avtWorldSpaceToImageSpaceTransform - wrong spatial bounds

[ARSanderson]

avtWorldSpaceToImageSpaceTransform is a "helper" transform for rendering. However, the handling of the resulting spatial bounds are not updated. They are only marked as being InvalidateSpatialMetaData.

This action maybe correct given this class is a "helper" transform for point sampling. However, avtResampleFilter also uses these results.

The issue is further compounded by the fact within avtWorldSpaceToImageSpaceTransform rectilinear grids can be passed through. And the view may be further transformed if it is for a perspective view.

As such, the class and its usage is rather bastardized. Which is probably why InvalidateSpatialMetaData was set and not attempt was made to set the actual bounds.

I have no solution but felt it needs to be documented because it caused down stream issue for some test code I was creating. To my knowledge, it is not yet causing any down stream issues.

[markcmiller86]

In looking at avtSampleFilter code, I see the call to avtWorldSpaceToImageSpace transform is embedded in that and it appears to be there to satisfy the interface of avtSamplePointExtractor.

I am thinking that avtSamplePointExtractor is designed to work in some kind of normalized space of the dataset, independent of any view related stuff. But, it may have all originally been integrated with VisIt within the context of a ray-tracing algorithm and so the classes in use here operate kinda sorta within that context.

In avtSamplePointExtractor, just before calling avtWorldSpaceToImageSpace transform to do the sample point extraction, there is a comment that the pipeline is being pinched off so that what is done there doesn't propagate up the pipeline. That sounds highly indicative to me that were using avtWorldSpaceToImageSpace transform locally only to do work specific to this step in the pipeline and that changes from it should NOT propagate. But, maybe parts of the contract (e.g. spatial bounds validity) are erroneously propagating?

Just thinking out loud.

[brugger1]

@ARSanderson is this causing you problems?

[ARSanderson]

Similar to #17111 (comment) I do no need to use it in my pipeline any longer, so it is not a bottleneck.

But the question remains, when a filter InvalidateSpatialMetaData what does that really mean to down stream filters? And if invalidated but the filter can update the spatial bound should it not do that?? avtTransform::UpdateDataObjectInfo updates the spatial bound but yet also InvalidateSpatialMetaData. So what does that mean to down stream filters?