test: generate test corpus for B1/B2/B3 (Word+Excel+PPTX+PDF with hyperlinks/footnotes/charts/SmartArt)

scripts/verify_test_corpus.py

@@ -0,0 +1,41 @@
+"""Quick verification of the test corpus."""
+import zipfile
+import fitz
+
+print("=== test_word.docx ===")
+with zipfile.ZipFile("sample_files/test_corpus/test_word.docx") as z:
+    has_fn = "word/footnotes.xml" in z.namelist()
+    rels = z.read("word/_rels/document.xml.rels").decode()
+    n_hl = rels.count('TargetMode="External"')
+    print(f"  footnotes.xml: {has_fn}, external hyperlinks: {n_hl}")
+
+print("=== test_excel.xlsx ===")
+from openpyxl import load_workbook
+wb = load_workbook("sample_files/test_corpus/test_excel.xlsx")
+ws = wb["Sales"]
+n_comments = sum(1 for row in ws.iter_rows() for c in row if c.comment)
+n_hl = sum(1 for row in ws.iter_rows() for c in row if c.hyperlink)
+n_charts = len(ws._charts) if hasattr(ws, "_charts") else 0
+print(f"  comments: {n_comments}, hyperlinks: {n_hl}, charts: {n_charts}")
+print(f"  sheets: {wb.sheetnames}")
+
+print("=== test_pptx.pptx ===")
+with zipfile.ZipFile("sample_files/test_corpus/test_pptx.pptx") as z:
+    has_diag = "ppt/diagrams/data1.xml" in z.namelist()
+    if has_diag:
+        diag = z.read("ppt/diagrams/data1.xml").decode()
+        n_texts = diag.count("<a:t>")
+    else:
+        n_texts = 0
+    print(f"  diagrams/data1.xml: {has_diag}, <a:t> nodes: {n_texts}")
+
+print("=== test_pdf.pdf ===")
+doc = fitz.open("sample_files/test_corpus/test_pdf.pdf")
+print(f"  pages: {len(doc)}")
+total_links = 0
+total_images = 0
+for p in doc:
+    total_links += len(p.get_links())
+    total_images += len(p.get_images())
+print(f"  total links: {total_links}, total images: {total_images}")
+doc.close()