tabella-docs 2.0.3__tar.gz → 2.0.4__tar.gz

{tabella_docs-2.0.3 → tabella_docs-2.0.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tabella-docs
-Version: 2.0.3
+Version: 2.0.4
 Summary: Portable document lifecycle suite — create (.docx/.xlsx/.pptx/.pdf/.odt/.ods/.odp), convert (pandoc), and read (anydoc + LiteParse OCR) documents. Agent Plugins v1 compatible.
 Author-email: D Yiapanis <d@yiapanis.co>
 License-Expression: MIT

{tabella_docs-2.0.3 → tabella_docs-2.0.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "tabella-docs"
-version = "2.0.3"
+version = "2.0.4"
 description = "Portable document lifecycle suite — create (.docx/.xlsx/.pptx/.pdf/.odt/.ods/.odp), convert (pandoc), and read (anydoc + LiteParse OCR) documents. Agent Plugins v1 compatible."
 readme = "README.md"
 license = "MIT"

{tabella_docs-2.0.3 → tabella_docs-2.0.4}/tabella/__init__.py

@@ -13,7 +13,7 @@ from pathlib import Path
 
 from .bootstrap import resolve_venv_python
 
-__version__ = "2.0.3"
+__version__ = "2.0.4"
 
 # ── Configuration ─────────────────────────────────────────────
 

{tabella_docs-2.0.3 → tabella_docs-2.0.4}/tabella/read.py

@@ -104,7 +104,8 @@ def read(path: str, max_chars: Optional[int] = None, extract_images: bool = True
       1. Plain text → direct file read (fast path)
       2. Office formats + text PDFs → anydoc (Markdown output)
       3. Images → LiteParse (OCR)
-      4. PDFs that anydoc returns empty → LiteParse (OCR fallback for scanned PDFs)
+      4. PDFs with no text layer — anydoc returns empty or raises NeedsOcrError →
+         LiteParse (OCR fallback for scanned PDFs)
     """
     suffix = Path(path).suffix.lower()
 
@@ -126,15 +127,20 @@ def read(path: str, max_chars: Optional[int] = None, extract_images: bool = True
         anydoc_error = None
         try:
             text = _run_anydoc(path)
-        except (RuntimeError, ImportError) as e:
-            anydoc_error = str(e)
+        except Exception as e:
+            # A fallback boundary: any primary-engine failure means "try the
+            # other engine". anydoc raises NeedsOcrError for a PDF with no text
+            # layer, and its hierarchy (NeedsOcrError -> ConvertError ->
+            # Exception) is theirs, not ours — enumerating it is what made the
+            # OCR lane below unreachable for scanned PDFs.
+            anydoc_error = f"{type(e).__name__}: {e}"
 
         # 4. If anydoc returned empty text (scanned PDF) or failed, try LiteParse OCR
         if not text.strip():
             try:
                 data = _run_liteparse(path, extract_images=extract_images)
                 text = data.get("text") or ""
-            except (RuntimeError, ImportError) as liteparse_err:
+            except Exception as liteparse_err:
                 if anydoc_error:
                     raise RuntimeError(f"anydoc failed: {anydoc_error}; liteparse fallback also failed: {liteparse_err}")
                 raise RuntimeError(f"both anydoc and liteparse returned no text. liteparse: {liteparse_err}")

{tabella_docs-2.0.3 → tabella_docs-2.0.4}/tabella_docs.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: tabella-docs
-Version: 2.0.3
+Version: 2.0.4
 Summary: Portable document lifecycle suite — create (.docx/.xlsx/.pptx/.pdf/.odt/.ods/.odp), convert (pandoc), and read (anydoc + LiteParse OCR) documents. Agent Plugins v1 compatible.
 Author-email: D Yiapanis <d@yiapanis.co>
 License-Expression: MIT

{tabella_docs-2.0.3 → tabella_docs-2.0.4}/tabella_docs.egg-info/SOURCES.txt

@@ -15,5 +15,6 @@ tabella_docs.egg-info/top_level.txt
 tests/test_bootstrap_paths.py
 tests/test_ensure_venv_win.py
 tests/test_odt_table.py
+tests/test_read_ocr_fallback.py
 tests/test_security_gates.py
 tests/test_text_section_alias.py

tabella_docs-2.0.4/tests/test_read_ocr_fallback.py

@@ -0,0 +1,49 @@
+"""doc_read on an image-only PDF must reach the OCR lane, not raise.
+
+anydoc signals a missing text layer by *raising* NeedsOcrError (MRO:
+NeedsOcrError -> ConvertError -> Exception, so not a RuntimeError), where the
+fallback was gated on empty text and caught only RuntimeError/ImportError.
+Net effect: the fallback was unreachable for exactly the input it exists for.
+
+Runs the real pipeline — PIL makes the page, Tesseract reads it — because a
+mocked empty return is what let this ship.
+"""
+
+from pathlib import Path
+
+import pytest
+
+pytest.importorskip("PIL")
+pytest.importorskip("anydoc")
+pytest.importorskip("liteparse")
+
+from PIL import Image, ImageDraw, ImageFont  # noqa: E402
+
+
+def _page(tmp_path: Path):
+    img = Image.new("RGB", (1200, 400), "white")
+    draw = ImageDraw.Draw(img)
+    try:
+        font = ImageFont.truetype(
+            "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", 64
+        )
+    except OSError:
+        font = ImageFont.load_default(size=64)
+    draw.text((60, 120), "SCANNED HELLO 42", fill="black", font=font)
+    return img
+
+
+def test_image_only_pdf_falls_back_to_ocr(tmp_path):
+    png, pdf = tmp_path / "scan.png", tmp_path / "scan.pdf"
+    img = _page(tmp_path)
+    img.save(png)
+    img.save(pdf, "PDF", resolution=150.0)
+
+    from tabella.read import read
+
+    from_png = " ".join(read(str(png)).split())
+    from_pdf = " ".join(read(str(pdf)).split())
+
+    assert "HELLO" in from_png.upper(), f"OCR lane itself is broken: {from_png!r}"
+    assert from_pdf, "image-only PDF returned no text"
+    assert from_pdf == from_png, f"PDF and PNG disagree: {from_pdf!r} vs {from_png!r}"