What are the necessary elements for creating a new node

[PlethoraChutney]

Let's say, hypothetically, that someone wanted to write a new node that let you select atoms/residues/what have you by their ChimeraX selection string. That person might be confused with where to start writing a node.

It seems that the business logic for a node involves building a node tree in nodes.py. For instance, the chain selection logic lives in nodes.py, lines 446--534.

To get this new node to appear, you then need to add two things to ui.py. First, a layout_function subclass which describes how the menu looks, and a bpy.types.Operator subclass which describes what happens. This would be menu_chain_selection_custom and MOL_OT_Chain_Selection_Custom, respectively.

This is where it gets confusing for me. Modifying what you've got in (e.g.) menu_chain_selection_custom does modify what is displayed in the addon, but trying to add a new box or whatever doesn't work. Working on PDB caching made me wonder if I also would need to add a bpy.props.StringProperty to load.py?

But then I looked for MOL_color_element and I can't even find where that's defined at all!

I guess my question is: for any given node, what properties need to be added to what files for what functionality. I've tried googling around but it seems like a lot of blender addons are just single-file __init__.py things, which muddy the waters.

Thank you in advance!

[BradyAJohnston]

The majority of the nodes don't involve python at all, but because GN is still in its infancy, things that involved loops and text parsing all require some calls to python scripts.

Creating New Nodes

There are two types of nodes that I have created so far.

The majority of the nodes are pre-made by hand, and live inside of the assets/node_append_file.blend file that is in the repo. This gets shipped with the addon and when the user clicks the button the pre-made node is just appended to the file and then added to the node tree.

The other types of nodes are the ones that need to be structure-specific, so basically anything that would require a loop. You can't loop currently inside of GN, so in creating those nodes I have to make the loops myself when generating them using python. Those are the majority of the nodes defined inside of nodes.py. This enables the chain selection nodes that are the correct size for the number of chains for the structure, and to have the names correct etc.

MOL_color_element is one of the pre-defined nodes (for the first 20 elements, just for convenience) so it doesn't exist inside of the nodes.py. Selections for chains & ligands, as well as colouring by chain, are all created procedurally based on the information from the structure for the number and names of the chains.

A new custom selection string node

There isn't yet useful text parsing inside of GN, so the text needs to be parsed by python then a custom node group created. If you look at resid_multiple_selection() function in nodes.py this does exactly this. You can input a custom resid selection string, and it will build a custom node tree that creates the selection. It can be used by clicking MolecularNodes -> Selections -> Res ID and a custom selection input pops up, which allows for the custom numerical ranges and the resulting nodes that are created:

If you want to implement a new custom selection that would mimic that of ChimeraX, I would certain welcome it but it might be a bit finnicky. You would have to write your own text parses to take the inputted text, then createa new node group which combines the boolean selections of chain_id, res_id and res_name etc. You could copy the operator MOL_OT_Residues_Selection_Custom which provides the popup then executes the function for creating the node.

It's kind of all over the place, and ideally at some point text parsing will be possible inside of GN which will avoid this back and forth, but currently I believe this is what we are limited to doing.

Hopefully this clears up some of how the logic of the nodes work. Happy to talk about it plenty more!

[PlethoraChutney]

Ah, awesome. Thanks so much!