Spaces:
Paused
Paused
File size: 25,010 Bytes
c3d8221 d612fd0 e02c52e 5e7e2ec 39fa14e 7c45b63 e02c52e b8e7cc8 d612fd0 aa040e6 d612fd0 e02c52e d612fd0 c0335c0 a8706e1 d612fd0 c3d8221 d612fd0 a8706e1 d612fd0 39fa14e d612fd0 39fa14e c0335c0 5e7e2ec d612fd0 5e7e2ec d612fd0 aa040e6 c3d8221 d612fd0 5e7e2ec 39fa14e d612fd0 c3d8221 a8706e1 ffa044a 5e7e2ec c3d8221 5e7e2ec 89130fe 5e7e2ec a8706e1 d612fd0 c3d8221 5e7e2ec c3d8221 d612fd0 5e7e2ec c3d8221 5e7e2ec c3d8221 39fa14e 88e6883 134724c 5e7e2ec c3d8221 5e7e2ec c3d8221 e304a12 5e7e2ec a8706e1 d612fd0 aa040e6 a8706e1 5e7e2ec c3d8221 5e7e2ec c3d8221 aa040e6 eaf094d aa040e6 c3d8221 aa040e6 5e7e2ec 2b0a6be f91e4ae 2b0fa1b f91e4ae 2b0fa1b f91e4ae 2b0fa1b f91e4ae 2b0fa1b f91e4ae 2b0fa1b f91e4ae 56d718d aa040e6 5e7e2ec aa040e6 56d718d 2b0fa1b 56d718d 5e7e2ec 5d10220 d612fd0 c3d8221 aa040e6 c3d8221 5e7e2ec c3d8221 aa040e6 c3d8221 669ba77 c3d8221 39fa14e 5e7e2ec c3d8221 5e7e2ec c3d8221 d612fd0 c3d8221 d612fd0 39fa14e 2b0fa1b 39fa14e bcc1a18 7c45b63 39fa14e bcc1a18 6e7e94c bcc1a18 6e7e94c bcc1a18 6e7e94c 39fa14e c3d8221 2b0fa1b 39fa14e 2b0fa1b 39fa14e 2b0fa1b 39fa14e 2b0fa1b 7c45b63 39fa14e 7c45b63 39fa14e 2b0fa1b 39fa14e 2b0fa1b 39fa14e 2b0fa1b 39fa14e 2b0fa1b c3d8221 d612fd0 39fa14e aa040e6 39fa14e aa040e6 2b0fa1b 39fa14e c3d8221 c8b0bcc 39fa14e aa040e6 5e7e2ec 39fa14e 5e7e2ec c3d8221 39fa14e c3d8221 39fa14e 206f06a 39fa14e c3d8221 39fa14e aa040e6 c0335c0 5e7e2ec 39fa14e aa040e6 d612fd0 39fa14e c0335c0 5e7e2ec 39fa14e 5e7e2ec c0335c0 39fa14e c0335c0 39fa14e c0335c0 39fa14e c0335c0 5e7e2ec c0335c0 39fa14e d612fd0 c3d8221 5e7e2ec 39fa14e c3d8221 5e7e2ec 2b0fa1b c3d8221 39fa14e c3d8221 39fa14e c0335c0 5e7e2ec c0335c0 5e7e2ec 39fa14e 5e7e2ec c3d8221 5d10220 39fa14e aa040e6 5e7e2ec c3d8221 e02c52e 39fa14e a8706e1 eaf094d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 |
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
process_report.py
保持你原有功能(读取 Excel -> 聚合 -> 导出 Excel/HTML -> 通过 Resend 发送)的前提下,
仅进行了最小必要增强:
1) 增强导出的 Excel 样式,解决:部分边框不全、Q 列以后列无边框、去掉最后一行(若为汇总/空行)的问题;
2) 在 Resend 发送失败并怀疑为 Gmail "550-5.7.1 unsolicited mail" 时,自动尝试用更“简洁/友好”的邮件内容重试一次;
3) 可选 SMTP 回退:当 Resend 失败且你提供 SMTP_* 环境变量时,会尝试通过 SMTP 重发(仅作回退);
其他数据处理逻辑(表头识别、分组、聚合、进度判定)保持原样,未做不必要修改.
