How nshaibu/pydantic-mini works
Project Type and Purpose
• It’s a Python library/tool that provides a minimal, stdlib-only alternative to Pydantic’s BaseModel.
• Purpose: Define dataclass-style models with built-in type/value validation, custom validators, and serialization to/from JSON, dict, and CSV.
File Structure
• Top-level: packaging & CI files (.github/, pyproject.toml, requirements-*.txt)
• src/pydantic_mini/: the runtime modules:
• base.py – core BaseModel implementation
• typing.py – MiniAnnotated, Attrib definitions
• formatters.py – load/dump logic for JSON, dict, CSV
• exceptions.py – ValidationError, etc.
• utils.py – helper functions (e.g., InitVar support)
• tests/: unit tests that validate functionality.
README for Insights
• Core flows:
a. User defines a subclass of BaseModel with annotated fields.
b. .loads() invokes formatters, then BaseModel initializer.
c. BaseModel introspects annotations, applies type checks, runs validators.
d. On success, returns model instance; on failure, raises ValidationError.
e. .dump() serializes instances via formatters.
• Extensibility via: custom field validators, class-level validate methods, default factories, InitVar hooks.
Main Components & Their Relationships
a. User Code (their BaseModel subclasses)
b. BaseModel Core
c. Field Metadata / Annotation Layer (MiniAnnotated & Attrib)
d. Validation Engine
e. Serialization/Deserialization Layer (formatters)
f. Exceptions Module
g. Utility Module
h. Python stdlib (dataclasses, typing, json, csv, datetime)
• User code → BaseModel (definition + instantiation)
• BaseModel → typing layer (to parse field metadata)
• BaseModel → utils (InitVar, default factories)
• BaseModel → validation engine (calls built-in & custom validators)
• Validation engine → exceptions (throws ValidationError)
• BaseModel → formatters (loads/dumps) → stdlib modules
• Tests/CI pipelines are external: they invoke package build & run unit tests.
Architectural Patterns & Principles
• Modular separation: core logic (base) vs metadata (typing) vs IO (formatters) vs errors (exceptions).
• Single responsibility per module.
• Dependency‐inversion: BaseModel depends on abstract Annotated metadata, not concrete validators.
• Extensibility: hook points (validate_<field> methods, validate hook, __model_init__).
• Stdlib‐only: no third-party dependencies.