Python ImportError Reference — 67+ “Cannot import name X from Y” Fixes (2026)

itsourcecode.com hosts 67+ documented fixes for Python ImportError messages — the error class raised when a library is found but a specific name or sub-module within it can’t be imported. Almost every ImportError in modern Python comes from version conflicts: you have the library installed, but the version you have removed/renamed the specific function or class you’re trying to import. Browse the fixes below by library or scenario.

What is the difference between ImportError and ModuleNotFoundError?

This trips up almost every Python developer:

• ModuleNotFoundError = the whole package isn’t installed. import requests when requests isn’t pip-installed. Fix with pip install requests. We have a separate ModuleNotFoundError reference (198+ posts).
• ImportError = the package IS installed, but a specific name/sub-module within it can’t be imported. from requests import nonexistent. Fix usually means upgrading or downgrading the library to the version that contains that name.

ModuleNotFoundError is actually a subclass of ImportError, introduced in Python 3.6. If you catch ImportError, you catch both.

How to fix any ImportError in 4 steps

1. Read the exact name and source module in quotes. The error has two parts: what you tried to import and which module it should be in. Example: ImportError: cannot import name 'parse_rule' from 'werkzeug.routing' — the name is parse_rule, the module is werkzeug.routing.
2. Check the installed version of the source library. Run pip show <library>. Note the version number.
3. Check when the name was added or removed. Look at the library’s CHANGELOG.md, GitHub release notes, or PyPI page. Often a function/class was removed in a major version (e.g., Werkzeug 2.0+ removed many internal helpers).
4. Pin the working version or update your import. Two paths: (a) Downgrade to the version that has the name: pip install werkzeug<2.0. (b) Find the new import location: from werkzeug.routing.rules import parse_rule (or whatever the new path is). Update your code.

Most common ImportError patterns (in order of frequency)

• Library version conflict — name was removed/renamed in a newer version (60%+ of errors here)
• Python 2 → 3 migration — collections.Mapping moved to collections.abc.Mapping in Python 3.10
• Sub-package reorganization — sklearn moved many internals out of sklearn.externals
• Library reorg between versions — Flask, Werkzeug, Jinja2 frequently move internal helpers
• Beta or development imports — trying to use features from main branch that aren’t in your installed version
• Circular imports — your own code structure has two modules that import each other
• DLL load failed — binary dependency (numpy, scipy) needs system library installed

Featured ImportError fixes by source library

🐍 Werkzeug / Flask / Jinja2 (web framework family)

Werkzeug 2.0+ moved many internal helpers. Flask 2.0+ moved imports. Jinja2 3.1+ removed deprecated names.

• cannot import name ‘parse_rule’ from ‘werkzeug.routing’
• cannot import name ‘cached_property’ from ‘werkzeug’
• cannot import name ‘environmentfilter’ from ‘jinja2’
• cannot import name ‘escape’ from ‘jinja2’
• cannot import name ‘url’ from ‘django.conf.urls’ (Django 4.0+)

🤖 Machine Learning (TensorFlow, PyTorch, sklearn, torchmetrics)

ML libraries reorganize internals frequently between major versions.

• cannot import name ‘container_abcs’ from ‘torch._six’
• cannot import name ‘get_num_classes’ from ‘torchmetrics’
• cannot import name ‘get_config’ from ‘tensorflow’ (eager context)
• cannot import name ‘joblib’ from ‘sklearn.externals’
• cannot import name ‘safe_weights_name’ from ‘transformers’
• numba needs numpy 1.21 or less

🗄️ Database & SQLAlchemy

• cannot import name ‘_rfc_1738_quote’ from ‘sqlalchemy.engine.url’
• cannot import name ‘bigquery’ from ‘google.cloud’
• cannot import name ‘CompilationException’ from ‘dbt.exceptions’

🐍 Python stdlib / typing / collections

Python 3.10+ moved many names between standard library locations. Older code breaks.

