fix(tsv): honor infer_table_structure when partitioning TSV files

CHANGELOG.md

@@ -1,3 +1,9 @@
+## 0.27.19
+
+### Fixes
+
+- **`partition_tsv()` honours `infer_table_structure`.** The function had no `infer_table_structure` parameter, so the argument was absorbed by `**kwargs` and dropped: `Table.metadata.text_as_html` was set even for `partition(..., infer_table_structure=False)`, or for a `skip_infer_table_types` list containing `tsv`. `partition_csv()`, `partition_xlsx()`, `partition_docx()` and `partition_odt()` all gate that field on the flag, and `decide_table_extraction()` in `partition/auto.py` passes it to every non-special partitioner. The parameter now has the same name, default and docstring as its siblings.
+
 ## 0.27.18
 
 ### Fixes

test_unstructured/partition/test_tsv.py

@@ -23,6 +23,7 @@
 from unstructured.common.html_table import HtmlTable
 from unstructured.documents.elements import Table
 from unstructured.errors import UnprocessableEntityError
+from unstructured.partition.auto import partition
 from unstructured.partition.tsv import partition_tsv
 
 EXPECTED_FILETYPE = "text/tsv"
@@ -286,3 +287,26 @@ def test_partition_tsv_with_implicit_index_columns_matches_pandas_within_the_lim
         index=False, header=True, na_rep=""
     )
     assert table.text == HtmlTable.from_html_text(expected).text
+
+
+# -- `infer_table_structure` ---------------------------------------------------------------------
+
+
+@pytest.mark.parametrize("infer_table_structure", [True, False])
+def test_partition_tsv_from_filename_infer_table_structure(infer_table_structure: bool):
+    elements = partition_tsv(
+        example_doc_path("stanley-cups.tsv"), infer_table_structure=infer_table_structure
+    )
+
+    has_text_as_html = elements[0].metadata.text_as_html is not None
+    assert has_text_as_html == infer_table_structure
+
+
+@pytest.mark.parametrize("infer_table_structure", [True, False])
+def test_partition_tsv_via_partition_respects_infer_table_structure(infer_table_structure: bool):
+    elements = partition(
+        example_doc_path("stanley-cups.tsv"), infer_table_structure=infer_table_structure
+    )
+
+    has_text_as_html = elements[0].metadata.text_as_html is not None
+    assert has_text_as_html == infer_table_structure

unstructured/__version__.py

@@ -1 +1 @@
-__version__ = "0.27.18"  # pragma: no cover
+__version__ = "0.27.19"  # pragma: no cover

unstructured/partition/tsv.py

@@ -30,6 +30,7 @@ def partition_tsv(
     *,
     file: Optional[IO[bytes]] = None,
     include_header: bool = False,
+    infer_table_structure: bool = True,
     **kwargs: Any,
 ) -> list[Element]:
     """Partitions TSV files into document elements.
@@ -42,6 +43,12 @@ def partition_tsv(
         A file-like object using "rb" mode --> open(filename, "rb").
     include_header
         Determines whether or not header info info is included in text and medatada.text_as_html.
+    infer_table_structure
+        If True, any Table elements that are extracted will also have a metadata field
+        named "text_as_html" where the table's text content is rendered into an html string.
+        I.e., rows and cells are preserved.
+        Whether True or False, the "text" field is always present in any Table element
+        and is the text content of the table (no structure).
     """
     exactly_one(filename=filename, file=file)
 
@@ -72,7 +79,7 @@ def partition_tsv(
     metadata = ElementMetadata(
         filename=filename,
         last_modified=get_last_modified_date(filename) if filename else None,
-        text_as_html=html_table.html,
+        text_as_html=html_table.html if infer_table_structure else None,
     )
     metadata.detection_origin = DETECTION_ORIGIN