Apply random selection factor precisely for all operators #87228

New Issue

Campbell Barton · 2021-04-06T07:32:07+02:00

Campbell Barton commented

2021-04-06 07:32:07 +02:00

Support accurate random selection.

Recently 9c20228128 was committed that ensures the number of selected items meets the requested factor.

Before this change, the selection probability was controlled by the factor, however the actual number of selected items would vary.

This change was only made for edit-mesh, there are other random selection operators that should use the more accurate behavior.

- CURVE_OT_select_random 1c8c91384c
- LATTICE_OT_select_random 1c8c91384c
- OBJECT_OT_select_random 1c8c91384c
- PARTICLE_OT_select_random
- MBALL_OT_select_random_metaelems

Adding this as a TODO, and marking as "Good First Issue", since following the logic in MESH_OT_select_random should be fairly straightforward.

Support accurate random selection. Recently 9c20228128 was committed that ensures the number of selected items meets the requested factor. Before this change, the selection probability was controlled by the factor, however the actual number of selected items would vary. This change was only made for edit-mesh, there are other random selection operators that should use the more accurate behavior. - - [x] `CURVE_OT_select_random` 1c8c91384c - - [x] `LATTICE_OT_select_random` 1c8c91384c - - [x] `OBJECT_OT_select_random` 1c8c91384c - - [ ] `PARTICLE_OT_select_random` - - [ ] `MBALL_OT_select_random_metaelems` Adding this as a TODO, and marking as "Good First Issue", since following the logic in `MESH_OT_select_random` should be fairly straightforward.

Campbell Barton commented

2021-04-06 07:32:07 +02:00

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

Campbell Barton commented

2021-04-06 07:32:07 +02:00

Added subscriber: @ideasman42

Yaser commented

2021-04-09 20:40:06 +02:00

Added subscriber: @nest

Yaser commented

2021-04-09 20:40:06 +02:00

Greetings. I'm interested in doing this issue. Maybe partially only as I'm still new in contributing here.
I used blender regularly before and I believe I understand what the problem is.

I have looked into the code mentioned, MESH_OT_select_random in editmesh_select.c and generally understood the logic, it randomizes the list of verts (or edges ,faces) and selects x items from it based on the random factor.
Then I took a look in object_select.c in object_select_random_exec but did not understand this line (1463) : CTX_DATA_BEGIN (C, Base *, base, selectable_bases) {
Is it doing a loop over all the objects selectable? If so, is selectabe_bases a list?

I apologize if my question is missing something or this is not the place to ask.
Thanks.

Greetings. I'm interested in doing this issue. Maybe partially only as I'm still new in contributing here. I used blender regularly before and I believe I understand what the problem is. I have looked into the code mentioned, MESH_OT_select_random in editmesh_select.c and generally understood the logic, it randomizes the list of verts (or edges ,faces) and selects x items from it based on the random factor. Then I took a look in object_select.c in object_select_random_exec but did not understand this line (1463) : CTX_DATA_BEGIN (C, Base *, base, selectable_bases) { Is it doing a loop over all the objects selectable? If so, is selectabe_bases a list? I apologize if my question is missing something or this is not the place to ask. Thanks.

Brennen Green commented

2021-04-18 00:30:59 +02:00

Added subscriber: @dreymatic

Brennen Green commented

2021-04-18 19:42:10 +02:00

Hi, I've been working on implementing this for OBJECT_OT_select_random. I believe that I am essentially done except for one part of the specification that was referenced here: https://developer.blender.org/D1927 by @ideasman42

The only part of the specification I have not quite figured out for the Object implementation is finding out the total number of bases:
Base **bases = MEM_mallocN(sizeof(*bases) * TOT_BASES, __func__); TOT_BASES is the value I need to figure out. After I figure this out I should be able to test and submit my diff :)

I was curious if anyone had any advice or recommendation for this?

Hi, I've been working on implementing this for OBJECT_OT_select_random. I believe that I am essentially done except for one part of the specification that was referenced here: https://developer.blender.org/D1927 by @ideasman42 The only part of the specification I have not quite figured out for the Object implementation is finding out the total number of bases: `Base **bases = MEM_mallocN(sizeof(*bases) * TOT_BASES, __func__);` TOT_BASES is the value I need to figure out. After I figure this out I should be able to test and submit my diff :) I was curious if anyone had any advice or recommendation for this?

