import-analyzer-py

Warning

This project is in early development and will undergo significant changes. The API is not stable and may change without notice.

A Python import analyzer with cross-file analysis, unused import detection, circular import warnings, and autofix.

Comparison with Other Tools

Feature	This tool	Ruff	Autoflake	Pyflakes	Pylint	Unimport
Detect unused imports	✅	✅	✅	✅	✅	✅
Autofix	✅	✅	✅	❌	❌	✅
Cross-file analysis	✅	❌	❌	❌	❌	❌
Re-export tracking	✅	❌¹	❌	❌	❌²	❌
Cascade detection	✅	❌	❌	❌	❌	❌
Circular import warnings	✅	❌	❌	❌	❌	❌
Unreachable file warnings	✅	❌	❌	❌	❌	❌
Indirect import fix	✅	❌	❌	❌	❌	❌
Respects __all__	✅	✅	⚠️³	✅	✅	✅
noqa: F401 support	✅	✅	✅	❌⁴	✅⁵	✅
Full scope analysis (LEGB)	✅	✅	⚠️⁶	⚠️⁶	✅	✅
String annotations	✅	✅	✅	✅	✅	✅
TYPE_CHECKING blocks	✅	✅	✅	⚠️⁷	✅	✅
Type comments (# type:)	❌	✅	✅	✅	✅	✅
Redundant alias (x as x)	❌	✅	❌	❌	❌	❌
Star import suggestions	❌	❌	❌	❌	❌	✅
Redefinition warnings	❌	✅	❌	✅	✅	❌
Speed	Moderate	🚀 Fast	Moderate	Fast	Slow	Moderate

^{¹ Ruff suggests import X as X for __init__.py but doesn't track actual cross-file usage
² Pylint skips __init__.py by default but doesn't track actual re-export consumers
³ Autoflake only uses __all__ to skip star import expansion, not to preserve re-exports
⁴ Pyflakes has no noqa support; Flake8 adds it as a wrapper layer
⁵ Pylint uses # pylint: disable=unused-import
⁶ Autoflake/Pyflakes have basic scope analysis but miss some LEGB edge cases
⁷ Pyflakes treats TYPE_CHECKING as normal conditional code (works but not explicit)}

What makes this tool unique

Cross-file analysis — This is the only tool that follows imports from your entry point and tracks which imports are actually used by other files. This enables:

• Re-export preservation: If utils.py imports List and main.py does from utils import List, the import in utils.py is correctly identified as used
• Cascade detection: When removing an unused import makes another import unused, this tool finds all of them in a single pass (note: imports in __all__ are always preserved as public API)
• Indirect import fix: When you import a name from a re-exporter instead of its original source, this tool can rewrite the import to use the direct source
• Circular import detection: Warns about import cycles in your codebase
• Unreachable file detection: Identifies files that become dead code after fixing imports

Features we don't have (yet)

Based on analysis of other tools' source code:

• Type comments: # type: int style annotations (PEP 484) are not parsed
• Redundant alias detection: import X as X as explicit re-export marker (Ruff feature)
• Star import suggestions: Suggesting specific names to replace from x import * (Unimport feature)
• Redefinition warnings: Warning when an import is reassigned before use (Pyflakes feature)

Installation

pip install import-analyzer-py

Or install from source:

git clone https://github.com/cmyui/import-analyzer-py
cd import-analyzer-py
pip install .

pre-commit

Add to your .pre-commit-config.yaml:

repos:
  - repo: https://github.com/cmyui/import-analyzer-py
    rev: v0.1.5 # or latest version
    hooks:
      # Cross-file analysis (recommended) - analyzes entire project
      - id: import-analyzer

      # Or for faster single-file analysis on changed files only:
      # - id: import-analyzer-single-file

The cross-file hook runs with all features enabled by default: --fix-unused-imports --fix-indirect-imports --warn-implicit-reexports --warn-circular --warn-unreachable

To customize:

hooks:
  - id: import-analyzer
    args: [.] # check only, no warnings
  - id: import-analyzer
    args: [., --fix-unused-imports] # fix but no warnings

Usage

Cross-file mode (default)

Cross-file mode follows imports from an entry point and tracks re-exports across your codebase. This prevents false positives when imports are used by other files.

# Analyze from entry point (follows imports)
import-analyzer main.py

# Analyze entire directory
import-analyzer src/

# Fix all unused imports (including cascaded ones)
import-analyzer --fix-unused-imports main.py

# Fix indirect imports (rewrite to use direct sources)
import-analyzer --fix-indirect-imports main.py

# Fix indirect imports including same-package re-exports (stricter)
import-analyzer --fix-indirect-imports --strict-indirect-imports main.py

# Warn about implicit re-exports (imports used by other files but not in __all__)
import-analyzer --warn-implicit-reexports main.py

# Warn about circular imports
import-analyzer --warn-circular main.py

# Warn about files that become unreachable after fixing
import-analyzer --warn-unreachable main.py

# Quiet mode (summary only)
import-analyzer -q main.py

Single-file mode

For simple use cases or when you want to analyze files independently:

# Check files independently (no cross-file tracking)
import-analyzer --single-file myfile.py

# Check multiple files
import-analyzer --single-file src/*.py

Exit codes

Code	Meaning
0	No unused imports found (or --fix-unused-imports was used)
1	Unused imports found

Features

Cross-file analysis

• Re-export tracking: Imports used by other files are preserved
• Cascade detection: Finds all unused imports in a single pass, even when removing one exposes another
• Indirect import fix: Rewrites imports that go through re-exporters to use the original source directly
• Circular import detection: Warns about import cycles
• Implicit re-export warnings: Identifies re-exports missing from __all__
• Unreachable file detection: Warns about files that become dead code after fixing imports

