blender
/
blender
133
398
Fork 579
Code Issues 6.3k Pull Requests 747 Projects 19 Wiki Activity

Function Node: Hash Value #110769

Open
Charlie Jolly wants to merge 18 commits from CharlieJolly/blender:gn-hash-node into main
pull from: CharlieJolly/blender:gn-hash-node
merge into: blender:main

When changing the target branch, be careful to rebase the branch in your fork to match. See documentation.
Conversation 26 Commits 18 Files Changed 6 +201
Charlie Jolly commented 2023-08-03 17:02:34 +02:00
Member
Copy Link

This node hashes float, vector, color, string and integers to a hash.

Uses hash functions from BLI_noise.hh.

These are also used in White Noise node.

I'm not experienced enough with the hash functions to know if
the int/uint conversion is a real issue or not.

Originally suggested in comments for D15375

Based on archived patch: https://archive.blender.org/developer/D15835

Mentioned here: #103457

image

This node hashes float, vector, color, string and integers to a hash. Uses hash functions from BLI_noise.hh. These are also used in White Noise node. I'm not experienced enough with the hash functions to know if the int/uint conversion is a real issue or not. Originally suggested in comments for D15375 Based on archived patch: https://archive.blender.org/developer/D15835 Mentioned here: https://projects.blender.org/blender/blender/issues/103457 ![image](/attachments/79fd45ab-8915-472f-953e-5b0a6d3a6673)
image.png
49 KiB
image.png
44 KiB
image.png
32 KiB
image.png image.png
👍 2 ❤️ 2
Charlie Jolly added 2 commits 2023-08-03 17:02:39 +02:00
e24c6dbb4c Function Node: Hash Value 
This node hashes float, vector, color, string and integers to a hash.

Uses hash functions from BLI_hash and BLI_noise.

Suggested in comments for D15375

Based on https://archive.blender.org/developer/D15835
Charlie Jolly added this to the Nodes & Physics project 2023-08-03 17:02:50 +02:00
Iliya Katushenock added the
Interest
Geometry Nodes
 label 2023-08-03 17:04:39 +02:00
Jacques Lucke requested changes 2023-08-03 17:37:38 +02:00
Jacques Lucke left a comment
Member
Copy Link

We haven't discussed that in a meeting yet, but currently I think this is a useful low level building block.

We haven't discussed that in a meeting yet, but currently I think this is a useful low level building block.
source/blender/nodes/function/nodes/node_fn_hash_value.cc Outdated
@ -0,0 +88,4 @@
static auto exec_preset = mf::build::exec_presets::AllSpanOrSingle();
static auto fn_hash_float = mf::build::SI1_SO<float, int>(
"Hash Float", [](float a) { return get_default_hash(a); }, exec_preset);
Jacques Lucke commented 2023-08-03 17:36:15 +02:00
Member
Copy Link

This hash is not designed to generated very random results in these cases, it shouldn't be exposed in the node.
Just expose the noise::hash functions like you did below. Those are used indirectly by noise code already anyway.

This hash is not designed to generated very random results in these cases, it shouldn't be exposed in the node. Just expose the `noise::hash` functions like you did below. Those are used indirectly by noise code already anyway.
Charlie Jolly commented 2023-08-03 17:42:20 +02:00
Author
Member
Copy Link

I'll wait for meeting discussion but I guess the suggestion is to remove all the get_default_hash methods and just use the BLI_noise functions? This also simplifies node without the need for users to choose the method. This was originally added to explore what was useful or not.

I'll wait for meeting discussion but I guess the suggestion is to remove all the `get_default_hash` methods and just use the BLI_noise functions? This also simplifies node without the need for users to choose the method. This was originally added to explore what was useful or not.
Jacques Lucke commented 2023-08-03 17:47:01 +02:00
Member
Copy Link

I guess the suggestion is to remove all the get_default_hash methods and just use the BLI_noise functions?

Yeah, that's what I meant.

> I guess the suggestion is to remove all the get_default_hash methods and just use the BLI_noise functions? Yeah, that's what I meant.
CharlieJolly marked this conversation as resolved
Charlie Jolly added 1 commit 2023-08-03 18:17:43 +02:00
Iliya Katushenock commented 2023-08-03 18:20:43 +02:00
Member
Copy Link

I agree with Jacques, the node seems very useful, I often think about it when I work with groups.

  1. Replace NodeHashMode by simply socket types.
  2. Iterate over all sockets in node_update and compare types.
  3. Wrap all static auto fn_hash_vector in bke::attribute_math::convert_to_static_type (use socket_type_to_custom_data_type) (wrap in constexpr if statement to not generate code for unsupported types (See accumulate or blur attribute nodes as example)).
  4. Use simple for-loop in node_gather_link_searches (nodeStaticSocketLabel, std::string = "HashFor" + nodes::...(type) can be captured by lambda [name = std::move(name)]). And make sure, if connected sockets types is converteble (bke function).
  5. Store all node data in custom1.
  6. Rotation socket can be added too.
I agree with Jacques, the node seems very useful, I often think about it when I work with groups. 1. Replace `NodeHashMode` by simply socket types. 2. Iterate over all sockets in `node_update` and compare types. 3. Wrap all `static auto fn_hash_vector` in `bke::attribute_math::convert_to_static_type` (use `socket_type_to_custom_data_type`) (wrap in constexpr if statement to not generate code for unsupported types (See accumulate or blur attribute nodes as example)). 4. Use simple for-loop in `node_gather_link_searches` (`nodeStaticSocketLabel`, `std::string = "HashFor" + nodes::...(type)` can be captured by lambda `[name = std::move(name)]`). And make sure, if connected sockets types is converteble (bke function). 5. Store all node data in `custom1`. 6. Rotation socket can be added too.
👍 1
Iliya Katushenock commented 2023-08-03 18:26:29 +02:00
Member
Copy Link

