import io, os, re from typing import List, Dict import streamlit as st import pandas as pd # --- PDF text import pdfplumber from pypdf import PdfReader # --- OCR from pdf2image import convert_from_bytes import pytesseract from PIL import Image # ====================================================================== # SCHEMA TABELLA (colonne fisse) # ====================================================================== SCHEMA = [ "Piece","SKU","Title","Capacity","% Recycled","Weight","Color","Material / Resin","Class","Source File", # nuove colonne "Component","Function","General description of the packaging","Material Ref GCAS","Material Family" ] # ====================================================================== # ESTRATTORI LOW-LEVEL # ====================================================================== def extract_text_pages(pdf_bytes: bytes) -> List[str]: pages = [] # 1) pdfplumber try: with pdfplumber.open(io.BytesIO(pdf_bytes)) as pdf: for p in pdf.pages: pages.append(p.extract_text() or "") except Exception: pages = [] # 2) pypdf fallback if not pages or all(not (t or "").strip() for t in pages): try: reader = PdfReader(io.BytesIO(pdf_bytes)) pages = [(p.extract_text() or "") for p in reader.pages] except Exception: pages = [] return pages def run_ocr(pdf_bytes: bytes, lang: str, dpi: int, tesseract_cmd: str | None) -> List[str]: if tesseract_cmd: pytesseract.pytesseract.tesseract_cmd = tesseract_cmd images = convert_from_bytes(pdf_bytes, dpi=dpi) texts = [] config = "--psm 6 -c preserve_interword_spaces=1" for img in images: if not isinstance(img, Image.Image): img = img.convert("RGB") texts.append(pytesseract.image_to_string(img, lang=lang, config=config) or "") return texts # --- OCR rapido SOLO per il peso (prime pagine, DPI bassi, stop appena trovato) def run_ocr_for_weight(pdf_bytes: bytes, lang: str, tesseract_cmd: str | None, max_pages: int = 2, dpi_weight: int = 200) -> str: if tesseract_cmd: pytesseract.pytesseract.tesseract_cmd = tesseract_cmd images = convert_from_bytes(pdf_bytes, dpi=dpi_weight, first_page=1, last_page=max_pages) config = "--psm 6 -c preserve_interword_spaces=1" acc = [] for img in images: if not isinstance(img, Image.Image): img = img.convert("RGB") txt = pytesseract.image_to_string(img, lang=lang, config=config) or "" w = weight_from(txt) # definita sotto if w: return w acc.append(txt) return weight_from("\n".join(acc)) or "" # ====================================================================== # PARSING DOMINIO (euristiche/regex leggere) # ====================================================================== SKU_RE = re.compile(r"\b(?:Name|SKU|Part(?:\s*No\.?)?)\s*[:#]?\s*([A-Z0-9\-_/\.]{5,})", re.I) TITLE_RE = re.compile(r"\bTitle\s*[:\-]\s*(.+)", re.I) CLASS_RE = re.compile(r"\bClass\s*([A-Za-z ]+)", re.I) def _first(text: str, pattern: re.Pattern, group: int = 1) -> str: m = pattern.search(text or "") return m.group(group).strip() if m else "" def capacity_from(text: str) -> str: m = re.search(r"([0-9]+(?:[.,][0-9]+)?)\s*(L|Liter|ml|mL)\b", text or "", re.I) if not m: return "" unit = m.group(2).upper().replace("LITER","L").replace("ML","ml") return f"{m.group(1).replace(',', '.')} {unit}" def color_from(text: str) -> str: m = re.search(r"(?:Part\s*Color|Color)\s*[:\-]?\s*([A-Z ]{3,})", text, re.I) if m: return m.group(1).strip() m = re.search(r"\b([A-Z ]{4,}(?:GREEN|TRANSPARENT|WHITE|BLACK|BLUE|RED|CLEAR)[A-Z ]*)\b", text) return (m.group(1).strip() if m else "") def material_from(text: str) -> str: for line in (text or "").splitlines(): if re.search(r"\bRESIN\b", line, re.I): return line.strip() m = re.search(r"(SERIOPLAST.*?RESIN)", text, re.I) return m.group(1).strip() if m else "" # ====================================================================== # WEIGHT: prendi TUTTA la riga a partire da "Weight ..." e normalizza spazi/OCR # Esempio: "Weight 9 4 +/- 3 g" -> "Weight 94±3g" # ====================================================================== WEIGHT_LINE_RE = re.compile(r"(?is)\bweight\b[^\n\r]*") def _normalize_weight_line(s: str) -> str: s = (s or "").strip() # comprimi spazi ripetuti s = re.sub(r"\s+", " ", s) # togli spazi interni tra cifre (OCR: "9 4" -> "94") s = re.sub(r"(?