Workaround: Pin pandas <3 due to neutronbraggedge incompatibility

[KedoKudo]

Summary

iBeatles v1.2.1 includes a workaround that pins pandas <3 in the conda recipe to avoid a crash on startup.

Root Cause

The neutronbraggedge package uses positional indexing with [] on pandas Series objects, which fails in pandas 3.0:

# neutronbraggedge/material_handler/retrieve_material_metadata.py:68
self.lattice = float(_metadata[0])  # Fails in pandas 3.0

In pandas 3.0, [] is strictly for label-based access. Positional access requires .iloc[]:

self.lattice = float(_metadata.iloc[0])  # Correct

Affected Files in neutronbraggedge

1. retrieve_material_metadata.py:68,72 - Uses _metadata[0] and _metadata[1]
2. braggedge.py:249 - if file_type is 'csv': (SyntaxWarning, should use ==)
3. tof.py:61 - if units is not 's': (SyntaxWarning, should use !=)

Proper Fix

The proper fix should be implemented in the upstream neutronbraggedge package:

• Repository: https://github.com/ornlneutronimaging/BraggEdge
• Replace [] indexing with .iloc[] for positional access
• Replace is/is not with ==/!= for string comparisons

Workaround in iBeatles

Until the upstream fix is released, iBeatles pins pandas <3 in conda.recipe/meta.yaml.

Related

• See upstream issue: (to be created in BraggEdge repo)

[KedoKudo]

Upstream issue created: ornlneutronimaging/BraggEdge#5