This node seems generic enough to be found in utilities like Random Value node.

This node seems generic enough to be found in utilities like Random Value node.
Shmuel Israel commented 2023-08-04 00:47:58 +02:00
First-time contributor
Copy Link

btw how does this work? what does a string turn into? and can you convert a hash into a string? that would be pretty cool.
and what would a float look like after it's converted back and forth?

btw how does this work? what does a string turn into? and can you convert a hash into a string? that would be pretty cool. and what would a float look like after it's converted back and forth?
Iliya Katushenock commented 2023-08-04 01:08:53 +02:00
Member
Copy Link

Hash, this is the thing that may not be stabile very much from many things. You don't even have to think about converting data to and from hash.

Hash, this is the thing that may not be stabile very much from many things. You don't even have to think about converting data to and from hash.
Shmuel Israel commented 2023-08-04 06:51:57 +02:00
First-time contributor
Copy Link

@mod_moder you are saying it isn't stable and you can't convert things back and forth. in that case, why and how to you have these conversion to a hash and from a hash. I'm unsure on the behavior.

@mod_moder you are saying it isn't stable and you can't convert things back and forth. in that case, why and how to you have these conversion to a hash and from a hash. I'm unsure on the behavior.
Iliya Katushenock commented 2023-08-04 14:44:42 +02:00
Member
Copy Link

@shmuel You can get a random number for any data. It must be identical for the same data. But even just x = 10 * 4 / x = 10 + 10 + 10 + 10 for float will actually give us different real numbers and its hashs.

@shmuel You can get a random number for any data. It must be identical for the same data. But even just `x = 10 * 4` / `x = 10 + 10 + 10 + 10` for float will actually give us different real numbers and its hashs.
Charlie Jolly added 2 commits 2023-08-04 20:30:26 +02:00
db60c02ace Remove Hash To Float function 
Remove node storage and use custom1
Use SocketSearchOp
Address some comments from @mod_moder
Iliya Katushenock reviewed 2023-08-04 20:42:08 +02:00
source/blender/nodes/function/nodes/node_fn_hash_value.cc Outdated
@ -0,0 +128,4 @@
}
if (params.in_out() == SOCK_IN) {
if (type == CD_PROP_FLOAT) {
params.add_item(IFACE_("Value"), SocketSearchOp{"Value", type});
Iliya Katushenock commented 2023-08-04 20:42:08 +02:00
Member
Copy Link

To clarify a little: we must add all options if their type can be converted to type. So, for string is {sring}, for color, is {all, except of string and rotation}, ... . That's the point of the loop and the function to check if the socket types are convertible.

To clarify a little: we must add all options if their type can be converted to `type`. So, for string is {sring}, for color, is {all, except of string and rotation}, ... . That's the point of the loop and the function to check if the socket types are convertible.
Charlie Jolly commented 2023-08-07 17:25:21 +02:00
Author
Member
Copy Link

For inputs I'm not sure this is needed. I'd expect to hash by type so only having the one option reduces clutter. So plugging in a Float will use the Float To Hash function etc.

I have added Boolean socket and mapped that to Int.

For inputs I'm not sure this is needed. I'd expect to hash by type so only having the one option reduces clutter. So plugging in a Float will use the Float To Hash function etc. I have added Boolean socket and mapped that to Int.
Iliya Katushenock commented 2023-10-03 18:01:27 +02:00
Member
Copy Link

Node to convert Float to Int doesn't make sense if you create this via drag-and-drop from int output socket of other one node so it is connected to float input, but this still exist in the list to search/

Node to convert Float to Int doesn't make sense if you create this via drag-and-drop from int output socket of other one node so it is connected to float input, but this still exist in the list to search/
Charlie Jolly commented 2023-10-04 11:27:12 +02:00
Author
Member
Copy Link

Maybe you can add a screen shot because I'm not sure I understand.

Maybe you can add a screen shot because I'm not sure I understand.
Iliya Katushenock commented 2023-10-04 11:35:25 +02:00
Member
Copy Link

This search result makes no sense. But it still exists:

Drag and Drop search Result

Your node should also show all options, and not just those that are equal to current socket type.

This search result makes no sense. But it still exists: | Drag and Drop search | Result | | -- | -- | | ![](https://cdn.discordapp.com/attachments/340195875399663617/1159060583732215899/image.png) | ![](https://cdn.discordapp.com/attachments/340195875399663617/1159060583421841448/image.png) Your node should also show all options, and not just those that are equal to current socket type.
Charlie Jolly commented 2023-10-04 11:56:26 +02:00
Author
Member
Copy Link

Sorry, I still don't understand. Drag link only limits by type not node. Hash output is Integer. User might want to plug that into a math node.

