Skip to content

Validate and extract data into type-safe dataclasses

License

Notifications You must be signed in to change notification settings

takkaria/plucker

Repository files navigation

plucker

Build Status codecov Python Version

Validate and extract JSON-sourced data into type-safe dataclasses

wut

  • Tired of relying on vendor-provided, untyped Python libraries to interface with external APIs?
  • Want to just make a few simple HTTP requests without the weight of extra dependencies?
  • Do you only use a small subset of the data you get from external sources, picking out five fields when you are given thirty?
  • Are you more than slightly worried that the API might change under you and you wouldn't know?
  • Want to avoid just reaching into dictionaries to get the data you want?
  • Do you want to parse the JSON you get into something type-safe so that mypy will complain when you do wrong things?
  • Is writing fakes a bit too heavyweight for the APIs you're calling? Would producing an error on unexpected input work OK for now?

Enter plucker.

plucker is designed to validate, map and reduce regularly structured data into dataclasses. That data would typically be JSON from APIs but could be anything that mostly consists of Python dicts and lists when parsed.

plucker will either give you type-verified data, or it will fail with helpful error messages:

Data not in expected format; expected fred to be 'list' but it was 'dict':
.fred[].v
 ^^^^

Just pick the data you want using jq-style paths, map it so that it's the right type if you need to, and you have well-typed validated data to feed into the rest of your system.

Installation (soon...)

pip install plucker

Example

from typing import List
from dataclasses import dataclass
from enum import Enum, auto
from datetime import date

from plucker import pluck, Path


class Status(Enum):
    """A cintact's status."""
    CURRENT = auto()
    EXPIRED = auto()


@dataclass
class Contact:
    """A contact record."""
    name: str
    email: str


@dataclass
class Struct:
    """The typed dataclass we want our data collected into."""
    date: date
    id: int
    state: Status
    affected_records: List[int]
    contacts: List[Contact]


TO_STATUS = {"CUR": Status.CURRENT, "EXP": Status.EXPIRED}

input_ = {
    "date": "2021-01-01",
    "id": "1242",
    "payload": {
        "from": "CURRENT",
        "who": [
            {"name": "DM", "id": 1, "email": "dangermouse@example.com"},
            {"name": "Stiletto", "id": 23, "email": "baroni@example.com"},
        ]
    }
}

plucked = pluck(
    input_,
    Struct,
    date=Path(".date"),
    id=Path(".id").map(int),
    state=Path(".payload.from").map(TO_STATUS),
    affected_records=Path(".payload.who[].id"),
    people=Path(".payload.who[]").into(
        Contact,
        name=Path(".name"),
        email=Path(".email"),
    ),
)

expected = Struct(
    date=date(2021, 1, 1),
    id=1242,
    state=Status.CURRENT,
    affected_records=[1, 23],
    contacts=[
        Contact("DM", "dangermouse@example.com"),
        Contact("Stiletto", "baroni@example.com")
    ]
)

assert plucked == expected
# ^ it's True

Prior art

  1. dataclasses_json -> require the same structure between JSON and serialization, which means you have to specify an intermediate structure
  2. DRF serializers -> heavyweight and not type safe, destructure into dictionaries
  3. Elm's JSON decoders -> this design isn't really based on anything in there but ever since using them I wanted similar functionality in Python
  4. jq, an amazing commandline tool for querying JSON data
  5. Parse, don't validate
  6. Elm's error messages

License

MIT

Credits

A bunch of the tooling was taken from wemake-python-package but then heavily modified.