String attribute as component on pointcloud

Firstly use 2 featurereaders. The first Path and File Directory reader points to your raster folder. Then connect the PATH outport port into another featurereader where the dataset is set to the Attribute path_windows. 

The PointCloudComponentAdder allows you to define the DataType, along with AttributeManagers

example - make sure you use the Feature/Record Information tab when inspecting the Components 

Firstly use 2 featurereaders. The first Path and File Directory reader points to your raster folder. Then connect the PATH outport port into another featurereader where the dataset is set to the Attribute path_windows. 

The PointCloudComponentAdder allows you to define the DataType, along with AttributeManagers

example - make sure you use the Feature/Record Information tab when inspecting the Components 

 

OK, using the PointCloudComponentAdder did work to add the attributes as components. It also alerted me to the fact that one of my attributes/components, which was defined to be of an Int8 type, can’t be a NULL value. The error message was formatted strangely though: @AddPointCloudComponents: Parameter '<value>' has value '' which is invalid for component type 'Int8'.  Please ensure the specified value and type are compatible. No error or warning messages from the PointCloudCombiner.

Thanks ​@j.botterill 

This does not however explain why I can’t do the same action in the PointCloudCombiner with the Attributes to Preserve as Componenets option. According to the docs this should do exactly the same as the PointCloudComponentAdder. Any thoughts on this ​@DanAtSafe ?

Hi ​@esietinga, thanks for posting! Happy to chime in here. I don’t think point clouds support true NULL values in their components, so we might need to use a substitute value to represent "no data.”

Mapping NULL to 0 using NullAttributeMapper is a natural first instinct, but it comes with a tradeoff as you'd lose the ability to store a real 0 value in that component.
A better approach is to choose a different value that can't be mistaken for a real measurement. Depending on your component type, good options are:

Int16 → use -1
Real32 → use NaN (Not a Number)

Both of these are unlikely to conflict with real data, and NaN in particular behaves like a true NULL since it represents an undefined numeric state rather than an actual value. 

I think there is an issue regarding the PointCloudCombiner, I will set aside some time to test and reach out with an update. Let me know if there are any further questions. Happy to help, Kailin.

Hello ​@esietinga, the PointCloudCombiner does not appear to preserve components when the source dataset is raster or point cloud geometry type. Moving forward, we can refer to this as FMEENGINE-98600. Let me know if you have any further questions, Kailin.