Skip to main content

Generate command line options from dataclasses.

Project description

Table of Contents

  1. Install
  2. Usage
    1. Dataclass to command line options
      1. Simple types
      2. Complex types
      3. Private fields
      4. Nested dataclass
    2. APIs
      1. Example

Generate command line options from dataclasses.

# config.py
from dataclasses import dataclass, asdict, field
from smile_config import from_dataclass

@dataclass
class Train:
    """Train config."""

    batch_size: int = 64


@dataclass
class ML:
    lr: Annotated[float, dict(help="learning rate", type=float)] = 0.001
    train: Train = Train()
    cc: list[int] = field(default_factory=lambda: [10])


@dataclass
class Example:
    """Example config."""

    ml: ML = ML()
    x: bool = True
    a: int | None = None

config = from_dataclass(Example()).config

print(config)

# If autocomplete is not working, try to add the following line to your config file:
from typing import cast
config = cast(Example, config)

You can access the config as namedtuple.

> python config.py --ml.cc 10 10 --ml.lr 0.001 --no-x --a "1"
Example(ml=ML(lr=0.001, train=Train(batch_size=64), cc=[10, 10]), x=False, a=1)

Also, auto generate help message with default value.

> python config.py --help
Usage: config.py [-h] [--ml.lr float] [--ml.train.batch_size int] [--ml.cc int [int ...]] [--x | --no-x] [--a int]

Example config.

Options:
  -h, --help            show this help message and exit
  --x, --no-x           - (default: True)
  --a int               - (default: None)

Ml:
  --ml.lr float         learning rate (default: 0.001)
  --ml.cc int [int ...]
                        - (default: [10])

Ml.Train:
  --ml.train.batch_size int
                        - (default: 64)

Install

pip install -U smile_config

Usage

Dataclass to command line options

Simple types

Everything that argpase can handle. int, float, str, bool, and callable object.

@dataclass
class Simple:
    a: int = 1
    b: float = 2.0
    c: str = "hello"
    d: bool = False
    e: list[int] = field(default_factory=lambda: [10])

Will convert to:

parser.add_argument("--a", help="-", type=int, default=1)
parser.add_argument("--b", help="-", type=float, default=2.0)
parser.add_argument("--c", help="-", type=str, default="hello")
parser.add_argument("--d", help="-", type=bool, default=False, action="store_true")
parser.add_argument("--e", help="-", type=int, default=[10], nargs="+")

Complex types

Smile config uses Annotation to handle complex types, which will pass the second argument to parser.add_argument.

@dataclass
class C:
    x: Annotated[int, "Helps for x."] = 1

See the logic here:

The first argument is the type, e.g. int.

if the second argument is str, e.g. s, it will be passed as parser.add_argument("--x", help=s, ...).

If the second argument is a list, e.g. args, it will be passed as parser.add_argument("--x", ..., *args).

If the second argument is a dict, e.g. kwds, it will be passed as parser.add_argument("--x", ..., **kwds).

Private fields

Fields that start with _ will be ignored. Thus, please initialize it by default or in `__post_init__`.

Nested dataclass

Of course! It does support nested dataclass.

@dataclass
class A:
    a: int = 1

@dataclass
class B:
    a: A = A()

@dataclass
class C:
    a: A = A()
    b: B = B()
    c: int = 0
    _d: str = "private _d"


print(from_dataclass(C()).config)

# Output:
# C(a=A(a=1), b=B(a=A(a=1)), c=0, _d="private _d")

APIs

Smile config provides four APIs:

class Config:

    # the dataclass dict
    self.conf

    # the dataclass
    self.config

# Generate command line options from dataclass.
# For formatter: `from rich_argparse import RichHelpFormatter`
# `ns`: namespaces for types.
def from_dataclass(dc: Dataclass, *, formatter: HelpFormatter = RichHelpFormatter, ns: dict | None = None) -> Config:...

# Convert dict to an existing dataclass
def from_dict(dc: Type[Dataclass], d: dict) -> Dataclass:...

# Merge a dict with an existing dataclass instance
def merge_dict(dc: Dataclass, d: dict) -> Dataclass:...

Example

@dataclass
class Eg:
    a: int = 1
    b: bool = False

conf = from_dataclass(Eg())

print(conf)  # Config
# output: Eg(a=1, b=False)

print(conf.conf)  # dict
# output: {'a': 1, 'b': False}

print(conf.config)  # Eg
# output: Eg(a=1, b=False)

conf_dc = from_dict(Eg, {"a": 2, "b": True})  # Type[Eg] -> dict -> Eg
print(conf_dc)
# output: Eg(a=2, b=True)

conf_merge = merge_dict(conf_dc, {"a": 3})  # Eg -> dict -> Eg
print(conf_merge)
# output: Eg(a=3, b=True)

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

smile-config-0.12.0.tar.gz (9.9 kB view details)

Uploaded Source

Built Distribution

smile_config-0.12.0-py3-none-any.whl (11.3 kB view details)

Uploaded Python 3

File details

Details for the file smile-config-0.12.0.tar.gz.

File metadata

  • Download URL: smile-config-0.12.0.tar.gz
  • Upload date:
  • Size: 9.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/4.0.2 CPython/3.10.9

File hashes

Hashes for smile-config-0.12.0.tar.gz
Algorithm Hash digest
SHA256 7f19523a0bccf1fea370ce5ae3ae97dc5631410892212b6cd318fe664cbd465e
MD5 15377e7bcdc2d88a49307120ac84d67c
BLAKE2b-256 c70b5c79c7ce6fffd37c61e0263797d75f3e1b9a37ce08b9a9cf64b8d7c6e36d

See more details on using hashes here.

File details

Details for the file smile_config-0.12.0-py3-none-any.whl.

File metadata

File hashes

Hashes for smile_config-0.12.0-py3-none-any.whl
Algorithm Hash digest
SHA256 9d2886f12e1a39aba41fa0b4f8fa5e68a00093914a3cf489116f569f864e5db6
MD5 77aaadeb3eef08acdfbb7c647399665a
BLAKE2b-256 ce3394134171ed6a61886c9e060f35919e52b405ee8d7e9d95bb5f77f270c676

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page