Fixes

PragyaLint doesn't just report dead code — its --fix engine can actually remove it. Fixes are confidence-gated and explicit: nothing is modified unless you ask.

The safety model

Three guarantees keep the fixer safe:

Explicit — the fixer never runs automatically; you must pass --fix.
Confidence-gated — by default only high-confidence findings are fixed. Lower-confidence fixes require an explicit opt-in.
Protected exports — names listed in __all__ (or otherwise re-exported) are never touched without --force.

Usage

# Dry-run: preview changes (modifies nothing)
pragyalint --fix --dry-run

# Apply high-confidence fixes only (default)
pragyalint --fix

# Include medium-confidence fixes
pragyalint --fix exports --confidence medium+

# Include everything, overriding protected exports
pragyalint --fix all --confidence low+ --force

Fix types

Target	Default	First fix confidence	What it removes
`files`	high	on	Unreachable module files (never reaches entry points / `__init__.py`).
`imports`	high	on	Unused imports; trims unused names from multi-name imports (`from m import a, b` → `from m import a`).
`exports`	medium	off	Unused public exports; respects `__all__`, protected without `--force`.
`locals`	medium	off	Unused local functions/classes/variables.

Confidence thresholds

The --confidence flag sets the minimum confidence to fix:

high — only high (default)
medium+ — medium and high
low+ — low, medium, and high
all — everything

Recommended workflow

# 1. Review what would be removed
pragyalint --fix all --confidence medium+ --dry-run

# 2. Apply just the high-confidence removals
pragyalint --fix

# 3. Run your test suite
pytest

# 4. Commit the changes
git add -A && git commit -m "chore: prune dead code"

Warning: always review the dry-run output and run your tests after fixing. Medium/low-confidence fixes (exports, locals) can be wrong in code that uses reflection, dynamic dispatch, or plugin registration.