Exposing "sliced find functions".

[rob-myers]

Hey,

I'd like to access these functions in class dtNavMeshQuery:

• initSlicedFindPath
• updateSlicedFindPath
• finalizeSlicedFindPath
• finalizeSlicedFindPathPartial

so I can predict requestMoveTarget before it actually runs.

Simply extending recast-navigation.idl and copy-pasting their definitions into your files doesn't work:

• recast-navigation-wasm/src/NavMeshQuery.h
• recast-navigation-wasm/src/NavMeshQuery.cpp

Is there a "standard procedure" for migrating them?

Cheers,
Rob

[rob-myers]

Hmm... maybe I just forgot how to do this.

Looks like I need to write a wrapper C++ function in NavMeshQuery.cpp.

[rob-myers]

This works:

recast-navigation.idl

interface NavMeshQuery {

...
unsigned long initSlicedFindPath(unsigned long startRef, unsigned long endRef, [Const] float[] startPos, [Const] float[] endPos, [Const] dtQueryFilter filter, unsigned long options);

unsigned long updateSlicedFindPath(long maxIter, IntRef doneIters);

unsigned long finalizeSlicedFindPath(UnsignedIntArray path, IntRef pathCount, long maxPath);

unsigned long finalizeSlicedFindPathPartial(UnsignedIntArray existing, long existingSize, UnsignedIntArray path, IntRef pathCount, long maxPath);

...

}

NavMeshQuery.cpp

dtStatus NavMeshQuery::initSlicedFindPath(dtPolyRef startRef, dtPolyRef endRef,
    const float* startPos, const float* endPos,
    const dtQueryFilter* filter, const unsigned int options) {

    return m_navQuery->initSlicedFindPath(startRef, endRef, startPos, endPos, filter, options);
}

dtStatus NavMeshQuery::updateSlicedFindPath(const int maxIter, IntRef *doneIters) {
    return m_navQuery->updateSlicedFindPath(maxIter, &doneIters->value);
}

dtStatus NavMeshQuery::finalizeSlicedFindPath(UnsignedIntArray *path, IntRef *pathCount, const int maxPath) {
    return m_navQuery->finalizeSlicedFindPath(path->data, &pathCount->value, maxPath);
}

dtStatus NavMeshQuery::finalizeSlicedFindPathPartial(UnsignedIntArray *existing, const int existingSize,
    UnsignedIntArray *path, IntRef *pathCount, const int maxPath) {
    return m_navQuery->finalizeSlicedFindPathPartial(existing->data, existingSize, path->data, &pathCount->value, maxPath);
}

NavMeshQuery.h

dtStatus initSlicedFindPath(dtPolyRef startRef, dtPolyRef endRef,
  const float* startPos, const float* endPos,
  const dtQueryFilter* filter, const unsigned int options = 0);

dtStatus updateSlicedFindPath(const int maxIter, IntRef *doneIters);

dtStatus finalizeSlicedFindPath(UnsignedIntArray *path, IntRef *pathCount, const int maxPath);
	
dtStatus finalizeSlicedFindPathPartial(UnsignedIntArray *existing, const int existingSize,
  UnsignedIntArray *path, IntRef *pathCount, const int maxPath);

[isaac-mason]

Sorry - I haven't been as attentive to discussions because my pathfinding focuses have shifted - https://x.com/isaac_mason_/status/1957990076727849415

I'd be happy to accept a PR exposing this but probably won't get to it quickly myself.

[rob-myers]

Wow, looks awesome @isaac-mason .

Will it support DetourCrowd and e.g. offMeshConnections?

[isaac-mason]

I can answer your question with another tweet haha https://x.com/isaac_mason_/status/1959805909439557778
Basically yes and yes. I'm hoping to address some of the limitations I've run into & would have had to modify recastnavigation upstream for. So if you've got any you'd like to add to the laundry list I'm all ears :)

[rob-myers]

I've been handling the discontinuity between (a) nav traversal with separation flag (b) off mesh connections.

For example I added tmid (between tmin and tmax), the time when the agent starts the main segment of an offmesh connection. I also added tScale to permit slowing down when agent stops right after the connection.

This required forking your repo and upstream.

I wonder if such changes will be possible in your new system?

[isaac-mason]

Assuming you're referring to the crowd agent animation logic, I believe so! So far the crowd example is actually just composed of parts of the library, so you could copy in the crowd example and write your own offmesh anim logic.

I don't know what my longer term stance would be on "does crowd logic eventually become a built-in abstraction" but my guess is it'd be a gradual incorporation of parts over time.