ultraimport

Get control over your imports -- no matter how you run your code.

Overview | Installation | Quickstart | Examples | Documentation

Overview

Features

• Import any file from the file system as Python code:
  • Works independent of your sys.path
  • Works independent of your current working directory
  • Works independent of your top-level package
  • Works no matter if you run your code as a module or as a script
  • Does not care about __init__.py files
  • Use relative or absolute pathes
  • Use relative imports in scripts
• Dynamically wrap your code in a virtual namespace package
• Preprocess code for optimizations (see example)
• Recursively rewrite subsequent relative import statements (see example)
• Dependency injection (see example)
• Lazy loading (lazy imports for modules and callables)
• Make imports work in code you cannot modify (e.g., auto-generated files, legacy code)
• Fix circular imports through lazy imports or dependency injection
• Fix the error: ImportError: attempted relative import with no known parent package
• Fix the error: ValueError: attempted relative import beyond top-level package
• Better error messages

General Approach

ultraimport is built around an own implementation of the importlib.machinery.SourceFileLoader. This allows to take a different approach on finding code while still being compatible and integrate nicely with the normal Python import machinery. It also allows for some advanced use cases like virtual namespaces, pre-processing, lazy loading, dependency injection and last but not least much better error messages.

Is ultraimport supposed to replace the normal import statement?

No! You will continue to use the builtin import statements to import 3rd party libraries which have been installed system wide. ultraimport is meant to import local files whose locations you control because they are located relatively to some other files.

Issues

Currently, there is no integration with any Python language server for code completion in your IDE. Code completion in the Python REPL works as expected.

Installation

Install system wide:

pip install ultraimport

Install a local development version:

git clone https://github.com/ronny-rentner/ultraimport.git
pip install -e ./ultraimport

Quickstart

Note: You can find this quickstart example and others in the examples/ folder.

The quickstart folder looks like this:

quickstart
├── cherry.py
├── readme.md
├── red
│   ├── cherry.py
│   └── strawberry.py
├── run
│   └── run.py
└── yellow
    ├── banana.py
    └── lemon.py

The entry point is the script run.py located in the quickstart/run folder. If you want, you can directly execute the example script by running:

python /path/to/quickstart/run/run.py

Inside the run.py script, first we import ultraimport:

import ultraimport

1) Import from parent folder

This example shows how to import the Python module cherry.py from the parent folder. Note that __dir__ in the file path refers to the parent folder of the file that is executing the import. In this case, run.py is executing the import and it is located in a folder run and thus __dir__ refers to the run folder.

cherry = ultraimport('__dir__/../cherry.py')
# <module 'cherry' from '/home/ronny/Projects/py/ultraimport/examples/quickstart/cherry.py'>

2) Import from sibling folder

This example shows how to import another Python module from a sibling folder.

other_cherry = ultraimport('__dir__/../red/cherry.py')
# <module 'cherry' from '/home/ronny/Projects/py/ultraimport/examples/quickstart/red/cherry.py'>

3) Import single object

Import the Cherry object from cherry.py and alias it to the name my_class. You could also provide a list of strings instead of a single string if you want to import multiple objects.

my_class = ultraimport('__dir__/../red/cherry.py', 'Cherry')
# <class 'cherry.Cherry'>

4) Ensure type of imported object

You can make sure my_class is actually the type you expect, a class, and my_string is a string, otherwise a TypeError is thrown.

my_class, my_string = ultraimport('__dir__/../cherry.py', { 'MyClass': type, 'some_string': str })
# <class 'cherry.MyClass'>, "I am a string"

5) Import all objects

Using the known special string '*' allows to import all objects.

objs = ultraimport('__dir__/../cherry.py', '*')
# <class 'cherry.MyClass'>

6) Add imported objects to a namespace

You can also provide any dict as a namespace. A common value to use is globals(). If you set add_to_ns=True, the imported objects are added to the local scope of the caller.

ultraimport('__dir__/../cherry.py', '*', add_to_ns=globals())
# <class 'cherry.MyClass'>

7) Give imported module a known parent package

The next import would fail because the imported banana.py contains another relative import in line 1: from .. import cherry as relatively_imported_cherry.

try:
    banana = ultraimport('__dir__/../yellow/banana.py')
except Exception as e:
    # <class 'ultraimport.ultraimport.ExecuteImportError'>
    pass