"""
import os
import sys
import glob
import json
import base64
from io import BytesIO
from typing import Optional, Tuple, List
from datetime import datetime, date
import pandas as pd
import requests
import smtplib
from email.message import EmailMessage
# ====== 目录 & 环境 ======
INPUT_DIR = os.environ.get("INPUT_DIR", "/data/uploads")
OUTPUT_DIR = os.environ.get("OUTPUT_DIR", "/data/outputs")
os.makedirs(INPUT_DIR, exist_ok=True)
os.makedirs(OUTPUT_DIR, exist_ok=True)
RESEND_API_KEY = os.environ.get("RESEND_API_KEY")
FROM_EMAIL = os.environ.get("FROM_EMAIL")
FROM_NAME = os.environ.get("FROM_NAME") # 可选,显示名
TO_EMAIL = os.environ.get("TO_EMAIL")
TIMEZONE = os.environ.get("TZ", "Asia/Shanghai")
# SMTP 回退(可选)
SMTP_HOST = os.environ.get("SMTP_HOST")
SMTP_PORT = int(os.environ.get("SMTP_PORT", 587) or 587)
SMTP_USER = os.environ.get("SMTP_USER")
SMTP_PASS = os.environ.get("SMTP_PASS")
# ====== 业务相关字段 ======
ALIASES = {
"请购日期": ["请购日期", "请购日", "申请日期"],
"请购单号": ["请购单号", "请购单编号", "申请单号"],
"物料编码": ["物料编码", "物料号", "物料代码"],
"物料名称": ["物料名称", "品名", "名称"],
"纱支密度": ["纱支密度", "纱支/密度", "纱支 密度"],
"门幅(CM)": ["门幅(CM)", "门幅(CM)", "门幅cm", "门幅"],
"颜色": ["颜色", "色号/颜色", "色号"],
"主单位": ["主单位", "单位"],
"主数量": ["主数量", "数量", "请购数量"],
"需求日期": ["需求日期", "需求日", "交期", "要求到货日期"],
"供应商": ["供应商", "供货商", "供应商名称"],
"到货日期": ["到货日期", "实到日期", "收货日期"],
"到货主数量": ["到货主数量", "到货数量", "实到数量"],
"入库日期": ["入库日期", "入库日"],
"入库主数量": ["入库主数量", "入库数量"],
"计划到货日期": ["计划到货日期", "预计到货日期", "承诺到货日期", "计划到货日"],
}
EMAIL_COLS_DEFAULT = [
"请购日期","请购单号","物料编码","物料名称","纱支密度","门幅(CM)","颜色","主单位",
"主数量","需求日期","供应商","到货日期","到货主数量","入库日期","入库主数量","目前进度"
]
# ====== 工具函数(读取 / 归一化 / 聚合) ======
def _today() -> date:
return datetime.now().date()
def _normalize_columns(df: pd.DataFrame) -> pd.DataFrame:
mapped = {}
for std_name, variants in ALIASES.items():
for v in df.columns:
v_clean = str(v).strip()
if v_clean in variants:
mapped[v] = std_name
break
df = df.rename(columns=mapped)
return df
def _find_header_row(path: str, must_have: List[str] = None, try_rows: int = 10) -> int:
must_have = must_have or ["物料编码", "主数量"]
for r in range(try_rows):
try:
df_try = pd.read_excel(path, header=r, nrows=1)
except Exception:
continue
cols = [str(c).strip() for c in df_try.columns]
if all(any(m in c for c in cols) or m in cols for m in must_have):
return r
return 0
def read_system_export(path: str) -> pd.DataFrame:
header_row = _find_header_row(path)
try:
df = pd.read_excel(path, header=header_row)
except Exception:
df = pd.read_excel(path, header=0)
df = df.dropna(axis=1, how="all")
df = _normalize_columns(df)
if "物料名称" in df.columns:
df["物料名称"] = df["物料名称"].astype(str).fillna('')
contains_e = df["物料名称"].str.contains("鹅", na=False)
contains_ya = df["物料名称"].str.contains("鸭", na=False)
contains_huazhu = df["物料名称"].str.contains("华住", na=False)
contains_huazhu_special = df["物料名称"].str.contains("华住专用", na=False)
contains_yuanliao_special = df["物料名称"].str.contains("分原料绒", na=False)