<=\d)\s+(?=\d)", "", s) # unifica simboli ± s = re.sub(r"\+\s*/\s*-\s*|\+\s*-\s*", "±", s) s = s.replace("＋－", "±").replace("﹢", "+").replace("－", "-") # rimuovi spazi attorno a ± s = re.sub(r"\s*±\s*", "±", s) # rimuovi spazi prima dell'unità s = re.sub(r"\s+(?=(?:mg|g|kg)\b)", "", s, flags=re.I) # punti/virgole s = s.replace(",", ".") return s def weight_from(text: str) -> str: if not text: return "" # preferisci la prima riga che contiene anche l'unità lines = [m.group(0) for m in WEIGHT_LINE_RE.finditer(text)] for ln in lines: if re.search(r"\b(?:mg|g|kg)\b", ln, re.I): return _normalize_weight_line(ln) # se non trovata unità, restituisci comunque la prima occorrenza normalizzata return _normalize_weight_line(lines[0]) if lines else "" # --------------------- PIECE da "Packaging Component Type" --------------------- _ALLOWED_PIECES = { "ribbon": "ribbon", "bottle": "bottle", "film bundle": "film bundle", "container": "container", "label - adhesive": "LABEL - ADHESIVE", "label adhesive": "LABEL - ADHESIVE", "label-adhesive": "LABEL - ADHESIVE", "label - back": "LABEL - BACK", "back label": "LABEL - BACK", "label back": "LABEL - BACK", "closure": "CLOSURE", } _PACK_COMP_TYPE_RE = re.compile( r"Packaging\s+Component\s+Type\s*[:\-]?\s*([^\n\r]+)", re.I ) def _normalize_piece(s: str) -> str: s0 = (s or "").strip() s1 = re.sub(r"\s+", " ", s0) s2 = s1.lower() s2 = s2.replace("–", "-").replace("—", "-") s2 = s2.replace("label- ", "label ").replace(" -", " - ").strip() if s2 in _ALLOWED_PIECES: return _ALLOWED_PIECES[s2] s2 = s2.replace(" ", " ") if s2 in _ALLOWED_PIECES: return _ALLOWED_PIECES[s2] for key, canon in _ALLOWED_PIECES.items(): if key in s2: return canon return "" def piece_from(text: str, cls: str) -> str: m = _PACK_COMP_TYPE_RE.search(text or "") if m: val = m.group(1) normalized = _normalize_piece(val) if normalized: return normalized m2 = re.search(r"Packaging\s*Material\s*Type\s*([^\n]+)", text or "", re.I) if m2: seg = m2.group(1) norm = _normalize_piece(seg) if norm: return norm if cls: norm = _normalize_piece(cls) if norm: return norm if "bottle" in cls.lower(): return "bottle" if "cap" in cls.lower() or "closure" in cls.lower(): return "CLOSURE" if "corrugated" in cls.lower(): return "container" if "label" in cls.lower(): return "LABEL - BACK" return "" # --- Nuove colonne: euristiche base FUNCTION_RE = re.compile(r"\b(Primary|Secondary(?:\s*or\s*Tertiary)?|Tertiary)\b", re.I) def component_from(text: str, piece: str, cls: str) -> str: txt = text.lower() if "ink" in txt and "cartridge" in txt: return "Ink cartridge" if "ink foil" in txt: return "Ink foil" if "tape" in txt: return "Tape" if "label" in txt and ("psl" in txt or "wet glue" in txt or "iml" in txt or "htl" in txt): return "Labels" if "adhesive" in txt or "hot melt" in txt: return "Adhesive" if "cartonboard" in txt or "sheet" in txt: return "Cartonboard / Sheet" if "corrugated" in txt or "case" in txt or "outercase" in txt: return "Corrugated box" if "bundle" in txt: return "Bundle" if piece: return piece if cls: if "bottle" in cls.lower(): return "Bottle" if "cap" in cls.lower(): return "Closure" if "corrugated" in cls.lower(): return "Corrugated box" if "label" in cls.lower(): return "Labels" return "" def function_from(text: str) -> str: m = FUNCTION_RE.search(text or "") return m.group(1).title() if m else "" def material_ref_gcas_from(text: str) -> str: m = re.findall(r"\b(\d{7,9})\b", text or "") if m: seen = set(); out=[] for x in m: if x not in seen: seen.add(x); out.append(x) return ", ".join(out[:3]) m2 = re.findall(r"\((\d{5,})\s*kg\s*pack\)", text or "", re.I) if m2: seen=set(); out=[] for x in m2: if x not in seen: seen.add(x); out.append(x) return ", ".join(out[:3]) return "" def material_family_from(text: str) -> str: families = [ "Monolayer HDPE","Polypropylene (PP)","Paper","Flexible Film – Mono non Metallized", "Flexible - Label PSL WGL IML HTL","Rigid Paper – Corrugated Case", "Inks and solvents","Hot melt adhesive","Wet Glue Label", "Coated paper","Wood","Ink foil","Fasson PE 85 TOP White" ] t = text or "" for fam in families: if fam.lower() in t.lower(): return fam if re.search(r"\bHDPE\b", t): return "Monolayer HDPE" if re.search(r"\bPP\b|\bPolypropylene\b", t, re.I): return "Polypropylene (PP)" if "corrugated" in t.lower(): return "Rigid Paper – Corrugated Case" if "paper" in t.lower(): return "Paper" return "" def parse_record(pages: List[str], source_name: str) -> Dict[str, str]: full = "\n".join(pages or [""]) sku = _first(full, SKU_RE) title = _first(full, TITLE_RE) cls = _first(full, CLASS_RE) cap = capacity_from(title) or capacity_from(full) color = color_from(full) material = material_from(full) piece = piece_from(full, cls) # nuove colonne comp = component_from(full, piece, cls) func = function_from(full) gcas = material_ref_gcas_from(full) mfam = material_family_from(full) # WEIGHT: prendi l'intera riga "Weight ..." wght = weight_from(full) return { "Piece": piece or "", "SKU": sku or "", "Title": title or "", "Capacity": cap or "", "% Recycled": "–", "Weight": wght or "–", "Color": color or "", "Material / Resin": material or "", "Class": cls or "", "Source File": source_name, "Component": comp or "", "Function": func or "", "General description of the packaging": "", "Material Ref GCAS": gcas or "", "Material Family": mfam or "" } # ====================================================================== # UI STREAMLIT # ====================================================================== st.set_page_config(page_title="PDF → Table (OCR-ready)", layout="wide") st.title("📄→📊 PDF → Table (OCR-ready)") st.caption("Carica PDF (anche scansioni). Compilo la tabella con i campi richiesti; OCR mirato per il peso.") with st.sidebar: files = st.file_uploader("Seleziona PDF", type=["pdf"], accept_multiple_files=True) st.markdown("---") st.subheader("OCR") ocr_fallback = st.checkbox("Usa OCR se non c'è testo", value=True) ocr_lang = st.text_input("Lingue OCR (comma)", value="eng,ita") ocr_dpi = st.number_input("DPI OCR", 200, 600, 300, 50) tess_path = st.text_input("Percorso Tesseract (se non nel PATH)", value="") run_btn = st.button("▶️ Estrai") if not run_btn: st.info("Carica i PDF e premi **Estrai**.") st.stop() if not files: st.warning("Nessun PDF caricato.") st.stop() lang = "+".join([p.strip() for p in ocr_lang.split(",") if p.strip()]) or "eng" tess_cmd = tess_path.strip() or None rows, errors = [], [] for up in files: try: raw = up.read() pages = extract_text_pages(raw) # Se il PDF non ha testo estraibile, OCR completo una sola volta if ocr_fallback and not any((p or "").strip() for p in pages): pages = run_ocr(raw, lang=lang, dpi=int(ocr_dpi), tesseract_cmd=tess_cmd) rec = parse_record(pages, up.name) # Se Weight è vuoto, OCR rapido (prime pagine) e stop appena trovato if (not rec.get("Weight") or rec["Weight"] == "–") and ocr_fallback: w_ocr = run_ocr_for_weight(raw, lang=lang, tesseract_cmd=tess_cmd, max_pages=2, dpi_weight=200) if w_ocr: rec["Weight"] = w_ocr rows.append(rec) except Exception as e: errors.append((up.name, str(e))) if errors: with st.expander("Errori"): for name, err in errors: st.error(f"{name}: {err}") df = pd.DataFrame(rows, columns=SCHEMA) st.success(f"Creat{ 'e' if len(df)!=1 else 'a' } {len(df)} riga/e.") st.dataframe(df, use_container_width=True) c1, c2 = st.columns(2) with c1: st.download_button("⬇️ CSV", df.to_csv(index=False).encode("utf-8"), "table.csv", "text/csv") with c2: bio = io.BytesIO() with pd.ExcelWriter(bio, engine="openpyxl") as xw: df.to_excel(xw, index=False, sheet_name="data") st.download_button("⬇️ Excel", bio.getvalue(), "table.xlsx", "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet")