From bb68f25cd63ab76c2bb532ab18e0e49f7457d5e5 Mon Sep 17 00:00:00 2001
From: adenzhou1350 <209601943+adenzhou1350@users.noreply.github.com>
Date: Tue, 6 Oct 2026 16:17:27 +0800
Subject: [PATCH] Accept binary file-like SQL streams

---
 sqlparse/__init__.py        |  4 ++--
 sqlparse/lexer.py           |  4 ++--
 tests/test_binary_stream.py | 47 +++++++++++++++++++++++++++++++++++++
 3 files changed, 51 insertions(+), 4 deletions(-)
 create mode 100644 tests/test_binary_stream.py

diff --git a/sqlparse/__init__.py b/sqlparse/__init__.py
index 8411f5a3..6744a0af 100644
--- a/sqlparse/__init__.py
+++ b/sqlparse/__init__.py
@@ -30,11 +30,11 @@ def parse(
 
 
 def parsestream(
-    stream: str | IO[str], encoding: str | None = None
+    stream: str | bytes | IO[str] | IO[bytes], encoding: str | None = None
 ) -> Generator[sql.Statement, None, None]:
     """Parses sql statements from file-like object.
 
-    :param stream: A file-like object.
+    :param stream: SQL text or a text or binary file-like object.
     :param encoding: The encoding of the stream contents (optional).
     :returns: A generator of :class:`~sqlparse.sql.Statement` instances.
     """
diff --git a/sqlparse/lexer.py b/sqlparse/lexer.py
index 5d4a5f97..0e569905 100644
--- a/sqlparse/lexer.py
+++ b/sqlparse/lexer.py
@@ -12,7 +12,7 @@
 # http://pygments.org/
 # It's separated from the rest of pygments to increase performance
 # and to allow some customizations.
-from io import TextIOBase
+from io import IOBase
 from threading import Lock
 
 from sqlparse import keywords, tokens
@@ -116,7 +116,7 @@ def get_tokens(self, text, encoding=None):
 
         ``stack`` is the initial stack (default: ``['root']``)
         """
-        if isinstance(text, TextIOBase):
+        if isinstance(text, IOBase):
             text = text.read()
 
         if isinstance(text, str):
diff --git a/tests/test_binary_stream.py b/tests/test_binary_stream.py
new file mode 100644
index 00000000..b738a19e
--- /dev/null
+++ b/tests/test_binary_stream.py
@@ -0,0 +1,47 @@
+"""Text and binary streams should share the existing decoding path."""
+import gzip
+from io import BytesIO, StringIO
+
+import pytest
+
+import sqlparse
+
+
+@pytest.mark.parametrize('encoding', ['utf-8', 'latin-1'])
+@pytest.mark.parametrize('compressed', [False, True])
+def test_parsestream_binary(encoding, compressed):
+    # issue352: GzipFile is a binary stream, not a TextIOBase.
+    text = "SELECT 'caf\N{LATIN SMALL LETTER E WITH ACUTE}'; SELECT 2;"
+    data = text.encode(encoding)
+    if compressed:
+        stream = gzip.GzipFile(fileobj=BytesIO(gzip.compress(data)))
+    else:
+        stream = BytesIO(data)
+    with stream:
+        result = [str(stmt) for stmt in sqlparse.parsestream(stream, encoding)]
+        assert not stream.closed
+    assert result == [str(stmt) for stmt in sqlparse.parse(text)]
+
+
+def test_parsestream_binary_default_encoding():
+    text = "SELECT '\N{CJK UNIFIED IDEOGRAPH-65E5}';"
+    with BytesIO(text.encode('utf-8')) as stream:
+        assert [str(stmt) for stmt in sqlparse.parsestream(stream)] == [text]
+
+
+def test_parsestream_text():
+    text = "SELECT 'caf\N{LATIN SMALL LETTER E WITH ACUTE}';"
+    with StringIO(text) as stream:
+        assert [str(stmt) for stmt in sqlparse.parsestream(stream)] == [text]
+        assert not stream.closed
+
+
+def test_parsestream_binary_bad_encoding():
+    with BytesIO(b"SELECT '\xff';") as stream:
+        with pytest.raises(UnicodeDecodeError):
+            list(sqlparse.parsestream(stream, encoding='utf-8'))
+
+
+def test_parsestream_invalid_object():
+    with pytest.raises(TypeError, match='Expected text or file-like object'):
+        list(sqlparse.parsestream(object()))