Brennen Green commented

2021-04-19 19:55:33 +02:00

Implemented a fix for OBJECT_OT_select_random. Here's the diff: https://developer.blender.org/D11017 Let me know if I can submit the diff in a better way!

Peter Carter commented

2021-05-08 03:28:08 +02:00

Added subscriber: @pjc7287

Piotr Makal commented

2021-06-23 20:46:07 +02:00

Added subscriber: @pmakal

Piotr Makal commented

2021-06-23 20:46:07 +02:00

Uploaded diff for a review: https://developer.blender.org/D11685
@ideasman42 I would really appreciate it if you would check it :)

I haven't touch particles since gathering both points and keys to a one container in pure C code was a little bit to much for me (allocating memory and stuff). Also metaballs are not part of my diff since I just couldn't understand the code and the metaball behavior during debugging when doing random selection. In my testing it seem like every metaball compound (a shape consisted of many touching metaballs) was having only one element thus this loop is pointless:

if (BLI_rng_get_float(rng) < randfac) {
  if (select) {
    ml->flag |= SELECT;
  }
  else {
    ml->flag &= ~SELECT;
  }
}

}```

But I'm not metaball expert and https://wiki.blender.org/wiki/Modules doesn't point out any person I could ask about it so I just left it as is.

Uploaded diff for a review: https://developer.blender.org/D11685 @ideasman42 I would really appreciate it if you would check it :) I haven't touch particles since gathering both points and keys to a one container in pure C code was a little bit to much for me (allocating memory and stuff). Also metaballs are not part of my diff since I just couldn't understand the code and the metaball behavior during debugging when doing random selection. In my testing it seem like every metaball compound (a shape consisted of many touching metaballs) was having only one element thus this loop is pointless: ``` LISTBASE_FOREACH (MetaElem *, ml, mb->editelems) { ``` if (BLI_rng_get_float(rng) < randfac) { if (select) { ml->flag |= SELECT; } else { ml->flag &= ~SELECT; } } }``` ``` But I'm not metaball expert and https://wiki.blender.org/wiki/Modules doesn't point out any person I could ask about it so I just left it as is.

blender-admin commented

2021-06-28 09:04:34 +02:00

This issue was referenced by 1c8c91384c

This issue was referenced by 1c8c91384c8d7d23b9a966802182e3ef607103ac

blender-admin commented

2021-06-29 08:18:26 +02:00

This issue was referenced by 2d4ec90497

This issue was referenced by 2d4ec90497443dc28e342c539e65010c7f4a04bb

Matias Piipari commented

2022-01-06 23:56:34 +01:00

Added subscriber: @Matias-Piipari

Matias Piipari commented

2022-01-06 23:56:34 +01:00

Hi there! I took a stab at addressing the PARTICLE_OT_select_random case of this issue with the following diff: https://developer.blender.org/D13758 -- this being my first Blender contribution, please excuse the various ways in which I must be failing at following intended conventions :-)

I'm also not sure if there's also some more efficient way of dealing with the point-wise selection (RAN_POINTS), in particular the 1st pass in that case through the visible keys of the visible points is done just to get the allocation size for the array of point / key pairs (in the other RAN_HAIR case I treated the total point count as an upper bound for that array size).

Also, should there be some test scenarios for all of this?

Hi there! I took a stab at addressing the `PARTICLE_OT_select_random` case of this issue with the following diff: https://developer.blender.org/D13758 -- this being my first Blender contribution, please excuse the various ways in which I must be failing at following intended conventions :-) I'm also not sure if there's also some more efficient way of dealing with the point-wise selection (RAN_POINTS), in particular the 1st pass in that case through the visible keys of the visible points is done just to get the allocation size for the array of point / key pairs (in the other RAN_HAIR case I treated the total point count as an upper bound for that array size). Also, should there be some test scenarios for all of this?

San Rin commented

2022-04-03 18:44:09 +02:00

Added subscriber: @Rin-San

San Rin commented

2022-04-04 15:29:50 +02:00

Hi! I upload a diff to supplement MBALL_OT_select_random_metaelems : https://developer.blender.org/D14545
hope it will help

Hi! I upload a diff to supplement `MBALL_OT_select_random_metaelems` : https://developer.blender.org/D14545 hope it will help

Philipp Oeser removed the

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Apply random selection factor precisely for all operators #87228