to_remove = (contains_e | contains_ya | contains_huazhu | contains_yuanliao_special) & ~contains_huazhu_special
df = df[~to_remove]
for c in ["请购日期","需求日期","到货日期","入库日期","计划到货日期"]:
if c in df.columns:
df[c] = pd.to_datetime(df[c], errors="coerce")
for c in ["主数量","到货主数量","入库主数量"]:
if c in df.columns:
df[c] = pd.to_numeric(df[c], errors="coerce").fillna(0)
df = df.dropna(how="all")
# ====== 新增代码:开始 ======
# 增加清洗:如果一行连“请购单号”或“物料编码”这种关键ID都没有,则删除
key_cols_to_check = [c for c in ["请购单号", "物料编码"] if c in df.columns]
if key_cols_to_check:
df = df.dropna(subset=key_cols_to_check, how='all')
# ====== 新增代码:结束 ======
return df
def _first_nonnull(series: pd.Series):
return series.dropna().iloc[0] if not series.dropna().empty else None
def aggregate_for_email(df: pd.DataFrame) -> pd.DataFrame:
group_keys = [k for k in [
"请购单号","物料编码","物料名称","纱支密度","门幅(CM)","颜色","主单位","供应商"
] if k in df.columns]
if not group_keys:
raise RuntimeError("找不到用于分组的关键字段(如 请购单号/物料编码 等),请检查导入的表头。")
agg_map = {
"主数量": _first_nonnull,
"请购日期": _first_nonnull,
"需求日期": _first_nonnull,
"到货日期": "max",
"入库日期": "max",
"计划到货日期": "max",
"到货主数量": "sum",
"入库主数量": "sum"
}
final_agg_map = {k: v for k, v in agg_map.items() if k in df.columns}
grouped = df.groupby(group_keys, dropna=False).agg(final_agg_map).reset_index()
grouped["目前进度"] = grouped.apply(_calc_progress_row, axis=1)
# ====== 修改点在这里(新的 get_sort_key 排序逻辑) ======
def get_sort_key(status_text):
status_text = str(status_text)
# 1. 最紧急:逾期 + 未到货
if "逾期" in status_text and "未到货" in status_text:
return 1
# 2. 次紧急:逾期 + 部分到货
if "逾期" in status_text and "部分到货" in status_text:
return 2
# 3. 一般紧急:即将到货
if "未来7天" in status_text:
return 3
# 4. 处理中:部分到货(未逾期)
if "部分到货" in status_text: # 此时已排除逾期
return 4
# 5. 处理中:未到货(未到期)
if "未到货(未到期)" in status_text:
return 5
# 6. 已完成
if "完全到货" in status_text:
return 6
# 7. 其他(如 "处理中" 或其他未匹配状态)
return 7
# ====== 修改结束 ======
grouped['sort_key'] = grouped['目前进度'].apply(get_sort_key)
grouped = grouped.sort_values(by='sort_key').drop(columns=['sort_key'])
if "计划到货日期" in grouped.columns:
grouped = grouped.drop(columns=["计划到货日期"])
final_cols = [col for col in EMAIL_COLS_DEFAULT if col in grouped.columns]
final = grouped[final_cols]
date_cols_to_format = ["请购日期", "需求日期", "到货日期", "入库日期"]
for col in date_cols_to_format:
if col in final.columns:
final[col] = pd.to_datetime(final[col], errors='coerce').dt.strftime('%Y-%m-%d')
return final
def get_sort_key(status_text):
if "逾期" in status_text and "未到货" in status_text:
return 1
if "逾期" in status_text and "部分到货" in status_text:
return 2
if "未来7天" in status_text:
return 3
if "完全到货" in status_text:
return 4
return 5
grouped['sort_key'] = grouped['目前进度'].apply(get_sort_key)
grouped = grouped.sort_values(by='sort_key').drop(columns=['sort_key'])
if "计划到货日期" in grouped.columns:
grouped = grouped.drop(columns=["计划到货日期"])
final_cols = [col for col in EMAIL_COLS_DEFAULT if col in grouped.columns]
