscribers: @DanielBystedt, @mont29, @ideasman42

Philipp Oeser commented

2020-12-22 15:57:45 +01:00

Afraid I cannot make the final call here.
To me, this does not look like a bug, but there are cases where the described behavior would be benificial.

CC @mont29
CC @ideasman42
CC @DanielBystedt

Afraid I cannot make the final call here. To me, this does not look like a bug, but there are cases where the described behavior would be benificial. CC @mont29 CC @ideasman42 CC @DanielBystedt

Daniel Bystedt commented

2020-12-22 23:12:29 +01:00

Good find @Stan_Pancakes

I agree with what you write regarding not ending up with the expected selection after using the subdivide operator. I am stating this from a design perspective as it this might be the expected behavior based on how the code is written.

Below I did a test with the example you described above. I agree that after subdividing the "inner" and "outer" edge ring, the faces pointing in +Z and -Z should not be selected.

#83181 subdivide edges.mp4

Interesting enough, I ended up with the expected selection in this example below. After subdividing the selected faces I first changed to vertex select and then face select, which gave me a desired selection. I agree that the edge in the middle should not be selected after subdividing the faces in the example below

#83181 subdivide faces.mp4

Good find @Stan_Pancakes I agree with what you write regarding not ending up with the expected selection after using the subdivide operator. I am stating this from a design perspective as it this might be the expected behavior based on how the code is written. --- Below I did a test with the example you described above. I agree that after subdividing the "inner" and "outer" edge ring, the faces pointing in +Z and -Z should not be selected. [#83181 subdivide edges.mp4](https://archive.blender.org/developer/F9525862/T83181_subdivide_edges.mp4) --- Interesting enough, I ended up with the expected selection in this example below. After subdividing the selected faces I first changed to **vertex select** and then **face select**, which gave me a desired selection. I agree that the edge in the middle should not be selected after subdividing the faces in the example below [#83181 subdivide faces.mp4](https://archive.blender.org/developer/F9525867/T83181_subdivide_faces.mp4)

Stanislav Blinov commented

2020-12-22 23:32:09 +01:00

Yes, mode switching validates selection, thus enabling exclusive face selection mode while having a lone edge selected deselects that edge. This is why I maintain that such selection shouldn't be created by the op in the first place (i.e. that this behavior is actually a bug in the op). The thing is, the examples are equivalent: in both cases the stray edge is being selected due to flushing from vertices regardless of selection mode. It's just that the second example shows more clearly why this is a problem.

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Mesh: Subdivide op flushes selection too eagerly #83181