src/datamodel_code_generator/_builtin_formatter.py (1)

1718-1725: 🎯 Functional Correctness | 🔵 Trivial | 💤 Low value

Guard logic is sound but only accounts for \n-terminated backslash continuations.

Per Python's lexical analysis, multi-line STRING tokens can only arise from triple-quoted literals or backslash line continuations, so the general approach here is correct. However, the check "\\\n" in code will miss a backslash continuation that uses CRLF line endings (i.e., "\\\r\n"), since the substring \ immediately followed by \n wouldn't be present when a \r sits between them. If code could ever contain CRLF line endings alongside a backslash-continued single-quoted string, the guard would incorrectly skip tokenization and string_lines would stay empty, causing blank-line normalization logic to misclassify lines inside that string as top-level code.

This is likely a non-issue in practice since generated code from this formatter is presumably always \n-terminated, but worth confirming.

🛡️ Optional defensive fix

-    if ('"""' in code) or ("'''" in code) or ("\\\n" in code):
+    if ('"""' in code) or ("'''" in code) or ("\\\n" in code) or ("\\\r\n" in code):

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/datamodel_code_generator/_builtin_formatter.py` around lines 1718 - 1725,
The guard in _normalize_top_level_blank_lines only detects backslash
continuations with LF endings, so update the multi-line STRING precheck to also
recognize CRLF backslash continuations before deciding whether to tokenize. Keep
the existing tokenization-based approach in _normalize_top_level_blank_lines,
but broaden the continuation detection so string_lines is populated correctly
when code contains "\r\n" line endings.

tests/main/test_performance.py (1)

128-158: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Duplicate test bodies — consider parametrizing.

The two tests are identical except for the formatters argument and naming. Consolidate with pytest.mark.parametrize.

♻️ Suggested consolidation

-@pytest.mark.perf
-@pytest.mark.benchmark
-def test_perf_large_models_pydantic_v2_builtin(tmp_path: Path) -> None:
-    """Performance test: Generate 500 Pydantic v2 models with the built-in formatter."""
-    output_file = tmp_path / "output.py"
-    generate(
-        input_=PERFORMANCE_DATA_PATH / "large_models.json",
-        input_file_type=InputFileType.JsonSchema,
-        output=output_file,
-        output_model_type=DataModelType.PydanticV2BaseModel,
-        formatters=[Formatter.BUILTIN],
-    )
-    content = output_file.read_text()
-    assert content.count("class Model") >= 500
-
-
-@pytest.mark.perf
-@pytest.mark.benchmark
-def test_perf_large_models_pydantic_v2_noformat(tmp_path: Path) -> None:
-    """Performance test: Generate 500 Pydantic v2 models without formatting."""
-    output_file = tmp_path / "output.py"
-    generate(
-        input_=PERFORMANCE_DATA_PATH / "large_models.json",
-        input_file_type=InputFileType.JsonSchema,
-        output=output_file,
-        output_model_type=DataModelType.PydanticV2BaseModel,
-        formatters=[],
-    )
-    content = output_file.read_text()
-    assert content.count("class Model") >= 500
+@pytest.mark.perf
+@pytest.mark.benchmark
+@pytest.mark.parametrize(
+    ("formatters", "test_id"),
+    [([Formatter.BUILTIN], "builtin"), ([], "noformat")],
+    ids=["builtin", "noformat"],
+)
+def test_perf_large_models_pydantic_v2_formatting(tmp_path: Path, formatters: list, test_id: str) -> None:
+    """Performance test: Generate 500 Pydantic v2 models with varying formatting."""
+    output_file = tmp_path / "output.py"
+    generate(
+        input_=PERFORMANCE_DATA_PATH / "large_models.json",
+        input_file_type=InputFileType.JsonSchema,
+        output=output_file,
+        output_model_type=DataModelType.PydanticV2BaseModel,
+        formatters=formatters,
+    )
+    content = output_file.read_text()
+    assert content.count("class Model") >= 500

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/main/test_performance.py` around lines 128 - 158, These two performance
tests duplicate the same generate/assert flow and only differ in the formatters
argument, so consolidate them into one parametrized pytest test. Refactor the
duplicated logic in test_perf_large_models_pydantic_v2_builtin and
test_perf_large_models_pydantic_v2_noformat using pytest.mark.parametrize over
the formatter set while keeping the existing generate call and class count
assertion intact.