• cannot import name ‘Mapping’ from ‘collections’ (Py3.10+)
• cannot import name ‘TypeAlias’ from ‘typing’ (Py3.10+)
• cannot import name ‘ParamSpec’ from ‘typing_extensions’

📦 setuptools / pip / packaging

• cannot import name ‘Feature’ from ‘setuptools’ — Quick Fixes
• cannot import name ‘_unicodefun’ from ‘click’

🔁 Relative imports / circular imports / test modules

• attempted relative import beyond top-level package
• ImportError while importing test module
• Failed to import test module

⚙️ Binary / DLL / native library load failures

Errors at the C-extension level — numpy, scipy, OpenCV, or system libraries.

• numpy.core.multiarray failed to import
• ImportError: dlopen
• Failed to find libmagic, check your installation
• Initialization failed (TensorFlow)
• Unable to import required dependencies (numpy)
• lxml not found, please install it

2026 Updated Guides — featured ImportError fixes

• Failed to find libmagic — Quick Fixes — python-magic install
• cannot import name ‘Feature’ from ‘setuptools’ — setuptools migration
• ImportError: dlopen — macOS native library load

Frequently asked questions

I see ImportError but the module is installed — why?

You’re hitting “cannot import name X from Y” — the package IS installed, but the specific name (function, class, attribute) you’re trying to import isn’t in the version you have. This is different from ModuleNotFoundError. Check the library’s release notes for when the name was added/removed/renamed, then either: (a) install the version that has it, or (b) update your import to the new location.

Why does “from collections import Mapping” fail on Python 3.10?

Python 3.10 removed deprecated imports from collections that were moved to collections.abc in Python 3.3. The deprecation warnings ran for 7 years before removal. Fix: change from collections import Mapping to from collections.abc import Mapping. The same applies to Iterable, Hashable, Callable, and other ABCs.

How do I fix Flask / Werkzeug version conflicts?

Flask 2.0+ requires Werkzeug 2.0+ and removed many internal helpers. If you see cannot import name X from werkzeug..., you have two options: (1) pin compatible versions in requirements.txt: Flask==2.3.2 Werkzeug==2.3.7. (2) Update your code to use the new import locations (the Werkzeug 2.0 migration guide documents most of them). Don’t try to mix old Werkzeug with new Flask or vice versa.

What does “circular import” / “partially initialized module” mean?

Two of YOUR modules import from each other, creating a chicken-and-egg problem. Module A starts loading and imports B; B tries to import something from A; A isn’t fully loaded yet → ImportError. Fix by restructuring: (1) move shared code into a third module both can import from; (2) move the import statement inside the function instead of at module top; (3) refactor to remove the circular dependency.

Why “attempted relative import beyond top-level package”?

You used from ..parent import X in a script you ran directly with python script.py. Relative imports only work when Python knows the package structure — meaning the script must be loaded as part of a package, not run as a standalone file. Fix: run with python -m mypackage.script instead, OR use absolute imports.

How is this reference different from your ModuleNotFoundError reference?

ModuleNotFoundError reference (198+ posts) covers errors where the whole package isn’t installed — “No module named requests” means run pip install requests. This ImportError reference covers errors where the package IS installed but a specific name within it can’t be imported — usually a version conflict. Different root causes, different fixes. See ModuleNotFoundError reference →

How often is this ImportError reference updated?

New posts are added weekly. Existing posts are revised when major library versions ship breaking changes (Flask 3, Werkzeug 3, NumPy 2, etc.). This page was last refreshed in May 2026.

Related error categories

• ModuleNotFoundError Reference — 198+ “No module named X” install errors
• TypeError Reference — 220+ Python and JavaScript TypeError fixes
• AttributeError Reference — 173+ “object has no attribute X” fixes
• ValueError Reference — 100+ pandas/NumPy/sklearn ValueError fixes
• Python Tutorials — beginner-to-intermediate Python lessons

About this ImportError reference

This ImportError reference has been built since 2015 by PIES Information Technology Solutions in Binalbagan, Negros Occidental, Philippines. Every post comes from a real error encountered in production. Used by 12,000+ Python developers monthly across the Philippines, India, the United States, and beyond. If your ImportError isn’t here, send the full traceback to our contact form and we’ll add it.