Sorry, I still don't understand. Drag link only limits by type not node. Hash output is Integer. User might want to plug that into a math node.
Iliya Katushenock commented 2023-10-04 12:01:39 +02:00
Member
Copy Link

I mean, node_gather_link_searches have to create search items all input socket for all the types supported by the node, and the output one kind at the single call (noy only one input and one output for supported type). Only one limitation of what socket can not be converted to other type by implicit value conversion.

I mean, `node_gather_link_searches` have to create search items all input socket for all the types supported by the node, and the output one kind at the single call (noy only one input and one output for supported type). Only one limitation of what socket can not be converted to other type by implicit value conversion.
Charlie Jolly commented 2023-10-04 12:16:29 +02:00
Author
Member
Copy Link

It might be easier for you to provide code if you need me to change something here.

Or are you talking about weight as is done for math node?

It might be easier for you to provide code if you need me to change something here. Or are you talking about weight as is done for math node?
Iliya Katushenock commented 2023-10-04 12:42:46 +02:00
Member
Copy Link
static void node_gather_link_searches(GatherLinkSearchOpParams &params)
{
  static const std::array<eNodeSocketDatatype> supported_types{...};
  const eNodeSocketDatatype other_type = eNodeSocketDatatype(params.other_socket().type);

  if (ELEM(other_type, SOCK_STRING)) {
    return;
  }

  for (const eNodeSocketDatatype item_type : supported_types) {
    if (params.in_out() == SOCK_OUT) {
      if (!ELEM(SOCK_ROTATION, other_type, item_type)) {
        params.add_item(IFACE_("Hash"), SocketSearchOp{"Hash", CD_PROP_INT32});
      }
      continue;
    }
    if (!ELEM(SOCK_ROTATION, other_type, item_type) || item_type == other_type) {
      params.add_item(IFACE_("Value"), SocketSearchOp{"Value", type});
    }
  }
}
```Cpp static void node_gather_link_searches(GatherLinkSearchOpParams &params) { static const std::array<eNodeSocketDatatype> supported_types{...}; const eNodeSocketDatatype other_type = eNodeSocketDatatype(params.other_socket().type); if (ELEM(other_type, SOCK_STRING)) { return; } for (const eNodeSocketDatatype item_type : supported_types) { if (params.in_out() == SOCK_OUT) { if (!ELEM(SOCK_ROTATION, other_type, item_type)) { params.add_item(IFACE_("Hash"), SocketSearchOp{"Hash", CD_PROP_INT32}); } continue; } if (!ELEM(SOCK_ROTATION, other_type, item_type) || item_type == other_type) { params.add_item(IFACE_("Value"), SocketSearchOp{"Value", type}); } } } ```
Charlie Jolly commented 2023-10-04 13:38:51 +02:00
Author
Member
Copy Link

I'm still not understanding what this is doing different other than using an array and excluding Rotation socks.

I'm still not understanding what this is doing different other than using an array and excluding Rotation socks.
Iliya Katushenock commented 2023-10-04 20:22:19 +02:00
Member
Copy Link
With all possible node types Current one
| With all possible node types | Current one | | -- | -- | | ![](https://cdn.discordapp.com/attachments/340195875399663617/1159193380962697276/image.png) | ![](https://cdn.discordapp.com/attachments/340195875399663617/1159193381361176617/image.png) |
Charlie Jolly commented 2023-10-05 12:27:27 +02:00
Author
Member
Copy Link

Ok, I get what you are saying but we're not consistent here. E.g. Map Range doesn't list all available sockets. Listing all sockets also adds noise that the user needs to go through. In most cases, users will be hashing the data type associated with the socket so current behaviour seems favourable to me.

Ok, I get what you are saying but we're not consistent here. E.g. Map Range doesn't list all available sockets. Listing all sockets also adds noise that the user needs to go through. In most cases, users will be hashing the data type associated with the socket so current behaviour seems favourable to me.
Iliya Katushenock commented 2023-10-11 18:58:12 +02:00
Member
Copy Link

That was my mistake.

That was my mistake.
mod_moder marked this conversation as resolved
Charlie Jolly added 1 commit 2023-08-07 17:25:21 +02:00
Charlie Jolly added 1 commit 2023-08-07 17:31:10 +02:00
Charlie Jolly added 2 commits 2023-10-03 14:16:43 +02:00
f7f381d71a Merge branch 'main' into gn-hash-node 
# Conflicts:
#	source/blender/blenkernel/BKE_node.h
#	source/blender/makesrna/intern/rna_nodetree.cc
#	source/blender/nodes/function/node_function_register.cc
#	source/blender/nodes/function/node_function_register.hh
Charlie Jolly added 1 commit 2023-10-03 14:19:51 +02:00
Charlie Jolly added 1 commit 2023-10-04 11:28:12 +02:00
Charlie Jolly added 1 commit 2023-10-04 11:57:45 +02:00
Jacques Lucke commented 2023-10-06 17:06:45 +02:00
Member
Copy Link

The code generally looks good. I think the main thing that's missing in the description is the motivation and examples for how someone should use this. Like, why should this be used instead of White Noise for example?

The code generally looks good. I think the main thing that's missing in the description is the motivation and examples for how someone should use this. Like, why should this be used instead of White Noise for example?
Charlie Jolly commented 2023-10-06 17:29:52 +02:00
Author
Member
Copy Link

