Skip to content

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

fix(transforms): pass through ragged tables instead of misaligning columns (#1713)

Parideboyc7665ca08863
diff --git a/headroom/transforms/tabular_ingest.py b/headroom/transforms/tabular_ingest.py
index 07b889b13..373b9c771 100644
--- a/headroom/transforms/tabular_ingest.py
+++ b/headroom/transforms/tabular_ingest.py
@@ -134,6 +134,13 @@ def parse_tabular(
if not headers or not rows:
return None
+ # Ragged tables (rows whose cell count differs from the header count)
+ # can't be zipped into records without shifting values under the wrong
+ # column — a compressed table must never state facts the original
+ # didn't (#1652). Treat them as non-tabular and pass through verbatim.
+ width = len(headers)
+ if any(len(row) != width for row in rows):
+ return None
return headers, rows, fmt
diff --git a/tests/test_transforms_tabular.py b/tests/test_transforms_tabular.py
index 858708892..9316ad22f 100644
--- a/tests/test_transforms_tabular.py
+++ b/tests/test_transforms_tabular.py
@@ -153,6 +153,54 @@ def test_parse_markdown_table_drops_separator() -> None:
assert all("---" not in cell for row in rows for cell in row)
+def test_parse_tabular_rejects_ragged_fixed_width(monkeypatch) -> None:
+ # Rows with differing cell counts can't be zipped under the headers
+ # without misattributing columns (#1652) — must pass through.
+ import headroom.transforms.tabular_ingest as ti
+
+ monkeypatch.setattr(
+ ti,
+ "detect_content_type",
+ lambda _c: DetectionResult(ContentType.TABULAR, 0.9, {"format": "fixed_width"}),
+ )
+ ragged = (
+ "tool installed latest status\n"
+ "rtk 0.42.4 0.43.0 update available\n"
+ "rtk ✓ 0.42.4 0.42.4 - up-to-date"
+ )
+ assert ti.parse_tabular(ragged) is None
+
+
+def test_parse_tabular_rejects_ragged_markdown(monkeypatch) -> None:
+ import headroom.transforms.tabular_ingest as ti
+
+ monkeypatch.setattr(
+ ti,
+ "detect_content_type",
+ lambda _c: DetectionResult(ContentType.TABULAR, 0.9, {"format": "markdown"}),
+ )
+ ragged = "| a | b | c |\n| --- | --- | --- |\n| 1 | 2 | 3 |\n| 4 | 5 |"
+ assert ti.parse_tabular(ragged) is None
+
+
+def test_compress_passes_through_ragged_table(monkeypatch) -> None:
+ import headroom.transforms.tabular_ingest as ti
+
+ monkeypatch.setattr(
+ ti,
+ "detect_content_type",
+ lambda _c: DetectionResult(ContentType.TABULAR, 0.9, {"format": "fixed_width"}),
+ )
+ ragged = (
+ "tool installed latest status\n"
+ "rtk 0.42.4 0.43.0 update available\n"
+ "rtk ✓ 0.42.4 0.42.4 - up-to-date"
+ )
+ result = TabularCompressor().compress(ragged)
+ assert not result.was_modified
+ assert result.compressed == ragged
+
+
def test_parse_tabular_returns_none_for_non_tabular() -> None:
assert parse_tabular("just a normal paragraph here") is None