IO for dummies! We make IO as easy as possible by providing a unified save/load interface, using the most common and recommendable default options for IO between various object types and file types. (Users may pass additional keyword arguments to the underlying IO methods.)
dummio simplifies IO calls for some file types. For example, instead of
with open(file_path, 'r', encoding='utf-8') as file:
data = json.load(file)
you can simply
data = dummio.json.load(file_path)
In some coding applications it is desirable to pass an IO module as an argument to a function. Here it is convenient to pass a dummio submodule, since all dummio submodules have the same save and load interface, having equivalent signatures (except for differences hidden in **kwargs).
So far we support:
- text, pickle, and dill
- simple dictionaries:
- json
- yaml
- pandas dataframes:
- csv
- parquet
- onnx.ModelProto instances
- pydantic models (relying on the built-in json serialization methods)
Filepaths passed to save and load methods can be of type str, pathlib.Path, or universal_pathlib.UPath such that many of these methods will "just work" against cloud paths like UPath("s3://bucket/key"), UPath("gs://bucket/key"), or UPath("az://container/key").
dummio has no required dependencies. For example, calling from dummio.pandas import df_parquet will raise a helpful message to install pandas if you have not already done so.
Basic IO methods can be accessed directly as dummio.text, dummio.json, etc:.
import dummio
text = "hello world"
data = {"key": text}
path = "io_example_file"
# Text
dummio.text.save(text, path=path)
assert text == dummio.text.load(path)
# YAML
dummio.yaml.save(data)
assert data == dummio.yaml.load(path)
We're on pypi, so pip install dummio.
git clone [email protected]:zkurtz/dummio.git
cd dummio
pip install uv
uv sync --group extras
source .venv/bin/activate
pre-commit install