The code generally looks good. I think the main thing that's missing in the description is the motivation and examples for how someone should use this. Like, why should this be used instead of White Noise for example?

This was originally motivated as a response to the comments in D15375. 😝

image

https://archive.blender.org/developer/D15375

White Noise only handles float data so this compliments that to allow users to hash data for indexes etc

@mod_moder Has also mentioned it here Attribute Map node #103457

> The code generally looks good. I think the main thing that's missing in the description is the motivation and examples for how someone should use this. Like, why should this be used instead of White Noise for example? This was originally motivated as a response to the comments in D15375. 😝 ![image](/attachments/07b2cb40-6613-4148-bd68-3dedb7f1785f) https://archive.blender.org/developer/D15375 White Noise only handles float data so this compliments that to allow users to hash data for indexes etc @mod_moder Has also mentioned it here Attribute Map node #103457
image.png
39 KiB
Iliya Katushenock commented 2023-10-06 17:41:52 +02:00
Member
Copy Link

Hash is a good thing to have during groups building from non-integer noises.
Multidimensional hash can be used in recreating all noise textures as node groups (as the base noise node).

Hash is a good thing to have during groups building from non-integer noises. Multidimensional hash can be used in recreating all noise textures as node groups (as the base noise node).
👍 1
Jacques Lucke requested review from Simon Thommes 2023-10-09 11:48:12 +02:00
Jacques Lucke requested review from Hans Goudey 2023-10-09 11:48:12 +02:00
Simon Thommes requested changes 2023-10-09 14:57:55 +02:00
Simon Thommes left a comment
Member
Copy Link

To me it's not immediately obvious, why there are 3 inputs for the Integer type. I can see needing 2, but that should be enough to hash any amount of integers together by folding.
Also this seems a bit strange in comparison with the other types. I'd propose the following:
Add a Seed integer field input, same as the random value node has, for all types. That way it makes the inputs and behavior between types more consistent and intuitive imo. Any type will have one input for the value of that type and a second input for the seed. That happens to be 2 integers for the integer type.

The main input to be hashed should be called Value regardless of the type for consistency with other nodes, e.g. store named attribute.

I'm not sure about the name of the node and the output. I think it's fine to introduce the term hash, as it's descriptive, but we don't have that anywhere in the UI yet afaik. It might be worth considering the term ID instead in something like ID from Value, Random ID, ... to avoid introducing new terms.

@dfelinto , what do you think?

To me it's not immediately obvious, why there are 3 inputs for the Integer type. I can see needing 2, but that should be enough to hash any amount of integers together by folding. Also this seems a bit strange in comparison with the other types. I'd propose the following: Add a `Seed` integer field input, same as the random value node has, for all types. That way it makes the inputs and behavior between types more consistent and intuitive imo. Any type will have one input for the value of that type and a second input for the seed. That happens to be 2 integers for the integer type. The main input to be hashed should be called `Value` regardless of the type for consistency with other nodes, e.g. store named attribute. I'm not sure about the name of the node and the output. I think it's fine to introduce the term hash, as it's descriptive, but we don't have that anywhere in the UI yet afaik. It might be worth considering the term `ID` instead in something like `ID from Value`, `Random ID`, ... to avoid introducing new terms. @dfelinto , what do you think?
👍 3
Hans Goudey commented 2023-10-10 10:08:46 +02:00
Member
Copy Link

Design-wise, combining hash-to-float in a node that otherwise just creates integers is a bit confusing. Maybe that should be a separate node? It also seems a bit arbitrary for the integer mode to have three inputs where the float mode only has one. I guess in an ideal world the number of inputs would be dynamic. Other than that, I do think this node makes sense!

Design-wise, combining hash-to-float in a node that otherwise just creates integers is a bit confusing. Maybe that should be a separate node? It also seems a bit arbitrary for the integer mode to have three inputs where the float mode only has one. I guess in an ideal world the number of inputs would be dynamic. Other than that, I do think this node makes sense!
Simon Thommes commented 2023-10-10 11:01:43 +02:00
Member
Copy Link

combining hash-to-float

Where is that happening? Did I miss something? Afaict the node is always just creating integer hashes.

> combining hash-to-float Where is that happening? Did I miss something? Afaict the node is always just creating integer hashes.
Hans Goudey commented 2023-10-10 11:09:56 +02:00
Member
Copy Link

Where is that happening? Did I miss something? Afaict the node is always just creating integer hashes.

I just saw it in the screenshot in the description.

>Where is that happening? Did I miss something? Afaict the node is always just creating integer hashes. I just saw it in the screenshot in the description.
Charlie Jolly commented 2023-10-10 11:36:12 +02:00
Author
Member
Copy Link

