README.md

Kagglize Module

Use your Python modules (code split up over mutiple files) for Kaggle submissions.

Why

When writing complex Python programs, it is practial to split up the code over several files and directories. The directory structure might look something like this:

my_module
├── __init__.py
├── some_file.py
├── some_directory
│   ├── __init__.py
│   └── another_file.py
...

Unfortunately, the online competition platform Kaggle uses Jupyter notebooks and single Python scripts edited in a web-based IDE. It does not support uploading a locally developed python module consisting of mutiple files. (Technically uploading single files works but is prohibitively time-intensive and error-prone to do for a whole module, especially when resubmitting on a regular basis.)

This tools helps you with that, by merging all your files into a single Python script. When executed this script regenerates all your original files, so you can import and run code from them.

Usage

Install via pip:

$ pip install kagglize-module

Run:

$ kagglize-module
Usage: kagglize-module [OPTIONS] MODULE_ROOT_PATH

  Generate a single python script to recreate files of a python module.

Options:
  --output-file PATH      Location of generated script.  [default: output.py]
  --filename-regex TEXT   Filter filenames with regex.  [default: ^.*\.py$]
  --import / --no-import  Generate import statement.  [default: True]
  --help                  Show this message and exit.

Example:

$ kagglize-module my_module  # the module root path

Generates following python code to output.py:

from pathlib import Path

Path('my_module').mkdir(parents=True, exist_ok=True)
Path('my_module/__init__.py').open('w').write('')
Path('my_module/some_file.py').open('w').write('<content of some_file.py as string>')
Path('my_module/some_directory').mkdir(parents=True, exist_ok=True)
Path('my_module/some_directory/__init__.py').open('w').write('')
Path('my_module/some_directory/another_file.py').open('w').write('<content of another_file.py as string>')
...

import my_module

You can copy and paste this code into a Jupyter notebook on Kaggle and run it to load your module.

Disclaimer

This is intended only to upload your own code and not for uploading 3rd party modules.

Use at your own risk (see LICENSE). I cannot guarantee that Kaggle allows using this and will not disqualify your submission. When asking regarding its usage for the Abstraction and Reasoning Challenge I got a positive answer from a Kaggle Team Member.