Sampler silently ignores modifier condition evaluation failures #121

Summary

Modifier condition evaluation failures are currently logged as warnings and treated as "modifier does not apply." This can silently disable intended logic and flatten distributions.

Why This Matters

A malformed or mismatched when condition should not quietly degrade behavior in critical runs.

Current Behavior (Code)

In /Users/adithyasrinivasan/Projects/extropy/extropy/population/sampler/modifiers.py:

• apply_modifiers_and_sample(...) wraps eval_condition(...) in try/except (line ~59-66)
• exceptions are logged and ignored

In /Users/adithyasrinivasan/Projects/extropy/extropy/utils/eval_safe.py:

• default condition failures return False (line ~272)

Expected Behavior

For strict-quality sampling, condition-eval failures should fail validation/sampling (or at least be escalated), not become silent no-ops.

Proposed Fix

1. Add strict mode behavior for modifier condition errors:

• collect structured errors (attribute, modifier index, condition, exception)
• fail sampling when any occur (in strict mode)

2. Keep permissive mode for backward compatibility, but expose counts in output.

3. Improve diagnostics:

• include offending condition string and referenced unknown values/names where possible.

Tests

• malformed condition causes strict sampling failure
• permissive mode logs and continues with explicit warning count
• valid conditions unaffected

Acceptance Criteria

• No silent modifier disablement in strict runs.
• Users can reliably detect condition-eval failures before trusting sample outputs.

Pipeline Impact

Prevents hidden rule dropouts in step 3 and reduces iterative debugging.