Single-file analysis

• Detects unused import X and from X import Y statements
• Handles aliased imports (import X as Y, from X import Y as Z)
• Recognizes usage in:
  • Function calls and attribute access
  • Type annotations (including forward references / string annotations)
  • Decorators and base classes
  • Default argument values
  • __all__ exports
• Skips __future__ imports (they have side effects)
• Respects # noqa: F401 comments (matches flake8 behavior)
• Full scope analysis with LEGB rule:
  • Correctly handles function parameters that shadow imports
  • Handles class scope quirks (class body doesn't enclose nested functions)
  • Supports comprehension scopes and walrus operator bindings
  • Respects global and nonlocal declarations

Directory exclusions

The tool automatically skips common non-source directories:

• Virtual environments: .venv, venv, .env, env
• Build artifacts: build, dist, *.egg-info
• Cache directories: __pycache__, .mypy_cache, .pytest_cache, .ruff_cache
• Version control: .git, .hg, .svn
• Other: node_modules, .tox, .nox, .eggs

Autofix

• Safely handles empty blocks by inserting pass
• Partial removal from multi-import statements
• Handles semicolon-separated statements
• Handles backslash line continuations

Examples

Single-file example

Before:

import os
import sys  # unused
from typing import List, Optional  # List unused
from pathlib import Path

def get_home() -> Optional[Path]:
    return Path(os.environ.get("HOME"))

After (--fix-unused-imports):

import os
from typing import Optional
from pathlib import Path

def get_home() -> Optional[Path]:
    return Path(os.environ.get("HOME"))

Cross-file re-export example

# utils.py
from typing import List  # NOT unused - re-exported to main.py

# main.py
from utils import List
x: List[int] = []

Running import-analyzer main.py correctly preserves the List import in utils.py because it's used by main.py.

Cascade detection example

# main.py
from helpers import List  # unused - not used locally

# helpers.py
from utils import List    # becomes unused when main.py's import is removed

# utils.py
from typing import List   # becomes unused when helpers.py's import is removed

Running import-analyzer --fix-unused-imports main.py removes all three imports in a single pass.

Note: Imports listed in __all__ are always considered "used" (public API declaration) and won't be removed by cascade detection. This matches the behavior of flake8, ruff, and autoflake.

Indirect import fix example

The tool can fix two patterns of indirect imports:

Pattern 1: from-import style

# core.py
CONFIG = {...}  # Original definition

# utils/__init__.py
from core import CONFIG  # Re-exports CONFIG

# app.py (before)
from utils import CONFIG  # Indirect - importing from re-exporter

Running import-analyzer --fix-indirect-imports app.py rewrites the import to use the direct source:

# app.py (after)
from core import CONFIG  # Direct - importing from source

Pattern 2: import + attribute access style

# logger.py
LOGGER = Logger()  # Original definition

# models/__init__.py
from logger import LOGGER  # Re-exports LOGGER

# app.py (before)
import models
models.LOGGER.info("hello")  # Indirect - accessing via re-exporter

Running import-analyzer --fix-indirect-imports app.py:

# app.py (after)
import models
import logger
logger.LOGGER.info("hello")  # Direct - accessing via source

This also works with nested access like pkg.internal.LOGGER and handles aliases throughout the re-export chain. Function-local and class-body imports are preserved in their original scope.

noqa comments

The tool respects # noqa comments matching flake8 behavior:

import os  # noqa: F401  - kept (F401 = unused import)
import sys  # noqa       - kept (bare noqa suppresses all)
import re  # noqa: E501  - flagged (wrong code)

For multi-line imports, noqa applies per-line:

from typing import (
    List,  # noqa: F401  - kept
    Dict,  # flagged (no noqa)
)

Known Limitations

• Star imports: from X import * cannot be analyzed statically
• Dynamic imports: importlib.import_module() calls are not tracked
• Namespace packages: PEP 420 namespace packages are not supported

Development

Setup

git clone https://github.com/cmyui/import-analyzer-py
cd import-analyzer-py
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Running tests

# Run with pytest
pytest tests/ -v

# Run with tox (multiple Python versions)
tox

# Run with coverage
tox -e py

Project structure

.
├── import_analyzer/
│   ├── __init__.py          # Public API exports
│   ├── __main__.py          # Entry point for python -m
│   ├── _main.py             # CLI and orchestration
│   ├── _data.py             # Data classes (ImportInfo, ModuleInfo, etc.)
│   ├── _ast_helpers.py      # AST visitors for import/usage collection
│   ├── _detection.py        # Single-file detection logic
│   ├── _autofix.py          # Autofix logic
│   ├── _resolution.py       # Module resolution (resolves imports to files)
│   ├── _graph.py            # Import graph construction
│   ├── _cross_file.py       # Cross-file analysis with cascade detection
│   └── _format.py           # Output formatting
├── tests/
│   ├── detection_test.py
│   ├── aliased_imports_test.py
│   ├── shadowed_imports_test.py
│   ├── scope_analysis_test.py
│   ├── special_imports_test.py
│   ├── type_annotations_test.py
│   ├── autofix_test.py
│   ├── file_operations_test.py
│   ├── cli_test.py
│   ├── resolution_test.py   # Module resolution tests
│   ├── graph_test.py        # Import graph tests
│   ├── cross_file_test.py   # Cross-file analysis tests
│   └── format_test.py       # Output formatting tests
├── pyproject.toml
├── tox.ini
└── .github/workflows/ci.yml

License

MIT