final = grouped[final_cols]
date_cols_to_format = ["请购日期", "需求日期", "到货日期", "入库日期"]
for col in date_cols_to_format:
if col in final.columns:
final[col] = pd.to_datetime(final[col], errors='coerce').dt.strftime('%Y-%m-%d')
return final
def _calc_progress_row(row: pd.Series) -> str:
today = pd.Timestamp(_today())
main_qty = float(row.get("主数量", 0) or 0)
arr_qty = float(row.get("到货主数量", 0) or 0)
demand_date = row.get("需求日期", pd.NaT)
arrival_date = row.get("到货日期", pd.NaT)
plan_arrival = row.get("计划到货日期", pd.NaT)
if main_qty > 0 and arr_qty >= main_qty * 0.97:
return "完全到货"
parts: List[str] = []
shortage = max(0.0, main_qty - arr_qty)
if arr_qty > 0:
parts.append(f"部分到货 缺货{shortage:g}米")
elif pd.isna(arrival_date) or arr_qty == 0:
parts.append("未到货")
if (pd.isna(arrival_date) or arr_qty < main_qty) and pd.notna(demand_date):
overdue_days = (today - demand_date).days
if overdue_days > 0:
parts.append(f"已逾期{overdue_days}天")
if pd.notna(plan_arrival):
try:
days_ahead = (plan_arrival.date() - today.date()).days
if 0 <= days_ahead <= 7:
parts.append(f"未来7天到货(计划{str(plan_arrival.date())})")
except Exception:
pass
if not parts:
if pd.notna(demand_date) and today <= pd.Timestamp(demand_date.date()):
return "未到货(未到期)"
return "处理中"
return ";".join(parts)
def _find_latest_input(input_dir: str) -> Optional[str]:
files = []
for pat in ("*.xlsx", "*.xls"):
files.extend(glob.glob(os.path.join(input_dir, pat)))
if not files:
return None
files.sort(key=os.path.getmtime, reverse=True)
return files[0]
# ====== 导出 Excel:增强样式,解决边框/最后一行问题(仅替换导出逻辑) ======
def _df_to_styled_excel_bytes(df: pd.DataFrame) -> bytes:
"""
增强导出:
- 如果最后一行是全空或包含合计/总计/nan文字,删除它(用户不需要最后一行)
- 另外:如果最后一行“看起来像汇总行”(大部分为数值),也会删除
- 设置统一边框,仅覆盖到实际数据的最后一列(不延伸到 Q 列之后)
- 标红逾期行
- 自动列宽,但不破坏原数据
"""
# ---------- 更稳健的最后一行判断与删除 ----------
if len(df) > 0:
last = df.tail(1)
# 1) 如果整行全为空 -> 删除
if last.isna().all(axis=1).iloc[0]:
df = df.iloc[:-1, :]
else:
# 将最后一行所有单元格值转换为字符串并清洗(去 BOM / 零宽等)
def _clean_val(x):
try:
s = str(x)
except Exception:
s = ""
s = s.replace('\ufeff', '').replace('\u200b', '').strip().lower()
return s
last_strs = last.applymap(_clean_val).iloc[0].tolist()
joined = " ".join([s for s in last_strs if s])
# 2) 如果包含明确关键字 -> 删除
keywords = ["合计", "总计", "小计", "汇总", "nan", "total", "sum"]
if any(k in joined for k in keywords):
df = df.iloc[:-1, :]
else:
# 3) 判断“数值比例”:如果最后一行中多数单元格看起来是数值(>=60%),则认定为汇总行并删除
num_cols = len(last_strs)
numeric_count = 0
for v in last_strs:
if v == "":
continue
# 尝试解析为 float(去掉千分分隔符逗号)
v2 = v.replace(',', '')
try:
float(v2)
numeric_count += 1
continue
except Exception:
pass
# 允许带单位的数字(例如 "301575.6米")——提取前面数字
import re
m = re.match(r'^[\+\-]?\d+(\.\d+)?', v2)
if m:
numeric_count += 1
# 规则阈值:数值比例 >= 0.6 或者数值个数 >= max(2, 40%*cols) 时视作汇总
if num_cols > 0:
ratio = numeric_count / num_cols
else:
ratio = 0.0
if (numeric_count >= max(2, int(num_cols * 0.4))) or (ratio >= 0.6):
df = df.iloc[:-1, :]
