Confit
Confit is a complete and easy-to-use configuration framework aimed at improving the reproducibility of experiments by relying on the Python typing system, minimal configuration files and command line interfaces.
Getting started
Install the library with pip:
pip install confit
Confit only abstracts the boilerplate code related to configuration and leaves the rest of your code unchanged.
Here is an example:
script.py
+ from confit import Cli, Registry, RegistryCollection
+ class registry(RegistryCollection):
+ factory = Registry(("test_cli", "factory"), entry_points=True)
+ @registry.factory.register("submodel")
class SubModel:
# Type hinting is optional but recommended !
def __init__(self, value: float, desc: str = ""):
self.value = value
self.desc = desc
+ @registry.factory.register("bigmodel")
class BigModel:
def __init__(self, date: datetime.date, submodel: SubModel):
self.date = date
self.submodel = submodel
+ app = Cli(pretty_exceptions_show_locals=False)
# you can use @confit.validate_arguments instead if you don't plan on using the CLI
+ @app.command(name="script", registry=registry)
def func(modelA: BigModel, modelB: BigModel, seed: int = 42):
"""
Display the configured model dates.
Parameters
----------
modelA : BigModel
The first model whose date is displayed.
modelB : BigModel
The second model whose date is displayed.
seed : int
Random seed.
"""
assert modelA.submodel is modelB.submodel
print("modelA.date:", modelA.date.strftime("%B %-d, %Y"))
print("modelB.date:", modelB.date.strftime("%B %-d, %Y"))
+ if __name__ == "__main__":
+ app()
Create a new config file
The following also works with YAML files
config.cfg
# CLI sections
[script]
modelA = ${modelA}
modelB = ${modelB}
# CLI common parameters
[modelA]
@factory = "bigmodel"
date = "2003-02-01"
[modelA.submodel]
@factory = "submodel"
value = 12
[modelB]
date = "2003-04-05"
submodel = ${modelA.submodel}
and run the following command from the terminal
python script.py --config config.cfg --seed 43
The generated CLI also has a readable help message based on the function signature and docstrings:
python script.py --help
Display the configured model dates.
Parameters
----------
--config <Path>
Load a config file to fill in the following params. Can be repeated.
--modelA <BigModel>
The first model whose date is displayed.
--modelA.<field> VALUE
--modelB <BigModel>
The second model whose date is displayed.
--modelB.<field> VALUE
--seed <int> (default: 42)
Random seed.
You can still call the function method from your code, but now also benefit from
argument validation !
from script import func, BigModel, SubModel
# To seed before creating the models
from confit.utils.random import set_seed
seed = 42
set_seed(seed)
submodel = SubModel(value=12)
func(
# BigModel will cast date strings as datetime.date objects
modelA=BigModel(date="2003-02-01", submodel=submodel),
# Since the modelB argument was typed, the dict is cast as a BigModel instance
modelB=dict(date="2003-04-05", submodel=submodel),
seed=seed,
)
modelA.date: February 1, 2003
modelB.date: April 5, 2003
Serialization
You can also serialize registered classes, while keeping references between instances:
from confit import Config
submodel = SubModel(value=12)
modelA = BigModel(date="2003-02-01", submodel=submodel)
modelB = BigModel(date="2003-02-01", submodel=submodel)
print(Config({"modelA": modelA, "modelB": modelB}).to_str())
[modelA]
@factory = "bigmodel"
date = "2003-02-01"
[modelA.submodel]
@factory = "submodel"
value = 12
[modelB]
@factory = "bigmodel"
date = "2003-02-01"
submodel = ${modelA.submodel}
Error handling
You also benefit from informative validation errors:
func(
modelA=dict(date="hello", submodel=dict(value=3)),
modelB=dict(date="2010-10-05", submodel=dict(value="hi")),
)
ConfitValidationError: 2 validation errors for __main__.func()
-> modelA.date
invalid date format, got 'hello' (str)
-> modelB.submodel.value
value is not a valid float, got 'hi' (str)
Visit the documentation for more information!
Subcommands
Confit applications can be composed into groups of subcommands:
app = Cli()
training = Cli()
@training.command(name="run")
def run(epochs: int = 10):
print(f"Training for {epochs} epochs")
app.add_subcommands(training, name="training")
Run the nested command with:
python script.py training run --config config.yml --epochs 20
Acknowledgement
We would like to thank Assistance Publique – Hôpitaux de Paris and AP-HP Foundation for funding this project.
Metadata
Release files for confit 0.13.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| confit-0.13.0.tar.gz | 49.5 kB | Without provenance | Details for confit-0.13.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| confit-0.13.0-py3-none-any.whl | Python 3 | none | any | Details for confit-0.13.0-py3-none-any.whl |
Total release size: 85.4 kB
Release files / confit-0.13.0.tar.gz
| Download URL | confit-0.13.0.tar.gz |
|---|---|
| Size | 49.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
e0b67d2aca98a6f14838c29a3f989f3698d7042de1ee47221134fbd58f4024da
|
|
BLAKE2b-256 checksum How to use checksums |
f054a276a5f377b5fc1acd96bf7bf836a1c9d569ac96785968b0077837efc597
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.9.25
|
Release files / confit-0.13.0-py3-none-any.whl
| Download URL | confit-0.13.0-py3-none-any.whl |
|---|---|
| Size | 35.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
982d665eae48bc54c16afa41330ca7058598f9a7cf3684e2949745092545331c
|
|
BLAKE2b-256 checksum How to use checksums |
002dc02e653010906039f5d811a738cba76d3e0531d6c09eb20a325353f70a29
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.9.25
|
Security for confit 0.13.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.