fix: Support text partitioning from ZipExtFile objects

test_unstructured/partition/common/test_common.py

@@ -2,6 +2,8 @@
 from multiprocessing import Pool
 
 import numpy as np
+from io import BytesIO
+from unstructured.partition.common.common import convert_to_bytes
 import pytest
 from PIL import Image
 from unstructured_inference.constants import IsExtracted
@@ -466,3 +468,10 @@ def test_normalize_layout_element_layout_element_text_source_metadata():
     assert hasattr(element, "metadata")
     assert hasattr(element.metadata, "is_extracted")
     assert element.metadata.is_extracted == "true"
+
+def test_convert_to_bytes_preserves_seekable_file_position():
+    file = BytesIO(b"hello world")
+    file.seek(6)
+
+    assert convert_to_bytes(file) == b"hello world"
+    assert file.tell() == 6

test_unstructured/partition/test_auto.py

@@ -3,6 +3,7 @@
 from __future__ import annotations
 
 import json
+import zipfile
 import os
 import pathlib
 import tempfile
@@ -1118,6 +1119,28 @@ def test_auto_partition_forwards_include_page_breaks_to_partition_pdf():
 
 # -- metadata_filename ----------------------------------------------------
 
+def test_partition_txt_from_zip_ext_file(tmp_path):
+    zip_path = tmp_path / "text_file.zip"
+
+    with zipfile.ZipFile(zip_path, "w") as zf:
+        zf.writestr(
+            "hello.txt",
+            "Hello from inside a zip file.\nThis should be partitioned successfully.",
+        )
+
+    with zipfile.ZipFile(zip_path, "r") as zf:
+        info = zf.infolist()[0]
+
+        with zf.open(info) as file:
+            elements = partition(
+                file=file,
+                metadata_filename=info.filename,
+            )
+
+    assert [element.text for element in elements] == [
+        "Hello from inside a zip file.",
+        "This should be partitioned successfully.",
+    ]
 
 def test_auto_partition_forwards_metadata_filename_via_kwargs():
     with open(example_doc_path("fake-text.txt"), "rb") as f:

unstructured/partition/common/common.py

@@ -371,7 +371,8 @@ def spooled_to_bytes_io_if_needed(file: _T | SpooledTemporaryFile[bytes]) -> _T
 def convert_to_bytes(file: bytes | IO[bytes]) -> bytes:
     """Extract the bytes from `file` without preventing it from being read again later.
 
-    As a convenience to simplify client code, also returns `file` unchanged if it is already bytes.
+    As a convenience to simplify client code, also returns `file` unchanged if it is already
+    bytes.
     """
     if isinstance(file, bytes):
         return file
@@ -389,6 +390,30 @@ def convert_to_bytes(file: bytes | IO[bytes]) -> bytes:
         with open(file.name, "rb") as f:
             return f.read()
 
+    if hasattr(file, "read"):
+        original_position = None
+
+        if hasattr(file, "tell") and hasattr(file, "seek"):
+            try:
+                original_position = file.tell()
+                file.seek(0)
+            except (OSError, ValueError):
+                original_position = None
+
+        data = file.read()
+
+        if original_position is not None:
+            try:
+                file.seek(original_position)
+            except (OSError, ValueError):
+                pass
+
+        if isinstance(data, str):
+            return data.encode()
+
+        if isinstance(data, bytes):
+            return data
+
     raise ValueError("Invalid file-like object type")