# ---------- 原有的导出处理(后续代码保持不变) ----------
# 下面从你原来的代码继续(不要删改),例如:
# bio = BytesIO()
# writer = pd.ExcelWriter(bio, engine='xlsxwriter')
# ...
bio = BytesIO()
writer = pd.ExcelWriter(bio, engine='xlsxwriter')
sheet_name = '采购执行表'
df.fillna('').to_excel(writer, sheet_name=sheet_name, index=False)
workbook = writer.book
worksheet = writer.sheets[sheet_name]
# 样式
header_format = workbook.add_format({
'bold': True, 'font_name': 'Arial', 'font_size': 10,
'border': 1, 'align': 'center', 'valign': 'vcenter', 'bg_color': '#f6f6f6'
})
border_fmt = workbook.add_format({
'font_name': 'Arial', 'font_size': 10, 'border': 1
})
overdue_fmt = workbook.add_format({
'font_name': 'Arial', 'font_size': 10, 'border': 1, 'font_color': 'red'
})
# 写表头
for col_num, value in enumerate(df.columns.values):
worksheet.write(0, col_num, value, header_format)
# 仅在实际数据范围内画边框(最后一列 = df.shape[1]-1)
nrows = len(df)
# 如果 df 为空,确保 ncols 不为负
ncols = max(df.shape[1] - 1, 0)
# conditional_format with 'no_errors' ensures each cell gets the format; header is row 0, data rows 1..nrows
# last row index for conditional_format should be nrows (since header 0, data rows occupy 1..nrows)
worksheet.conditional_format(1, 0, nrows, ncols, {'type': 'no_errors', 'format': border_fmt})
# 标红逾期行(与原逻辑保持一致)
try:
progress_col_idx = df.columns.get_loc('目前进度')
for row_num in range(len(df)):
status_text = str(df.iloc[row_num, progress_col_idx])
if "逾期" in status_text:
worksheet.set_row(row_num + 1, None, overdue_fmt)
except Exception:
# 如果没有 '目前进度' 列会抛出,但不影响导出
pass
# 自动列宽(限制最大宽度)
for i, col in enumerate(df.columns):
try:
column_len = df[col].astype(str).str.len().max()
column_len = max(column_len, len(col) * 2)
worksheet.set_column(i, i, min(column_len, 40))
except Exception:
worksheet.set_column(i, i, 20)
writer.close()
bio.seek(0)
return bio.read()
# ====== 构造 HTML 邮件体(增强友好性:加入简短说明和签名) ======
def _build_html_body(df: pd.DataFrame, title: str, simple: bool = False) -> str:
"""
如果 simple=True,会生成更简洁、带文本说明+签名的邮件体(用于 Gmail 友好的重试)
默认生成详细表格版。
"""
if simple:
# 简洁版:短说明 + 附件提示 + 联系信息(更容易被过滤器接受)
intro = f"<p>您好,</p><p>附件为 {title},请查收。如有疑问,请回复本邮件联系。</p>"
signature = "<p>--<br>采购部<br>联系人: 采购组<br>电话/邮箱: example@company.com</p>"
html = f"<html><head><meta charset='utf-8'></head><body>{intro}{signature}</body></html>"
return html
# 默认完整版:表格 + 说明 + 签名
table_html = df.fillna('').to_html(index=False, escape=False, na_rep="")
intro = f"<p>您好,</p><p>以下为 {title},请查阅(表中为系统聚合后的采购执行明细):</p>"
signature = "<p>--<br>采购部<br>联系人: 采购组<br>电话/邮箱: example@company.com</p>"
html = f"""<html><head><meta charset="utf-8" /><style>
table {{ border-collapse: collapse; font-size: 13px; }}
table, th, td {{ border: 1px solid #ccc; padding: 6px; }}
th {{ background:#f6f6f6; }}
