When you click Stash Input, the node cooks the input and caches the geometry in a data parameter. The node otherwise doesn’t cook its input and simply outputs the cached geometry. (You can even disconnect the input and the node will continue to output its stored geometry.) You can click Stash Input again to re-cache if the input changes.

This node is conceptually very similar to locking a Null SOP, with several additional features:

• The interaction is slightly more convenient.

• You can modify or overwrite the cached geometry with hou.Geometry objects through scripting.

• You can link the geometry parameter to a parameter on a parent asset. This allows assets to inject geometry into a network without making the contents editable.

• The node never passes the input directly to the output, so Houdini doesn’t see downstream nodes as depending on upstream nodes. This is useful if a live connection would create a dependency cycle.

• The stashed geometry is saved as part of the node instance when you save the scene, so stashing large geometry can bloat the size of the .hip file.

• You can discard the cached geometry by disconnecting the node’s input and clicking Stash Input.

• Because the caching process is manual, it is independent of time. For example, you can run a simulation process (for example, a terrain erosion), cache the final geometry, disconnect the simulation, and then go back to frame 1 and use the geometry in an animation.