When making clusters, user objects, or custom components how should my output data be formatted?

Via David Rutten on this discussion

"The rules are quite complex and sometimes change per version, but there are two important reasons for the behaviour:

1. The ways datatrees change should be predictable per component. I.e. if you provide a single number the output should not differ fundamentally from when you provide two numbers as input.
2. A component should not destroy information stored in datatrees, as they are a sort of history record telling you where data came from.

The current logic is outlined in this document.

In short:

• If all inputs have item access and all outputs have item access then the datatree layout is copied from the master input parameter.
• If all inputs have list access and all outputs have list access, then the datatree layout is copied from the master input parameter.
• If some inputs have item while some outputs have list access, those outputs will have their trees grown by one layer.
• Input and outputs which have tree access are completely at the mercy of the component and their behaviour will differ on a per component basis."