Update app.py

app.py

@@ -333,139 +333,87 @@ def unified_search(bs, bv, bc, cs, cv, cc):
             })
 
         # --------------------------
-        # 비교
+        # 비교 (Section 매칭 기반으로 전면 개편)
         # --------------------------
         if bs and bv and bc and cs and cv and cc:
             df_base = display_data(bs, bv, bc)
             df_comp = display_data(cs, cv, cc)
 
             if 'section' not in df_base.columns or 'section' not in df_comp.columns:
-                return pd.DataFrame({"Error": ["section 없음"]})
-
-            df_base = df_base.rename(columns={
-                "section": f"Section_{bv}",
-                "description": f"Description_{bv}"
-            })
-
-            df_comp = df_comp.rename(columns={
-                "section": f"Section_{cv}",
-                "description": f"Description_{cv}"
-            })
-
-            base_sec = f"Section_{bv}"
-            comp_sec = f"Section_{cv}"
-            base_col = f"Description_{bv}"
-            comp_col = f"Description_{cv}"
+                return pd.DataFrame({"Error": ["해당 데이터에 'section' 컬럼이 존재하지 않습니다."]})
+
+            # 공백 차이로 인한 매칭 실패를 방지하기 위해 띄어쓰기를 제거한 merge_key 생성
+            df_base['merge_key'] = df_base['section'].astype(str).str.replace(" ", "")
+            df_comp['merge_key'] = df_comp['section'].astype(str).str.replace(" ", "")
+
+            # 컬럼명이 겹치지 않도록 임시 변경
+            df_b = df_base[['merge_key', 'section', 'description']].rename(
+                columns={'section': 'sec_b', 'description': 'desc_b'}
+            )
+            df_c = df_comp[['merge_key', 'section', 'description']].rename(
+                columns={'section': 'sec_c', 'description': 'desc_c'}
+            )
+
+            # Section(merge_key) 기준으로 병합 (Full Outer Join)
+            # 기준, 비교 어느 한 쪽에만 있어도 데이터가 보존됨
+            merged = pd.merge(df_b, df_c, on='merge_key', how='outer')
+
+            # Section 번호 순서대로 자연 정렬 (1.1, 1.2, 1.10 등 숫자를 인식하여 정렬)
+            merged['sort_key'] = merged['merge_key'].apply(natural_sort_key)
+            merged = merged.sort_values('sort_key').drop(columns=['sort_key'])
+
+            # 최종 출력할 컬럼명 세팅
+            base_sec_col = f"Section_{bv}"
+            base_desc_col = f"Description_{bv}"
+            comp_sec_col = f"Section_{cv}"
+            comp_desc_col = f"Description_{cv}"
 
             result_rows = []
-            used_comp = set()
-            threshold = 0.6
-
-            def similarity(a, b):
-                return difflib.SequenceMatcher(None, a, b).ratio()
-
-            # 1️⃣ base 기준 매칭
-            for _, b_row in df_base.iterrows():
-                b_sec = b_row[base_sec]
-                b_desc = str(b_row[base_col])
-
-                best_score = 0
-                best_j = None
 
-                for j, c_row in df_comp.iterrows():
-                    if j in used_comp:
-                        continue
-
-                    c_desc = str(c_row[comp_col])
-
-                    if "<img" in b_desc or "<img" in c_desc:
-                        continue
-
-                    score = similarity(b_desc, c_desc)
-
-                    if score > best_score:
-                        best_score = score
-                        best_j = j
-
-                if best_score >= threshold and best_j is not None:
-                    c_row = df_comp.loc[best_j]
-
-                    b_diff, c_diff = highlight_diff(
-                        b_desc, str(c_row[comp_col])
-                    )
+            for _, row in merged.iterrows():
+                # NaN 값을 빈 문자열로 안전하게 변환
+                sec_b = "" if pd.isna(row.get('sec_b')) else str(row['sec_b'])
+                sec_c = "" if pd.isna(row.get('sec_c')) else str(row['sec_c'])
+                desc_b = "" if pd.isna(row.get('desc_b')) else str(row['desc_b'])
+                desc_c = "" if pd.isna(row.get('desc_c')) else str(row['desc_c'])
 
+                # 1. 기준에 있고 비교에 없는 경우
+                if desc_b and not desc_c:
                     result_rows.append({
-                        base_sec: b_sec,
-                        base_col: b_diff,
-                        comp_sec: c_row[comp_sec],
-                        comp_col: c_diff
+                        base_sec_col: sec_b,
+                        base_desc_col: desc_b,
+                        comp_sec_col: "",
+                        comp_desc_col: ""
                     })
-
-                    used_comp.add(best_j)
-
-                else:
+                # 2. 기준에 없고 비교에만 있는 경우
+                elif not desc_b and desc_c:
                     result_rows.append({
-                        base_sec: b_sec,
-                        base_col: b_desc,
-                        comp_sec: "",
-                        comp_col: ""
+                        base_sec_col: "",
+                        base_desc_col: "",
+                        comp_sec_col: sec_c,
+                        comp_desc_col: desc_c
                     })
+                # 3. 둘 다 있는 경우
+                elif desc_b and desc_c:
+                    # 이미지 태그가 포함되어 있거나, 텍스트가 완전히 동일하면 무거운 Diff 생략
+                    if "<img" in desc_b or "<img" in desc_c or desc_b == desc_c:
+                        b_final, c_final = desc_b, desc_c
+                    else:
+                        b_final, c_final = highlight_diff(desc_b, desc_c)
 
-            # 2️⃣ 남은 comp 삽입
-            inserts = {}
-
-            for j, c_row in df_comp.iterrows():
-                if j in used_comp:
-                    continue
-
-                c_desc = str(c_row[comp_col])
-
-                best_pos = None
-                best_score = 0
-
-                for idx, r in enumerate(result_rows):
-                    b_desc = str(r.get(base_col, ""))
-
-                    if not b_desc or "<img" in b_desc or "<img" in c_desc:
-                        continue
-
-                    score = similarity(b_desc, c_desc)
-
-                    if score > best_score:
-                        best_score = score
-                        best_pos = idx
-
-                new_row = {
-                    base_sec: "",
-                    base_col: "",
-                    comp_sec: c_row[comp_sec],
-                    comp_col: c_desc
-                }
-
-                if best_pos is not None:
-                    inserts.setdefault(best_pos + 1, []).append(new_row)
-                else:
-                    inserts.setdefault(len(result_rows), []).append(new_row)
-
-            # 3️⃣ insert 적용
-            final_rows = []
-
-            for idx, row in enumerate(result_rows):
-                final_rows.append(row)
-                if idx in inserts:
-                    final_rows.extend(inserts[idx])
-
-            if len(result_rows) in inserts:
-                final_rows.extend(inserts[len(result_rows)])
-
-            final_df = pd.DataFrame(final_rows)
+                    result_rows.append({
+                        base_sec_col: sec_b,
+                        base_desc_col: b_final,
+                        comp_sec_col: sec_c,
+                        comp_desc_col: c_final
+                    })
 
-            return final_df[[base_sec, base_col, comp_sec, comp_col]]
+            return pd.DataFrame(result_rows)
 
         # --------------------------
-        # ❗ 모든 조건 미충족 → 무조건 return
+        # 모든 조건 미충족 (예외 처리)
         # --------------------------
-        return pd.DataFrame({"Info": ["조회 조건이 올바르지 않습니다"]})
+        return pd.DataFrame({"Info": ["조회 조건이 올바르지 않습니다. 양쪽 버전을 모두 선택해주세요."]})
 
     except Exception as e:
         traceback.print_exc()