hash-to-float was removed, some changes were discussed in Blender chat a while ago but the node didn't have any tlc for a bit!

`hash-to-float` was removed, some changes were discussed in Blender chat a while ago but the node didn't have any tlc for a bit!
Charlie Jolly commented 2023-10-10 11:58:43 +02:00
Author
Member
Copy Link

To me it's not immediately obvious, why there are 3 inputs for the Integer type. I can see needing 2, but that should be enough to hash any amount of integers together by folding.

The hash functions take 1, 2, 3 or 4 inputs. Exposing 3 seemed a good compromise here for hashing integer based positions.

Also this seems a bit strange in comparison with the other types. I'd propose the following:
Add a Seed integer field input, same as the random value node has, for all types. That way it makes the inputs and behavior between types more consistent and intuitive imo. Any type will have one input for the value of that type and a second input for the seed. That happens to be 2 integers for the integer type.

This complicates the code somewhat, I'm not clear why you would need a seed here? This could also be achieved by adding an offset to the input value.

The main input to be hashed should be called Value regardless of the type for consistency with other nodes, e.g. store named attribute.

We're not consistent here, see Map Range which has Value and Vector inputs depending on type.

I'm not sure about the name of the node and the output. I think it's fine to introduce the term hash, as it's descriptive, but we don't have that anywhere in the UI yet afaik. It might be worth considering the term ID instead in something like ID from Value, Random ID, ... to avoid introducing new terms.

@dfelinto , what do you think?

I'm not opinionated on this.

> To me it's not immediately obvious, why there are 3 inputs for the Integer type. I can see needing 2, but that should be enough to hash any amount of integers together by folding. The hash functions take 1, 2, 3 or 4 inputs. Exposing 3 seemed a good compromise here for hashing integer based positions. > Also this seems a bit strange in comparison with the other types. I'd propose the following: > Add a `Seed` integer field input, same as the random value node has, for all types. That way it makes the inputs and behavior between types more consistent and intuitive imo. Any type will have one input for the value of that type and a second input for the seed. That happens to be 2 integers for the integer type. This complicates the code somewhat, I'm not clear why you would need a seed here? This could also be achieved by adding an offset to the input value. > > The main input to be hashed should be called `Value` regardless of the type for consistency with other nodes, e.g. store named attribute. We're not consistent here, see Map Range which has `Value` and `Vector` inputs depending on type. > > I'm not sure about the name of the node and the output. I think it's fine to introduce the term hash, as it's descriptive, but we don't have that anywhere in the UI yet afaik. It might be worth considering the term `ID` instead in something like `ID from Value`, `Random ID`, ... to avoid introducing new terms. > > @dfelinto , what do you think? I'm not opinionated on this.
Hans Goudey commented 2023-10-10 12:05:48 +02:00
Member
Copy Link

We're not consistent here, see Map Range which has Value and Vector inputs depending on type.

Let's just use the word "Value," it's simpler and using other words is redundant with the socket type itself.

The hash functions take 1, 2, 3 or 4 inputs. Exposing 3 seemed a good compromise here for hashing integer based positions.

Since there's a performance cost for hashing more values, I'd probably go with 2 for now until the amount of inputs could be dynamic. But it's not a big deal either way.

>We're not consistent here, see Map Range which has Value and Vector inputs depending on type. Let's just use the word "Value," it's simpler and using other words is redundant with the socket type itself. >The hash functions take 1, 2, 3 or 4 inputs. Exposing 3 seemed a good compromise here for hashing integer based positions. Since there's a performance cost for hashing more values, I'd probably go with 2 for now until the amount of inputs could be dynamic. But it's not a big deal either way.
Simon Thommes commented 2023-10-10 12:30:04 +02:00
Member
Copy Link

Also this seems a bit strange in comparison with the other types. I'd propose the following:
Add a Seed integer field input, same as the random value node has, for all types. That way it makes the inputs and behavior between types more consistent and intuitive imo. Any type will have one input for the value of that type and a second input for the seed. That happens to be 2 integers for the integer type.

This complicates the code somewhat, I'm not clear why you would need a seed here? This could also be achieved by adding an offset to the input value.

To me it's a matter of the consistency of the design and usability. You could argue for the integer input type just as well, that you could just have a single input and offset it, if you need a different hash. I think it's reasonable to avoid such workarounds to allow multiple inputs hashed together. And rather than having those different for every type I think it's more intuitive, more consistent with other existing nodes and nicer design to introduce that second input as a seed integer value. how much does it complicate the code? To me it is still worth it, if it's not by that much. I'm quite confident that this is a nice solution on the user level.
