test: Run the model generator in unit tests

tests/unit/test_model_generation.py

@@ -0,0 +1,94 @@
+from __future__ import annotations
+
+import json
+import subprocess
+import sys
+from typing import TYPE_CHECKING
+
+from scripts.postprocess_generated_models import (
+    REPO_ROOT,
+    RESOURCE_INPUT_TYPEDDICTS,
+    build_alias_map,
+    postprocess_models,
+    postprocess_typeddicts,
+)
+
+if TYPE_CHECKING:
+    from pathlib import Path
+
+MINIMAL_SPEC = {
+    'openapi': '3.1.0',
+    'info': {'title': 'Test', 'version': '1.0.0'},
+    'paths': {},
+    'components': {
+        'schemas': {
+            'RunStatus': {'type': 'string', 'enum': ['READY', 'RUNNING', 'SUCCEEDED']},
+            'Run': {
+                'type': 'object',
+                'required': ['id', 'startedAt', 'status'],
+                'properties': {
+                    'id': {'type': 'string'},
+                    'startedAt': {'type': 'string', 'format': 'date-time'},
+                    'status': {'$ref': '#/components/schemas/RunStatus'},
+                },
+            },
+            # Post-processing fails unless every TypedDict seed is present in the generated file. Each gets its own
+            # field, so `reuse_model` doesn't turn them into empty subclasses of one shared class.
+            **{
+                name: {
+                    'type': 'object',
+                    'required': [f'{name[0].lower()}{name[1:]}Id'],
+                    'properties': {f'{name[0].lower()}{name[1:]}Id': {'type': 'string'}},
+                }
+                for name in RESOURCE_INPUT_TYPEDDICTS
+            },
+        }
+    },
+}
+
+
+def run_datamodel_codegen(*args: str) -> None:
+    # Running from the repository root makes datamodel-codegen pick up `[tool.datamodel-codegen]` from `pyproject.toml`.
+    subprocess.run(  # noqa: S603
+        [sys.executable, '-m', 'datamodel_code_generator', *args],
+        check=True,
+        cwd=REPO_ROOT,
+    )
+
+
+def test_model_generation_pipeline(tmp_path: Path) -> None:
+    """The `generate-models` pipeline (both codegen passes and post-processing) runs on a minimal spec."""
+    # PR CI never runs `poe generate-models`, so this catches a dependency update that breaks the generator.
+    spec_path = tmp_path / 'openapi.json'
+    spec_path.write_text(json.dumps(MINIMAL_SPEC))
+    models_path = tmp_path / '_models.py'
+    literals_path = tmp_path / '_literals.py'
+    typeddicts_path = tmp_path / '_typeddicts.py'
+
+    run_datamodel_codegen('--input', str(spec_path), '--output', str(models_path), '--alias-generator', 'to_camel')
+    run_datamodel_codegen(
+        '--input',
+        str(spec_path),
+        '--output',
+        str(typeddicts_path),
+        '--output-model-type',
+        'typing.TypedDict',
+        '--no-use-closed-typed-dict',
+    )
+    postprocess_models(models_path, literals_path)
+    postprocess_typeddicts(typeddicts_path, build_alias_map(models_path.read_text()))
+
+    models = models_path.read_text()
+    assert "@docs_group('Models')\nclass Run(BaseModel):" in models
+    assert 'alias_generator=to_camel' in models
+    assert 'started_at: AwareDatetime' in models
+    assert 'from apify_client._literals import RunStatus' in models
+    assert 'class RunStatus' not in models
+
+    literals = literals_path.read_text()
+    assert "RunStatus = Literal[\n    'READY',\n    'RUNNING',\n    'SUCCEEDED',\n] | str" in literals
+
+    typeddicts = typeddicts_path.read_text()
+    assert 'class RequestDict(TypedDict):\n    request_id: str' in typeddicts
+    assert 'class RequestCamelDict(TypedDict):\n    requestId: str' in typeddicts
+    assert 'class Run' not in typeddicts