pydantic-modelable-mypy 0.4.0

A mypy plugin teaching the type-checker about pydantic-modelable runtime model extensions

pydantic-modelable-mypy

A mypy plugin that teaches the type-checker about the model extensions pydantic_modelable performs at runtime.

Why

pydantic_modelable lets a class decorate a different, pre-existing model to add fields, union members or enum values to it. A decorator cannot re-type a third class in the standard type system, so mypy --strict reports the added fields as unknown attributes. This plugin observes those decorators and injects the corresponding members into the target model during analysis.

Usage

Install alongside pydantic-modelable and enable the plugin in your mypy configuration:

[mypy]
plugins = pydantic_modelable_mypy.plugin

What it covers

• as_attribute — the injected field is known for reads and construction.
• extends_union — the field is typed as the discriminated union of the base's subtypes (reads and construction), including subtypes defined in other modules, on full and incremental runs.
• extends_enum — the discriminator values injected as enum members are resolved by name (Palette.red).
• ModelableForwarder — decorators applied through a forwarder (@Fwd.as_attribute(...)) are resolved through the forwards_to chain (including chained forwarders) to the target Modelable and handled as above.

Limitation (extends_enum member names)

Enum member access on an unknown name offers no plugin hook, so injected member names can only be resolved from subtypes discovered during analysis. This is reliable on full (non-incremental) runs (e.g. CI with mypy fresh, or --no-incremental). On incremental / daemon runs, subtypes served from cache are not re-analysed, so member-name access (Palette.red) may report an unknown attribute until a clean run. Everything else ModelableStrEnum provides — iteration, membership, construction, .value — works in all modes, as does all of as_attribute and extends_union.