leith

Okay. I should post a useful update here.

A TOp network in a SOp does run. Or, can sometimes be forced to run inside Unreal's Houdini Engine. Proof follows:

I create a HDA that is a Geo object, that contains a pig_head node, connected to a TOp-SOp-Network node, connected to an output node.

When I drag that HDA into Unreal and then drag it into an unreal level, the Object SOp network executes, including the TOp network. Subsequent requests to ‘recook’ the asset, fail to re-run the TOp. However a request to “rebuild asset” does result in the TOp being run again.

In my case, my TOp network is just copying a text file from one place on my desktop to another, to prove that the TOp network is running, or not running.

Anyhow, this isn't the full integration, where a TOp network executes as a first class citizen within Unreal Engine, and dynamically generates unreal actors. I'd assume that's under development. Maybe? But we're not completely without TOp capability as it were. Even if it's a hack.

I suspect the reason the ‘recook’ fails but the ‘rebuild’ succeeds, is because the dependency graph isn't being marked ‘dirty’ by the recook request. I played around a bit with ways of randomizing the input to the TOp-SOp-Network node to force it to consider itself dirty every time, with no luck. But I suspect there are ways

Same general question here. I went so far as to dig through the GitHub source-code. I browsed the latest commits and only found some basic references to PDG updates. Which I think might have been simple updates to the HAPI files, but not actual implementation of PDG TOPs just yet? Or am I reading it all wrong?