Bug in numexpr v2.8.5 - Fix in GeoIPS or pin to numexpr 2.8.4 until further notice

[jsolbrig]

Requested Update

Description

There is a bug in numexpr 2.8.5 and above. They were attempting to fix some security issues involving the use of eval() and managed to break some behavior that previously worked. Our specific problem seems to relate to use of scientific notation in our equations.

For now, we will need to pin the numexpr version to 2.8.4 until further notice and track the changes in numexpr. For more information on the bug see pydata/numexpr#442.

The image below shoes the result of this problem.

Background and Motivation

• [ OPTIONAL: Provide additional detailed information and background related to this issue ]

Alternative Solutions

• [ OPTIONAL: If appropriate, list alternative fixes that could be considered]

Environment

• [ IF APPROPRIATE: Corresponding repository versions ]

Code to demonstrate issue

• [ IF APPROPRIATE: Can include steps here to identify/demonstrate bug fix or desired feature ]

Checklist for Completion

• <!--- Include a list of steps that must be complete in order to close this Issue --->