Randomness in arrays - first attempt #38527

New Issue

Patrice Bertrand · 2014-02-06T20:48:00+01:00

Patrice Bertrand commented

2014-02-06 20:48:00 +01:00

{F76260}Hello, this patch is a first attempt at adding a "noise" or "randomness" option in the array modifier.
Some discussion regarding this attempt can be found here: http://lists.blender.org/pipermail/bf-committers/2014-February/042760.html

I am posting this just to get some feedback, since this is my very first work on Blender code. I only include the Mod_array.c diff, since this is where all significant processing takes place. As you will notice, the loop acting on vertices was left untouched, so that the overhead as per the original array modifier is very low.

The feature works, including when mixed with other modifiers.. For the moment I made it simple, with just 3 parameters: add_noise is a boolean, included in amd-> flags, that is true when noise option has been requested, amd->noise_seed is an integer seed that is fed to RNG, and amd->noise_amplitude is a float that defines the range of the random translation. It would be very easy to add more customization, such as defining amplitude on the x, y and z axis and similarly adding rotation on each axis.

But, as I said in the comment referenced above, it appeared to me that a noise modifier would be a better approach, that could be used in a more flexible way on top of a number of array modifiers. The noise modifier would act on loose parts, keeping each loose part as it is and applying random, customizable, translation and rotation.

{[F76260](https://archive.blender.org/developer/F76260/array_noise.diff)}Hello, this patch is a first attempt at adding a "noise" or "randomness" option in the array modifier. Some discussion regarding this attempt can be found here: http://lists.blender.org/pipermail/bf-committers/2014-February/042760.html I am posting this just to get some feedback, since this is my very first work on Blender code. I only include the Mod_array.c diff, since this is where all significant processing takes place. As you will notice, the loop acting on vertices was left untouched, so that the overhead as per the original array modifier is very low. The feature works, including when mixed with other modifiers.. For the moment I made it simple, with just 3 parameters: add_noise is a boolean, included in amd-> flags, that is true when noise option has been requested, amd->noise_seed is an integer seed that is fed to RNG, and amd->noise_amplitude is a float that defines the range of the random translation. It would be very easy to add more customization, such as defining amplitude on the x, y and z axis and similarly adding rotation on each axis. But, as I said in the comment referenced above, it appeared to me that a noise modifier would be a better approach, that could be used in a more flexible way on top of a number of array modifiers. The noise modifier would act on loose parts, keeping each loose part as it is and applying random, customizable, translation and rotation.

Patrice Bertrand commented

2014-02-06 20:48:00 +01:00

Changed status to: 'Open'

Patrice Bertrand self-assigned this 2014-02-06 20:48:00 +01:00

Patrice Bertrand commented

2014-02-06 20:48:00 +01:00

Added subscriber: @PatriceBertrand

Warren Bahler commented

2014-02-10 17:13:23 +01:00

Added subscriber: @WarrenBahler

Warren Bahler commented

2014-02-10 17:27:20 +01:00

I don't have a build environment set up to test, but this is a feature I've though should be implemented for a while.

I think a simple 1 dimension random transform is sufficient for most cases;if the user needs more randomness he could add the array to a simple plane or empty and use dupli objects with a group to achieve a lot of variation.

or perhaps a "Dupligroup" option could be added directly to the modifier.

for example an array of empties could be assigned a dupli group which would instance a random object from the group; along with a random offset and transform this would be very powerful.
but that may be outside the scope of the array modifier?

I don't have a build environment set up to test, but this is a feature I've though should be implemented for a while. I think a simple 1 dimension random transform is sufficient for most cases;if the user needs more randomness he could add the array to a simple plane or empty and use dupli objects with a group to achieve a lot of variation. or perhaps a "Dupligroup" option could be added directly to the modifier. for example an array of empties could be assigned a dupli group which would instance a random object from the group; along with a random offset and transform this would be very powerful. but that may be outside the scope of the array modifier?

Patrice Bertrand commented

2014-02-11 10:12:13 +01:00

Thank you for your feedback.

Testing the feature at this point could be a waste of time, I was posting the patch to get some feedback on coding strategy and design from experienced developpers.
I am not familiar with dupli-xxx stuff, will look into it.
It seems to me the case of 2 dimensional arrays is really important, and really handled too poorly by my original approach, ie acting at the array modifier level. The second array modifier takes the output of the first as a whole, so it is impossible to add randomness to individual instances in the second line of the array, which really spoils it all. Here are 2 images that tell it all : The first one is a single array with random translation applied to each instance. The second image is what happens when adding the second dimension: the second line of items is a copy of the fist line, with random translation applied to the whole line.

So I am working on another approach, that will act after array modifiers, or after whatever other modifier, which will be more flexible and give far more control. Acting after array modifiers means you don't have the original object's mesh anymore, all you have is the derived mesh handed to you by previous modifiers. So I am trying to act on loose parts, which in many cases but not all, will be the original object.

Indeed, yet another approach would be to make up an array of empty, of 1, 2 or 3 dimensions, and then have a modifier that duplicates objects centered on each of these empties, each object having a random variation (translation / rotation / scale) applied to it. This would be quite flexible also. I might give it a try.

Thank you for your feedback. Testing the feature at this point could be a waste of time, I was posting the patch to get some feedback on coding strategy and design from experienced developpers. I am not familiar with dupli-xxx stuff, will look into it. It seems to me the case of 2 dimensional arrays is really important, and really handled too poorly by my original approach, ie acting at the array modifier level. The second array modifier takes the output of the first _as a whole_, so it is impossible to add randomness to individual instances in the second line of the array, which really spoils it all. [Here are 2 images that tell it all ](https://hubic.com/home/pub/?ruid=aHR0cHM6Ly9sYjEuaHViaWMub3ZoLm5ldC92MS9BVVRIX2E0ZjZlZjkxZDM0N2RiYmViNzEyNjJlZTk3M2EzNWU5L2RlZmF1bHQvLm92aFB1Yi8xMzkyMTA5NzA1XzEzOTQ3MDE3MDU/dGVtcF91cmxfc2lnPTgxMWUzZTQzMTIyNWY5ZGNmMjhjZGEwNDZlMjY2M2U1ZjlhYmJjZDQmdGVtcF91cmxfZXhwaXJlcz0xMzk0NzAxNzA1#): The first one is a single array with random translation applied to each instance. The second image is what happens when adding the second dimension: the second line of items is a copy of the fist line, with random translation applied to the whole line. So I am working on another approach, that will act _after_ array modifiers, or after whatever other modifier, which will be more flexible and give far more control. Acting after array modifiers means you don't have the original object's mesh anymore, all you have is the derived mesh handed to you by previous modifiers. So I am trying to act on loose parts, which in many cases but not all, will be the original object. Indeed, yet another approach would be to make up an array of empty, of 1, 2 or 3 dimensions, and then have a modifier that duplicates objects centered on each of these empties, each object having a random variation (translation / rotation / scale) applied to it. This would be quite flexible also. I might give it a try.

Patrice Bertrand commented

2014-02-15 21:54:59 +01:00

I have chosen to implement this as "noise modifier", of which I just submitted the patch. So this is closed.

Patrice Bertrand commented

2014-02-15 21:55:28 +01:00

Changed status from 'Open' to: 'Resolved'

Patrice Bertrand closed this issue

2014-02-15 21:55:28 +01:00

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Randomness in arrays - first attempt #38527