If you would try to excute banana.py directly, you'd get the famous ImportError: attempted relative import with no known parent package error. To solve this error, just give the banana.py module a known parent package by using the package parameter. The relative import from above has two dots, so it means go two levels up. Thus we need to give our banana.py module at least two levels of parent packages by using package=2.

banana = ultraimport('__dir__/../yellow/banana.py', package=2)
# <module 'quickstart.yellow.banana' from '/home/ronny/Projects/py/ultraimport/examples/quickstart/yellow/banana.py'>

As you can see above, the package was derived from the path automatically according to the package parameter value.

8) Embed module in a virtual namespace package

You can also provide a string for the package parameter to define the name of the package. In this case we create a package called fruit pointing at the parent directory of cherry.py, the file that is being imported.

cherry = ultraimport('__dir__/../red/cherry.py', package='some.fruit')
# <module 'some.fruit.cherry' from '/home/ronny/Projects/py/ultraimport/examples/quickstart/red/cherry.py'>

9) Integrate with normal imports

After creating the fruit namespace package as a side effect of the import, you can use it to do classical imports. Remember that the fruit package points to the directory red.

from some.fruit.strawberry import Strawberry
# <class 'some.fruit.strawberry.Strawberry'>

10) Create virtual namespace package

You could also explicitly create a virtual namespace pointing to the directory 'yellow'. Note that __dir__ in the path below refers to the parent folder of the file that is executing the import.

yellow_ns = ultraimport.create_ns_package('yellow', '__dir__/../yellow')
# <module 'yellow' (<_frozen_importlib_external._NamespaceLoader object at 0x7fba8de36920>)>

For further imports, the package_name yellow must be used as provided as the first argument.

from yellow import lemon
# <module 'yellow.lemon' from '/home/ronny/Projects/py/ultraimport/examples/quickstart/yellow/lemon.py'>

Let's add some other module cherry.py from a different directory to our virtual package. After you have added the module to the package, you can again use normal Python imports to access it.

ultraimport('__dir__/../red/cherry.py', package='yellow')
from yellow.cherry import Cherry
# <class 'yellow.cherry.Cherry'>

Documentation

The full interface documentation can be find in the docs/ folder. This is just an excerpt of the main function.

ultraimport function 🔗 source

ultraimport(
    file_path,
    objects_to_import=None,
    add_to_ns=None,
    preprocessor=None,
    package=None,
    caller=None,
    use_cache=True,
    lazy=False,
    recurse=False,
    inject=None,
    use_preprocessor_cache=True,
    cache_path_prefix=None
)

Import Python code files from the file system. This is the central main function of ultraimport.

Parameters:

• file_path (str): Path to the module file that should be imported. It can have any file extension. Please be aware that you must provide the file extension. The path can be relative or absolute. You can use the special string __dir__ to refer to the directory of the caller. If run from a Python REPL, the current working directory will be used for __dir__. If you use advanced debugging tools (or want to save some CPU cycles) you might want to set caller=__file__.

• objects_to_import (str | (Iterable[str] | Dict[str, object]): Can have several modes depending on the type of the parameter.

  • (str): Name of a single object to import from the module in file_path. The special value '*' selects all objects from that module.
  • (Iterable[str]): A list of names of objects to import.
  • (Dict[str, object]): The keys represent the names of the objects to import. The values define the expected types of those objects. A TypeError is thrown if the types don't match the expectation. If you set lazy=True, you must use a dict for objects_to_import and define the types.

• add_to_ns (Dict[str, object]): add the objects_to_import to the dict provided. A common value to use is globals(). If you set add_to_ns=True, the imported objects are added to the local scope of the caller.

• preprocessor (callable): Takes the source code as an argument and can return a modified version of the source code. Check out the debug-transform example on how to use the preprocessor.

• package (str | int): Can have several modes depending on if you provide a string or an integer.

  • (str) Generate one or more namespace packages and use them as parent package the imported module.
  • (int) Number of path parts (directories) to extract from the file_path to calculate the namespace package. This can help with subsequent relative imports in your imported files. If package is set to the default None, the module will be imported without setting it parent __package__.

• use_cache (bool): If set to False, allows re-importing of the same source file even if it was imported before. Otherwise a cached version of the imported module is returned.