Maybe someone more technical can evaluate this better on the code level @HooglyBoogly ?

> >Also this seems a bit strange in comparison with the other types. I'd propose the following: > > Add a Seed integer field input, same as the random value node has, for all types. That way it makes the inputs and behavior between types more consistent and intuitive imo. Any type will have one input for the value of that type and a second input for the seed. That happens to be 2 integers for the integer type. > > This complicates the code somewhat, I'm not clear why you would need a seed here? This could also be achieved by adding an offset to the input value. To me it's a matter of the consistency of the design and usability. You could argue for the integer input type just as well, that you could just have a single input and offset it, if you need a different hash. I think it's reasonable to avoid such workarounds to allow multiple inputs hashed together. And rather than having those different for every type I think it's more intuitive, more consistent with other existing nodes and nicer design to introduce that second input as a seed integer value. how much does it complicate the code? To me it is still worth it, if it's not by that much. I'm quite confident that this is a nice solution on the user level. Maybe someone more technical can evaluate this better on the code level @HooglyBoogly ?
Iliya Katushenock commented 2023-10-10 13:03:23 +02:00
Member
Copy Link

Concentrating the comments, the node declaration should be?:

  • Value input (any type).
  • Seed input (int).
  • Hash output.
Concentrating the comments, the node declaration should be?: - Value input (any type). - Seed input (int). - Hash output.
👍 1
Charlie Jolly commented 2023-10-10 15:15:20 +02:00
Author
Member
Copy Link

Also this seems a bit strange in comparison with the other types. I'd propose the following:
Add a Seed integer field input, same as the random value node has, for all types. That way it makes the inputs and behavior between types more consistent and intuitive imo. Any type will have one input for the value of that type and a second input for the seed. That happens to be 2 integers for the integer type.

This complicates the code somewhat, I'm not clear why you would need a seed here? This could also be achieved by adding an offset to the input value.

To me it's a matter of the consistency of the design and usability. You could argue for the integer input type just as well, that you could just have a single input and offset it, if you need a different hash. I think it's reasonable to avoid such workarounds to allow multiple inputs hashed together. And rather than having those different for every type I think it's more intuitive, more consistent with other existing nodes and nicer design to introduce that second input as a seed integer value. how much does it complicate the code? To me it is still worth it, if it's not by that much. I'm quite confident that this is a nice solution on the user level.
Maybe someone more technical can evaluate this better on the code level @HooglyBoogly ?

@HooglyBoogly @JacquesLucke how do you want me to implement the seed here? The current node is simply wrapping the noise::hash functions at present.

E.g.

For Float3, I can call hash_float(float4(a.x, a.y, a.z, float(seed))) or hash_float(float3(a.x, a.y, a.z) + float(seed))

For Rotation, I could call hash_float(float4(a.x, a.y, a.z, a.w) + float(seed))

For String, do we introduce a new function?

> > >Also this seems a bit strange in comparison with the other types. I'd propose the following: > > > Add a Seed integer field input, same as the random value node has, for all types. That way it makes the inputs and behavior between types more consistent and intuitive imo. Any type will have one input for the value of that type and a second input for the seed. That happens to be 2 integers for the integer type. > > > > This complicates the code somewhat, I'm not clear why you would need a seed here? This could also be achieved by adding an offset to the input value. > > To me it's a matter of the consistency of the design and usability. You could argue for the integer input type just as well, that you could just have a single input and offset it, if you need a different hash. I think it's reasonable to avoid such workarounds to allow multiple inputs hashed together. And rather than having those different for every type I think it's more intuitive, more consistent with other existing nodes and nicer design to introduce that second input as a seed integer value. how much does it complicate the code? To me it is still worth it, if it's not by that much. I'm quite confident that this is a nice solution on the user level. > Maybe someone more technical can evaluate this better on the code level @HooglyBoogly ? @HooglyBoogly @JacquesLucke how do you want me to implement the seed here? The current node is simply wrapping the `noise::hash` functions at present. E.g. For Float3, I can call `hash_float(float4(a.x, a.y, a.z, float(seed)))` or `hash_float(float3(a.x, a.y, a.z) + float(seed))` For Rotation, I could call `hash_float(float4(a.x, a.y, a.z, a.w) + float(seed))` For String, do we introduce a new function?
Iliya Katushenock commented 2023-10-10 15:25:43 +02:00
Member
Copy Link

@CharlieJolly blender::get_default_hash_2<T, int> (from bli)

