fix: PyPDF2 → pdfplumber — сохраняет структуру таблиц в PDF

deploy/services/parse.py

@@ -1,6 +1,5 @@
-"""Parse service — PDF (PyPDF2), DOCX (python-docx), plain text fallback."""
+"""Parse service — PDF (pdfplumber), DOCX (python-docx), plain text fallback."""
 import io
-from PyPDF2 import PdfReader
 
 
 def parse_file(filename, data):
@@ -20,15 +19,30 @@ def parse_file(filename, data):
 
 
 def _parse_pdf(data):
-    reader = PdfReader(io.BytesIO(data))
+    """Parse PDF with pdfplumber — preserves table structure (unlike PyPDF2)."""
+    import pdfplumber
     elements = []
-    for page in reader.pages:
-        text = page.extract_text()
-        if text:
-            for line in text.split("\n"):
-                line = line.strip()
-                if line:
-                    elements.append({"type": "paragraph", "text": line, "style": ""})
+    with pdfplumber.open(io.BytesIO(data)) as pdf:
+        for page in pdf.pages:
+            # Tables first — preserves column structure critical for specs
+            tables = page.extract_tables()
+            for table in tables:
+                if table:
+                    rows = []
+                    for row in table:
+                        if row:
+                            cells = [str(cell or "").strip() for cell in row]
+                            if any(cells):
+                                rows.append(cells)
+                    if rows:
+                        elements.append({"type": "table", "rows": rows})
+            # Remaining text as paragraphs
+            text = page.extract_text()
+            if text:
+                for line in text.split("\n"):
+                    line = line.strip()
+                    if line:
+                        elements.append({"type": "paragraph", "text": line, "style": ""})
     return {"status": "parsed", "element_count": len(elements), "elements": elements}