Context

linkml-map's eval_utils.py is a fork of linkml_runtime.utils.eval_utils with some improvements (fixed None handling bug, added eval_expr_with_mapping()), but the expression evaluator is still missing basic operators that users would reasonably expect.

The long-term goal (per #98) is to unify the expression language across the LinkML ecosystem. Before upstreaming anything, we should make the evaluator here actually good and well-tested.

Missing operators and features

The safe evaluator currently lacks:

• != (not equal) — despite == being supported
• and, or (logical operators via ast.BoolOp)
• not (logical negation via ast.UnaryOp / ast.Not)
• in, not in (membership testing)
• % (modulo)
• // (floor division)

Other improvements

• Remove dead Python 3.7 compat branches (ast.Num, ast.Str, ast.NameConstant, ast.Index) — the project requires 3.9+
• Add comprehensive tests for all operators and edge cases
• Document what the expression language supports

Out of scope (for now)

• List comprehensions, multi-statement blocks, assignment — these require unrestricted_eval=True and the asteval fallback. That's a separate concern.
• Upstreaming to linkml/linkml — do this once the evaluator is solid.

Related

Sub-issue of #98