Naming constraints break compatibility with existing datasets

[aeisenbarth]

_check_valid_name implemented stricter naming constraints in https://github.com/scverse/spatialdata/commit/137e1e06c946800599d55c45f18fe8a6a1fb06eb. We already have existing SpatialData datasets where . is used as separator for naming components with different meanings, like Slide1.A2.0.pre_maldi. When loading them with spatialdata 0.2.1, we get the error "Name must contain only alphanumeric characters, underscores, and hyphens". I would understand for Unicode characters, or characters reserved for SpatialData itself (/) or characters problematic on a subset of supported platforms (:, ?, \…).

1. Is there a reason for not allowing . anymore? Will it be used by SpatialData itself, or the NGFF specification?

2. What would be a solution, can . be allowed again or should we change the way how we create new datasets and adjust all existing datasets to the new naming constraints?

   • _ feels more like a replacement for space within a naming component
   • - feels more fitting for joining parts of a naming component, not separating them.
   • What allowed character would convey the meaning of separating components of a name?