tr:nth-child(even) {{background-color: #f2f2f2;}}
</style></head><body>{intro}{table_html}{signature}</body></html>"""
return html
# ====== 发送:Resend + Gmail 检测 + 可选 SMTP 回退(最小改动) ======
def _send_email_via_resend(subject: str, html_body: str, attachment_bytes: Optional[bytes], attachment_name: str) -> Tuple[bool, str]:
"""
使用 Resend API 发送邮件(带附件 base64)。
- 如果返回文本包含 Gmail 的 550-5.7.1 unsolicited 信息,将返回该信息,调用者可根据该信息选择重试(此文件中实现了自动简洁重试)。
"""
if not (RESEND_API_KEY and FROM_EMAIL and TO_EMAIL):
return False, "缺少 Resend 配置(RESEND_API_KEY / FROM_EMAIL / TO_EMAIL)"
if FROM_NAME:
from_field = f"{FROM_NAME} <{FROM_EMAIL}>"
else:
from_field = FROM_EMAIL
url = "https://api.resend.com/emails"
headers = {
"Authorization": f"Bearer {RESEND_API_KEY}",
"Content-Type": "application/json"
}
payload = {
"from": from_field,
"to": [TO_EMAIL],
"subject": subject,
"html": html_body,
# 增加 Reply-To 与 Message-ID,有助于通过垃圾邮件过滤
"headers": {
"Reply-To": FROM_EMAIL,
"Message-ID": f"<{int(datetime.now().timestamp())}@{FROM_EMAIL.split('@')[-1]}>"
}
}
if attachment_bytes is not None:
payload["attachments"] = [
{"filename": attachment_name, "content": base64.b64encode(attachment_bytes).decode("utf-8")}
]
try:
resp = requests.post(url, headers=headers, data=json.dumps(payload, ensure_ascii=False).encode('utf-8'), timeout=30)
try:
resp_text = resp.text
except Exception:
resp_text = repr(resp.content)
if 200 <= resp.status_code < 300:
return True, f"Resend OK: {resp_text}"
else:
# 检测是否为 Gmail 550-5.7.1 / unsolicited
lowered = resp_text.lower()
if "550-5.7.1" in lowered or "unsolicited" in lowered or "unsolicited mail" in lowered or "message rejected" in lowered:
# 明确标识为 Gmail 垃圾判定
extra = ";检测到可能的 Gmail unsolicited/550-5.7.1 拒收提示。"
return False, f"Resend 返回失败状态 {resp.status_code}: {resp_text}{extra}"
return False, f"Resend 返回失败状态 {resp.status_code}: {resp_text}"
except requests.exceptions.RequestException as e:
return False, f"邮件发送请求失败: {e}"
def _send_email_via_smtp(subject: str, html_body: str, attachment_bytes: Optional[bytes], attachment_name: str) -> Tuple[bool, str]:
"""
可选 SMTP 回退(仅在 Resend 失败时尝试)。
需要在环境变量中配置 SMTP_HOST/SMTP_PORT/SMTP_USER/SMTP_PASS。
"""
if not (SMTP_HOST and SMTP_USER and SMTP_PASS and FROM_EMAIL and TO_EMAIL):
return False, "缺少 SMTP 配置(SMTP_HOST / SMTP_USER / SMTP_PASS / FROM_EMAIL / TO_EMAIL)"
try:
msg = EmailMessage()
sender = f"{FROM_NAME} <{FROM_EMAIL}>" if FROM_NAME else FROM_EMAIL
msg["From"] = sender
msg["To"] = TO_EMAIL
msg["Subject"] = subject
msg["Reply-To"] = FROM_EMAIL
msg.set_content("这是 HTML 邮件,请使用支持 HTML 的客户端查看。")
msg.add_alternative(html_body, subtype='html')
if attachment_bytes is not None:
msg.add_attachment(attachment_bytes, maintype='application', subtype='vnd.openxmlformats-officedocument.spreadsheetml.sheet', filename=attachment_name)
smtp = smtplib.SMTP(SMTP_HOST, SMTP_PORT, timeout=30)
smtp.starttls()
smtp.login(SMTP_USER, SMTP_PASS)
smtp.send_message(msg)
smtp.quit()
return True, "SMTP 发送成功(回退)"
except Exception as e:
return False, f"SMTP 发送失败(回退): {e}"
# ====== 主流程:保持原逻辑,仅在发送处增加 Gmail 简洁重试与 SMTP 回退 ======
def run_once(file_path: Optional[str] = None) -> dict:
try:
if file_path is None:
file_path = _find_latest_input(INPUT_DIR)
if not file_path:
return {"ok": False, "msg": f"未在 {INPUT_DIR} 找到Excel输入文件"}
raw = read_system_export(file_path)
final = aggregate_for_email(raw)
out_name = f"采购执行表_{datetime.now().strftime('%Y%m%d')}.xlsx"
out_path = os.path.join(OUTPUT_DIR, out_name)
attach = _df_to_styled_excel_bytes(final)
final.to_excel(out_path, index=False)
subject = f"采购执行表自动推送 {datetime.now().date()}"
html = _build_html_body(final, title=f"采购执行表({datetime.now().date()})", simple=False)
# 首选:Resend
ok, info = _send_email_via_resend(subject, html, attachment_bytes=attach, attachment_name=out_name)
# 若 Resend 返回怀疑为 Gmail unsolicited(含 550-5.7.1),尝试一次“简洁模板”重试(只重试一次)
if not ok and ("550-5.7.1" in info or "unsolicited" in info.lower() or "unsolicited mail" in info.lower()):
# 生成更简洁的邮件内容(少表格、带签名提示),再次尝试 Resend
simple_html = _build_html_body(final, title=f"采购执行表({datetime.now().date()})", simple=True)
ok2, info2 = _send_email_via_resend(subject, simple_html, attachment_bytes=attach, attachment_name=out_name)
if ok2:
return {"ok": True, "msg": "Resend 初次被 Gmail 判断为 unsolicited,已用简洁模板重试并发送成功", "input": file_path, "output": out_path, "rows": len(final), "resend_info": info, "resend_retry_info": info2}
# 如果简洁重试也失败了,继续下面的 SMTP 回退逻辑或直接返回失败信息
info = f"{info};尝试简洁模板重试结果: {info2}"
# 如果仍未发送成功,并且配置了 SMTP 回退,则尝试 SMTP
if not ok and SMTP_HOST and SMTP_USER and SMTP_PASS:
ok_smtp, info_smtp = _send_email_via_smtp(subject, html, attachment_bytes=attach, attachment_name=out_name)
if ok_smtp:
return {"ok": True, "msg": "Resend 失败,但通过 SMTP 回退发送成功", "input": file_path, "output": out_path, "rows": len(final), "resend_info": info, "smtp_info": info_smtp}
else:
return {"ok": False, "msg": f"Resend 失败: {info};SMTP 回退也失败: {info_smtp}", "input": file_path, "output": out_path, "rows": len(final)}
return {"ok": ok, "msg": "邮件发送成功" if ok else f"邮件发送失败:{info}", "input": file_path, "output": out_path, "rows": len(final)}
except Exception as e:
import traceback
return {"ok": False, "msg": f"处理文件时发生严重错误: {e}", "traceback": traceback.format_exc()}
def main():
arg_file = sys.argv[1] if len(sys.argv) > 1 else None
result = run_once(arg_file)
print(json.dumps(result, ensure_ascii=False))
if not result.get("ok"):
sys.exit(1)
if __name__ == "__main__":
main()
|