@CharlieJolly `blender::get_default_hash_2<T, int>` (from `bli`)
Charlie Jolly commented 2023-10-10 15:42:14 +02:00
Author
Member
Copy Link

@CharlieJolly blender::get_default_hash_2<T, int> (from bli)

From previous review from Jacques it was decided to expose noise::hash functions and not default_hash.

> @CharlieJolly `blender::get_default_hash_2<T, int>` (from `bli`) From previous review from Jacques it was decided to expose `noise::hash` functions and not `default_hash`.
Charlie Jolly added 2 commits 2023-10-11 17:19:42 +02:00
Charlie Jolly added 1 commit 2023-10-11 22:52:10 +02:00
47cf039862 Add Seed socket and use this to hash the result 
Use a single Integer input
Use eNodeSocketDatatype not eCustomDataType
Use dynamic declaration and remove updatefunc
Charlie Jolly added 1 commit 2023-10-11 22:54:25 +02:00
Charlie Jolly commented 2023-10-11 22:55:53 +02:00
Author
Member
Copy Link

With Seed socket and one input for each hashable type

image

With Seed socket and one input for each hashable type ![image](/attachments/2843c7f3-015d-4918-977e-e71ee0b42a95)
image.png
44 KiB
👍 1
Simon Thommes commented 2023-10-12 12:45:52 +02:00
Member
Copy Link

That looks great to me now, thank you!

I still haven't actually tested functionality but as that is pretty straightforward, I'll trust the code review on that.

I'm still not 100% sure about exposing the 'new' term Hash in the UI, as we don't use that anywhere else. I just briefly talked to Andy as well and now I do think that since.
The term hash can totally be in the docs and output socket description imo. But since no other socket has that name, it seems like we're introducing a new concept, that doesn't really add that much being called its own term. So I would propose reusing the ID term here.

The only issue is that while hash works nicely as a Verb, ID does not. Personally I'd lean most towards ID from Value, but maybe there's a better name.
Since I might also be misinterpreting the similarity of the terms I'd still like this to be signed off by the dev team.
I'll put it on the agenda of the next meeting.

That looks great to me now, thank you! I still haven't actually tested functionality but as that is pretty straightforward, I'll trust the code review on that. I'm still not 100% sure about exposing the 'new' term `Hash` in the UI, as we don't use that anywhere else. I just briefly talked to Andy as well and now I do think that since. The term hash can totally be in the docs and output socket description imo. But since no other socket has that name, it seems like we're introducing a new concept, that doesn't really add that much being called its own term. So I would propose reusing the `ID` term here. The only issue is that while hash works nicely as a Verb, ID does not. Personally I'd lean most towards `ID from Value`, but maybe there's a better name. Since I might also be misinterpreting the similarity of the terms I'd still like this to be signed off by the dev team. I'll put it on the agenda of the next meeting.
👍 1
Charlie Jolly commented 2023-10-12 12:59:54 +02:00
Author
Member
Copy Link

@SimonThommes I'm not opinionated on the Hash name so I'll leave this for the meeting. One thing to consider is if we add a Hash to Float node at some point that the naming fits in with this.

@SimonThommes I'm not opinionated on the Hash name so I'll leave this for the meeting. One thing to consider is if we add a `Hash to Float` node at some point that the naming fits in with this.
Jacques Lucke commented 2023-10-12 13:50:07 +02:00
Member
Copy Link

We generally use X to Y instead of Y from X when naming nodes.

While I generally like "ID" here, it has one major problem for me. "ID" indicates uniqueness which a hashing function can't guarantee and there will inevitably be collisions where different input values map to the same output. A "hash" is less expected to be unique.

We generally use `X to Y` instead of `Y from X` when naming nodes. While I generally like "ID" here, it has one major problem for me. "ID" indicates uniqueness which a hashing function can't guarantee and there will inevitably be collisions where different input values map to the same output. A "hash" is less expected to be unique.
Hans Goudey requested changes 2024-01-30 04:57:11 +01:00
Hans Goudey left a comment
Member
Copy Link

I don't have a strong opinion on the name. I guess we should talk about it at a module meeting to make a final decision. Either way merge conflicts would need to be resolved before this is merged though.

I don't have a strong opinion on the name. I guess we should talk about it at a module meeting to make a final decision. Either way merge conflicts would need to be resolved before this is merged though.
Jacques Lucke added 2 commits 2024-02-26 13:07:05 +01:00
Simon Thommes commented 2024-02-26 19:26:47 +01:00
Member
Copy Link

Happy to confirm that this PR doesn't have the same floating point precision issue as the random value node (#118207) 👍

Happy to confirm that this PR doesn't have the same floating point precision issue as the random value node (#118207) 👍
🎉 1
Charlie Jolly commented 2024-02-27 12:36:26 +01:00
Author
Member
Copy Link

Sorry, I've not been active on Blender for a while. Thanks for picking this up.

Sorry, I've not been active on Blender for a while. Thanks for picking this up.
This pull request has changes conflicting with the target branch.
  • source/blender/blenkernel/BKE_node.hh
