diff --git a/benchmark/generate_samples.py b/benchmark/generate_samples.py
index fdc9463..780741d 100644
--- a/benchmark/generate_samples.py
+++ b/benchmark/generate_samples.py
@@ -576,6 +576,7 @@ def main() -> None:
     )
     defer_imports_path.write_text(defer_imports_contents, encoding="utf-8")
 
+    # Same defer_imports-influenced imports, but for a test in the tests directory
     tests_path = Path().resolve() / "tests" / "stdlib_imports.py"
     tests_path.write_text(defer_imports_contents, encoding="utf-8")
 
diff --git a/src/defer_imports/_core.py b/src/defer_imports/_core.py
index 82355ba..324d57a 100644
--- a/src/defer_imports/_core.py
+++ b/src/defer_imports/_core.py
@@ -37,9 +37,9 @@ class DeferredInstrumenter(ast.NodeTransformer):
     results are assigned to custom keys in the global namespace.
     """
 
-    def __init__(self, filepath: _tp.Union[_tp.StrPath, _tp.ReadableBuffer], data: SourceData, encoding: str) -> None:
-        self.filepath = filepath
+    def __init__(self, data: SourceData, filepath: _tp.Union[_tp.StrPath, _tp.ReadableBuffer], encoding: str) -> None:
         self.data = data
+        self.filepath = filepath
         self.encoding = encoding
         self.scope_depth = 0
 
@@ -269,12 +269,6 @@ def visit_Module(self, node: ast.Module) -> ast.AST:
         return self.generic_visit(node)
 
 
-def match_token(token: tokenize.TokenInfo, **kwargs: object) -> bool:
-    """Check if a given token's attributes match the given kwargs."""
-
-    return all(getattr(token, name) == val for name, val in kwargs.items())
-
-
 def sliding_window(iterable: _tp.Iterable[_tp.T], n: int) -> _tp.Iterable[tuple[_tp.T, ...]]:
     """Collect data into overlapping fixed-length chunks or blocks.
 
@@ -306,10 +300,10 @@ def check_source_for_defer_usage(data: _tp.Union[_tp.ReadableBuffer, str]) -> tu
         encoding = next(token_stream).string
 
     uses_defer = any(
-        match_token(tok1, type=tok_NAME, string="with")
-        and match_token(tok2, type=tok_NAME, string="defer_imports")
-        and match_token(tok3, type=tok_OP, string=".")
-        and match_token(tok4, type=tok_NAME, string="until_use")
+        (tok1.type == tok_NAME and tok1.string == "with")
+        and (tok2.type == tok_NAME and tok2.string == "defer_imports")
+        and (tok3.type == tok_OP and tok3.string == ".")
+        and (tok4.type == tok_NAME and tok4.string == "until_use")
         for tok1, tok2, tok3, tok4 in sliding_window(token_stream, 4)
     )
 
@@ -351,7 +345,7 @@ def source_to_code(  # pyright: ignore [reportIncompatibleMethodOverride]
         if not uses_defer:
             return super().source_to_code(data, path, _optimize=_optimize)  # pyright: ignore # See note above.
 
-        tree = DeferredInstrumenter(path, data, encoding).instrument()
+        tree = DeferredInstrumenter(data, path, encoding).instrument()
         return super().source_to_code(tree, path, _optimize=_optimize)  # pyright: ignore # See note above.
 
     def get_data(self, path: str) -> bytes:
diff --git a/src/defer_imports/_typing.py b/src/defer_imports/_typing.py
index ad5eb5d..9b1d8a9 100644
--- a/src/defer_imports/_typing.py
+++ b/src/defer_imports/_typing.py
@@ -70,7 +70,7 @@ def __getattr__(name: str) -> object:  # pragma: no cover  # noqa: PLR0911
     if name == "T":
         from typing import TypeVar
 
-        globals()["T"] = T = TypeVar("T")  # pyright: ignore [reportGeneralTypeIssues]
+        globals()[name] = T = TypeVar("T")  # pyright: ignore [reportGeneralTypeIssues]
         return T
 
     if name == "ReadableBuffer":
@@ -81,14 +81,14 @@ def __getattr__(name: str) -> object:  # pragma: no cover  # noqa: PLR0911
 
             ReadableBuffer = Union[bytes, bytearray, memoryview]
 
-        globals()["ReadableBuffer"] = ReadableBuffer
+        globals()[name] = ReadableBuffer
         return ReadableBuffer
 
     if name == "StrPath":
         import os
         from typing import Union
 
-        globals()["StrPath"] = StrPath = Union[str, os.PathLike[str]]
+        globals()[name] = StrPath = Union[str, os.PathLike[str]]
         return StrPath
 
     if name == "TypeAlias":
@@ -99,7 +99,7 @@ def __getattr__(name: str) -> object:  # pragma: no cover  # noqa: PLR0911
             class TypeAlias:
                 """Placeholder for typing.TypeAlias."""
 
-        globals()["TypeAlias"] = TypeAlias
+        globals()[name] = TypeAlias
         return TypeAlias
 
     msg = f"module {__name__!r} has no attribute {name!r}"
diff --git a/src/defer_imports/console.py b/src/defer_imports/console.py
index b987c44..676f30a 100644
--- a/src/defer_imports/console.py
+++ b/src/defer_imports/console.py
@@ -38,7 +38,7 @@ def runsource(self, source: str, filename: str = "<input>", symbol: str = "singl
 
         # Case 3: Input is complete.
         try:
-            tree = DeferredInstrumenter(filename, source, "utf-8").instrument(symbol)
+            tree = DeferredInstrumenter(source, filename, "utf-8").instrument(symbol)
             code = compile(tree, filename, symbol)
         except SyntaxError:
             # Case 1, again.
diff --git a/tests/test_deferred.py b/tests/test_deferred.py
index 08f249f..b8705f1 100644
--- a/tests/test_deferred.py
+++ b/tests/test_deferred.py
@@ -175,7 +175,7 @@ def test_instrumentation(before: str, after: str):
 
     before_bytes = before.encode()
     encoding, _ = tokenize.detect_encoding(io.BytesIO(before_bytes).readline)
-    transformed_tree = DeferredInstrumenter("<unknown>", before_bytes, encoding).instrument()
+    transformed_tree = DeferredInstrumenter(before_bytes, "<unknown>", encoding).instrument()
 
     assert f"{ast.unparse(transformed_tree)}\n" == after