VOC_Monitor/test/dia/trans_p2.py

174 lines
5.9 KiB
Python

import pandas as pd
from pathlib import Path
def parse_excel_file(input_path, diagram_type, line=1):
print(f"🔄 파싱 시작: {input_path} ({diagram_type})")
# 1. 엑셀 로드
raw = pd.read_excel(input_path, sheet_name=0, header=None)
# 2. 헤더 및 데이터 위치 탐색
header_row_idx = None
for i in range(min(15, len(raw))):
if "기지" in raw.iloc[i].astype(str).tolist():
header_row_idx = i
break
if header_row_idx is None:
print(f"⚠️ {input_path}: 헤더(기지)를 찾을 수 없습니다. 건너뜁니다.")
return None
# 3. 데이터프레임 구조화
header = raw.iloc[header_row_idx].astype(str).str.strip().tolist()
df = raw.iloc[header_row_idx + 1:].reset_index(drop=True)
new_cols = []
counts = {}
for col in header:
col_name = col if col != "nan" else "unnamed"
if col_name in counts:
counts[col_name] += 1
new_cols.append(f"{col_name}_{counts[col_name]}")
else:
counts[col_name] = 0
new_cols.append(col_name)
df.columns = new_cols
# 4. 필수 컬럼 식별
track_col = new_cols[0]
direction_col = next((c for c in [str(x) for x in new_cols] if "상하" in c), None)
run_col = next((c for c in [str(x) for x in new_cols] if "입출고" in c), None)
duty_col = next((c for c in [str(x) for x in new_cols] if "DIA" in c or "다이아" in c), None)
train_col = next((c for c in [str(x) for x in new_cols] if "열번" in c), None)
depot_col = next((c for c in [str(x) for x in new_cols] if "기지" in c), None)
# 역 컬럼 정의
meta_keywords = ["상하", "입출고", "DIA", "다이아", "열번", "기지", "track_no", "unnamed"]
station_cols = []
col_found_start = False
for i, col in enumerate(new_cols):
if col == depot_col:
col_found_start = True
continue
if col_found_start:
is_meta = any(k in col for k in meta_keywords)
if not is_meta:
station_cols.append(col)
# 5. 데이터 클렌징
df["train_number"] = pd.to_numeric(df[train_col], errors="coerce")
df = df.dropna(subset=["train_number"])
direction_map = {"": "up", "": "down"}
def classify_run_type(text):
if pd.isna(text) or text == "nan": return "regular"
text = str(text)
if "" in text: return "depot_out"
elif "" in text: return "depot_in"
else: return "regular"
df["direction"] = df[direction_col].map(direction_map).fillna("unknown")
df["run_type"] = df[run_col].apply(classify_run_type)
df["depot_detail"] = df[run_col].astype(str).replace("nan", "")
df["duty_id"] = pd.to_numeric(df[duty_col], errors="coerce").fillna(0).astype(int)
df["depot"] = df[depot_col].astype(str).replace("nan", "")
df["track_no"] = pd.to_numeric(df[track_col], errors="coerce").fillna(0).astype(int)
df["line"] = line
df["diagram_type"] = diagram_type
# 6. 세로형 변환 (Melt)
long_df = df.melt(
id_vars=[
"line", "diagram_type", "duty_id", "train_number",
"direction", "run_type", "depot", "depot_detail", "track_no"
],
value_vars=station_cols,
var_name="station",
value_name="time"
)
# 7. 후처리
long_df = long_df.dropna(subset=["time"])
def format_time(t):
if pd.isna(t) or t == "" or t == "nan": return None
try:
if isinstance(t, str):
# HH:MM:SS format
return t
ts = pd.to_datetime(str(t))
return ts.strftime("%H:%M:%S")
except:
return None
long_df["time"] = long_df["time"].apply(format_time)
long_df = long_df.dropna(subset=["time"])
station_order = {s: i + 1 for i, s in enumerate(station_cols)}
long_df["seq"] = long_df["station"].map(station_order)
return long_df
def convert_all_timetables(output_path="data/unified_timetable.parquet"):
config = [
{"file": "weekday.xlsx", "type": "weekday"},
{"file": "weekend.xlsx", "type": "weekend"},
{"file": "holiday.xlsx", "type": "holiday"}
]
all_dfs = []
for item in config:
file_path = Path("d:/py_train/voc_noti/test/dia") / item["file"]
if file_path.exists():
df = parse_excel_file(file_path, item["type"])
if df is not None:
all_dfs.append(df)
else:
print(f"❌ 파일 없음: {file_path}")
if not all_dfs:
print("❌ 파싱된 데이터가 없습니다.")
return
final_df = pd.concat(all_dfs, ignore_index=True)
# 정렬
final_df = final_df.sort_values(by=["diagram_type", "duty_id", "train_number", "seq"]).reset_index(drop=True)
# 최종 컬럼 순서 및 타입
final_df = final_df[[
"line", "diagram_type", "duty_id", "train_number",
"direction", "run_type", "depot", "depot_detail", "track_no",
"station", "seq", "time"
]]
# 타입 고정
final_df = final_df.astype({
"line": "int8",
"diagram_type": "string",
"duty_id": "int16",
"train_number": "int32",
"direction": "string",
"run_type": "string",
"depot": "string",
"depot_detail": "string",
"track_no": "int16",
"station": "string",
"seq": "int16",
"time": "string"
})
Path(output_path).parent.mkdir(parents=True, exist_ok=True)
final_df.to_parquet(output_path, index=False, engine="pyarrow", compression="zstd")
print(f"\n✅ 통합 데이터 생성 완료: {output_path}")
print(f" - 총 데이터 건수: {len(final_df)}")
print(f" - 다이어그램 타입별 건수:\n{final_df['diagram_type'].value_counts()}")
if __name__ == "__main__":
convert_all_timetables("d:/py_train/voc_noti/test/dia/data/unified_timetable.parquet")