A common way to create a ParaView Catalyst script is to use Paraview, load a simulation result file, setup a pipeline, add extractors, and use Export Catalyst Script.
This creates a python script with a pipeline starting with a file reader. When ParaView Catalyst sees the file reader, it looks at the registration name of the file reader and if that name matches a channel name in the conduit tree, then Catalyst ignores the file reader and uses a conduit source instead.
This mechanism is confusing to a lot of my users. In particular, there is no reason for them to know that they should rename the file reader to modify the registration name. Most of the time, they do not change it. When running ParaView Catalyst, every rank loads the result file, the machine runs out of RAM, and the user gets an awful error that does not help them know what they did wrong.
The fundamental issues are that:
1- ParaView cannot know what the registration name produced by the simulation code is going to be, since that decision is made by the simulation code.
2- ParaView Catalyst cannot know which source in the pipeline the user wants replaced by the conduit source.
I don’t have a good idea on how to fix this, which is why I’m posting in discord rather than opening and issue.
I think that what I would like to see is for ParaView to know whether it’s being run by Catalyst. In Catalyst mode, I’d like to see either a warning or an error noting that no sources have been replaced by the conduit source. The message would expose all the registration names in the script alongside those found in the conduit tree, and note that non match.
This behavior could be triggered by a checkbox in the Save Catalyst State Options, on by default. Code for this could be exported in the script (rather than being a new ParaView Catalyst feature).
Any thoughts on this or better ideas?