View command line instructions.

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u gn-hash-node:CharlieJolly-gn-hash-node
git checkout CharlieJolly-gn-hash-node
Sign in to join this conversation.
Reviewers
No reviewers
Jacques Lucke
Simon Thommes
Hans Goudey
Labels
Clear labels   
Interest
Alembic

  
Interest
Animation & Rigging

  
Interest
Asset Browser

  
Interest
Asset Browser Project Overview

  
Interest
Audio

  
Interest
Automated Testing

  
Interest
Blender Asset Bundle

  
Interest
BlendFile

  
Interest
Collada

  
Interest
Compatibility
This issue affects/is about backward or forward compatibility

  
Interest
Compositing

  
Interest
Core

  
Interest
Cycles

  
Interest
Dependency Graph

  
Interest
Development Management

  
Interest
EEVEE

  
Interest
EEVEE & Viewport

  
Interest
Freestyle

  
Interest
Geometry Nodes

  
Interest
Grease Pencil

  
Interest
ID Management

  
Interest
Images & Movies

  
Interest
Import Export

  
Interest
Line Art

  
Interest
Masking

  
Interest
Metal

  
Interest
Modeling

  
Interest
Modifiers

  
Interest
Motion Tracking

  
Interest
Nodes & Physics

  
Interest
OpenGL

  
Interest
Overlay

  
Interest
Overrides

  
Interest
Performance

  
Interest
Physics

  
Interest
Pipeline, Assets & IO

  
Interest
Platforms, Builds & Tests

  
Interest
Python API

  
Interest
Render & Cycles

  
Interest
Render Pipeline

  
Interest
Sculpt, Paint & Texture

  
Interest
Text Editor

  
Interest
Translations

  
Interest
Triaging

  
Interest
Undo

  
Interest
USD

  
Interest
User Interface

  
Interest
UV Editing

  
Interest
VFX & Video

  
Interest
Video Sequencer

  
Interest
Virtual Reality

  
Interest
Vulkan

  
Interest
Wayland
Issues relating to Wayland windowing support.

  
Interest
Workbench

  
Interest: X11

Issues relating to Xorg/X11 support.

  
Legacy
Blender 2.8 Project

  
Legacy
Milestone 1: Basic, Local Asset Browser

  
Legacy
OpenGL Error

  
Meta
Good First Issue

  
Meta
Papercut

  
Meta
Retrospective

  
Meta
Security
Issues relating to security: https://wiki.blender.org/wiki/Process/Vulnerability_Reports

  
Module
Animation & Rigging

  
Module
Core

  
Module
Development Management

  
Module
EEVEE & Viewport

  
Module
Grease Pencil

  
Module
Modeling

  
Module
Nodes & Physics

  
Module
Pipeline, Assets & IO

  
Module
Platforms, Builds & Tests

  
Module
Python API

  
Module
Render & Cycles

  
Module
Sculpt, Paint & Texture

  
Module
Triaging

  
Module
User Interface

  
Module
VFX & Video

  
Platform
FreeBSD

  
Platform
Linux

  
Platform
macOS

  
Platform
Windows

  
Priority
High

  
Priority
Low

  
Priority
Normal

  
Priority
Unbreak Now!

  
Status
Archived

  
Status
Confirmed

  
Status
Duplicate

  
Status
Needs Info from Developers

  
Status
Needs Information from User

  
Status
Needs Triage

  
Status
Resolved

  
Type
Bug

  
Type
Design

  
Type
Known Issue

  
Type
Patch

  
Type
Report

  
Type
To Do

No Label
Interest
Alembic
Interest
Animation & Rigging
Interest
Asset Browser
Interest
Asset Browser Project Overview
Interest
Audio
Interest
Automated Testing
Interest
Blender Asset Bundle
Interest
BlendFile
Interest
Collada
Interest
Compatibility
Interest
Compositing
Interest
Core
Interest
Cycles
Interest
Dependency Graph
Interest
Development Management
Interest
EEVEE
Interest
EEVEE & Viewport
Interest
Freestyle
Interest
Geometry Nodes
Interest
Grease Pencil
Interest
ID Management
Interest
Images & Movies
Interest
Import Export
Interest
Line Art
Interest
Masking
Interest
Metal
Interest
Modeling
Interest
Modifiers
Interest
Motion Tracking
Interest
Nodes & Physics
Interest
OpenGL
Interest
Overlay
Interest
Overrides
Interest
Performance
Interest
Physics
Interest
Pipeline, Assets & IO
Interest
Platforms, Builds & Tests
Interest
Python API
Interest
Render & Cycles
Interest
Render Pipeline
Interest
Sculpt, Paint & Texture
Interest
Text Editor
Interest
Translations
Interest
Triaging
Interest
Undo
Interest
USD
Interest
User Interface
Interest
UV Editing
Interest
VFX & Video
Interest
Video Sequencer
Interest
Virtual Reality
Interest
Vulkan
Interest
Wayland
Interest
Workbench
Interest: X11
Legacy
Blender 2.8 Project
Legacy
Milestone 1: Basic, Local Asset Browser
Legacy
OpenGL Error
Meta
Good First Issue
Meta
Papercut
Meta
Retrospective
Meta
Security
Module
Animation & Rigging
Module
Core
Module
Development Management
Module
EEVEE & Viewport
Module
Grease Pencil
Module
Modeling
Module
Nodes & Physics
Module
Pipeline, Assets & IO
Module
Platforms, Builds & Tests
Module
Python API
Module
Render & Cycles
Module
Sculpt, Paint & Texture
Module
Triaging
Module
User Interface
Module
VFX & Video
Platform
FreeBSD
Platform
Linux
Platform
macOS
Platform
Windows
Priority
High
Priority
Low
Priority
Normal
Priority
Unbreak Now!
Status
Archived
Status
Confirmed
Status
Duplicate
Status
Needs Info from Developers
Status
Needs Information from User
Status
Needs Triage
Status
Resolved
Type
Bug
Type
Design
Type
Known Issue
Type
Patch
Type
Report
Type
To Do
Milestone
Clear milestone
No items
No Milestone
Projects
Clear projects
No project
Nodes & Physics
Assignees
Clear assignees
No Assignees
6 Participants
Notifications
Due Date
The due date is invalid or out of range. Please use the format 'yyyy-mm-dd'.

No due date set.

Dependencies

No dependencies set.

Reference: blender/blender#110769
No description provided.
Delete Branch "CharlieJolly/blender:gn-hash-node"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?