• lazy (bool): Experimental wip If set to True and if objects_to_import is set to None, it will lazy import the module. If set to True and objects_to_import is a dict, the values of the dict must be the type of the object to lazy import from the module. Currently only the type callable is supported.

• recurse (bool): If set to True, a built-in preprocessor is activated to transparently rewrite all relative import statements (those with a dot like from . import something) to ultraimport() calls. Use this mode if you have no control over the source code of the impored modules.

• cache_path_prefix (str): Directory for storing preprocessed files. If you use the preprocessor feature or if you use the option recurse=True (which in turn uses the preprocessor feature) you will have the option to store the resulting code after preprocessing. By default, they are stored in parallel to the original source code files, but this option allows to override to location. One common setting is cache_path_prefix='__pycache__' to store the processed files along with the bytecode files. _Note_: Even when you change this directory, this will be hidden from Python. Towards Python, the preprocessed files will always look like they are in the same directory as the original source code files, even if they are not.

Returns: Depending on the parameters returns one of the following:

• object: If objects_to_import is None, returns a single module object.

• object: If objects_to_import is a str, returns the single object with the specified name from the imported module.

• dict: If objects_to_import has the value '*', returns a dict of all items from the imported module.

• list: If objects_to_import is a List[str], return a list of imported objects from the imported module.

Advanced Usage

See docs/advanced-usage.md

Better Error Messages

See docs/better-error-messages.md

Contributing

We love contributions!

ultraimport is open source, built on open source, and we'd love to have you hang out in our community.

The Issue: Relative Imports in Python

Classically, to do a relative import, a Python script run.py would look like this if you wanted to import the logger object from a logging.py module in the same directory:

from .logging import logger

def main():
    # do something

    logger('I did something')

if __name__ == 'main':
    main()

If you try to run the program in usual way, you'll get an error message:

$ python ./run.py
Traceback (most recent call last):
  File "/home/user/myprogram/./run.py", line 1, in <module>
    from .logging import logger
ImportError: attempted relative import with no known parent package

Python programs or scripts can be executed in a number of different ways and surprisingly, with some of the ways, it even works:

# Broken
python ~/myprogram/run.py

# Works
cd ~
python -c 'import myprogram.run'

# Works
python -m myprogram.run

# Broken
cd ~/myprogram
python -c 'import run'

# Broken
python -m run

# Broken
python ./run.py

# Broken
~/myprogram/run.py

You wonder: Why does Python come to a different conclusion depending on the way how I run the program?

The error ImportError: attempted relative import with no known parent package is rather erratic because the code has never changed. Sometimes you would also get ValueError: attempted relative import beyond top-level package.

There actually is a known parent package. It's the directory where the code lives in. Sometimes Python can see it, sometimes not.

Even if there was no parent package, what's the issue with importing a module that I only know from its relative position to my current module?

With ultraimport your program run.py will always find logging.py in the same directory, no matter how you run it. You could change it to look like this:

#!/usr/bin/env python3

# ultraimport needs to be installed and imported in the classical way
import ultraimport

# Import the 'logger' object from 'logging.py' that is located in the same
# directory as this file.
# Note: `__dir__` refers to the directory where this file is in.
logger = ultraimport('__dir__/logging.py', 'logger')

def main():
    # do something
    logger('I did something')

if __name__ == '__main__':
    main()
else:
    logger('I was imported')

As you can see, you'll have to import ultraimport in the classical way. It's intended to be installed as a system-wide library. Afterwards, you can import your own code based on relative or absolute file system paths so it can always be found.

With ultraimport, Python code can be executed in an way and the imports keep working:

# Works
python ~/myprogram/run.py

# Works
cd ~
python -c 'import myprogram.run'

# Works
python -m myprogram.run

# Works
cd ~/myprogram
python -c 'import run'

# Works
python -m run

# Works
python ./run.py

# Works
~/myprogram/run.py

Python Relative Import Limitations

https://peps.python.org/pep-0328/#relative-imports-and-name

Relative imports use a module's name attribute to determine that module's position in the package hierarchy. If the module's name does not contain any package information (e.g. it is set to 'main') then relative imports are resolved as if the module were a top level module, regardless of where the module is actually located on the file system.