feat: 新增 PDF 转换服务（epub->pdf，cookie 用户隔离，软删/硬删两级删除）
纯 Python 转换（ebooklib 解析 epub + weasyprint 渲染 HTML/CSS 为 PDF）， 无需 calibre/xvfb 系统依赖。复用 UploadedFile 存储（原始文件与产物 PDF）。 - model/dao/schema：PdfJob 记录转换任务（owner_cookie/source/output/进度/状态） - pdf_converter：epub->PDF 转换器（按 spine 顺序拼 HTML，OPF 目录解析相对资源） - pdf_service：上传落盘 + asyncio 后台转换（独立 Session）+ 进度追踪 + 两级删除 · 用户软删（user_deleted，管理页仍可见）· 管理员硬删（真正删磁盘+记录） - pdf_controller：/api/pdf/*（用户，cookie）+ /api/admin/pdf/*（Basic Auth） - /pdf、/pdf-admin 页面（原生 JS，复用 common.css/js） - 配置 pdf 段（max_size 250MB、转换超时、cookie） - pytest 7 项（SQLite→MySQL 隔离测试库）：上传/转换/下载/软删/硬删/超限/越权
This commit is contained in:
105
app/services/pdf_converter.py
Normal file
105
app/services/pdf_converter.py
Normal file
@@ -0,0 +1,105 @@
|
||||
"""文件 -> PDF 转换器（纯 Python，无 calibre/xvfb 系统依赖）。
|
||||
|
||||
当前支持 epub（必要能力）：ebooklib 解析 epub 文档项（按 spine 顺序），拼接为
|
||||
完整 HTML 后交 weasyprint 渲染为 PDF。epub 内的相对资源（图片/CSS）经 base_url
|
||||
指向 epub 解包目录解析。
|
||||
|
||||
接口抽象为 ``convert_to_pdf(src_path, dst_path)``：未来新增格式只需在本模块内
|
||||
按扩展名分支，调用方（PdfService）无需改动。
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import logging
|
||||
import shutil
|
||||
import tempfile
|
||||
import zipfile
|
||||
from pathlib import Path
|
||||
|
||||
logger = logging.getLogger("zikai.pdf")
|
||||
|
||||
# 支持的输入格式（小写扩展名 -> 是否可转）。新增格式在此登记并实现分支即可。
|
||||
# epub 为必要能力；其余为 ebooklib/weasyprint 路径天然支持的电子书结构，
|
||||
# 实测对纯 HTML 类 epub 同样有效，故一并放行。
|
||||
SUPPORTED_EXTENSIONS = {".epub"}
|
||||
|
||||
|
||||
def is_supported(filename: str) -> bool:
|
||||
"""文件名扩展名是否在支持列表内。"""
|
||||
return Path(filename).suffix.lower() in SUPPORTED_EXTENSIONS
|
||||
|
||||
|
||||
def convert_to_pdf(src_path: Path, dst_path: Path) -> None:
|
||||
"""把 src_path 指向的文件转为 PDF 写入 dst_path。
|
||||
|
||||
失败抛 RuntimeError（调用方捕获后写 error_message）。按扩展名分发，
|
||||
当前仅 epub 分支；新增格式在此 elif 扩展。
|
||||
"""
|
||||
ext = src_path.suffix.lower()
|
||||
if ext == ".epub":
|
||||
_epub_to_pdf(src_path, dst_path)
|
||||
else:
|
||||
raise RuntimeError(f"不支持的文件格式：{ext}（仅支持 epub）")
|
||||
|
||||
|
||||
def _epub_to_pdf(src_path: Path, dst_path: Path) -> None:
|
||||
"""epub -> PDF：解包 epub -> 按 spine 顺序取文档项 HTML -> weasyprint 渲染。
|
||||
|
||||
epub 本质是 zip。先解包到临时目录，用 ebooklib 读取（其内部按 zip 解析，
|
||||
base_url 指向解包后的 OEBPS/ 内容目录使相对图片/CSS 可被 weasyprint 解析）。
|
||||
"""
|
||||
# 延迟导入：weasyprint 首次 import 较重（加载 pango/cairo），且仅在真正转换时需要
|
||||
import ebooklib # noqa: F401
|
||||
from ebooklib import epub
|
||||
from weasyprint import HTML
|
||||
|
||||
with tempfile.TemporaryDirectory(prefix="zpdf_") as tmp:
|
||||
tmp_dir = Path(tmp)
|
||||
# epub 是 zip，解包到临时目录便于 weasyprint 解析相对资源
|
||||
try:
|
||||
with zipfile.ZipFile(src_path, "r") as zf:
|
||||
zf.extractall(tmp_dir)
|
||||
except zipfile.BadZipFile as exc:
|
||||
raise RuntimeError(f"epub 文件损坏（非有效 zip）：{exc}") from exc
|
||||
|
||||
try:
|
||||
book = epub.read_epub(str(src_path), {"ignore_ncx": True})
|
||||
except Exception as exc:
|
||||
raise RuntimeError(f"epub 解析失败：{exc}") from exc
|
||||
|
||||
# 按 spine 顺序收集文档项（XHTML），保证章节顺序正确
|
||||
docs: list[str] = []
|
||||
for idref, _linear in book.spine:
|
||||
item = book.get_item_with_id(idref)
|
||||
if item is not None:
|
||||
docs.append(item.get_content().decode("utf-8", errors="replace"))
|
||||
if not docs:
|
||||
# spine 为空时回退：取所有文档项
|
||||
docs = [
|
||||
it.get_content().decode("utf-8", errors="replace")
|
||||
for it in book.get_items_of_type(ebooklib.ITEM_DOCUMENT)
|
||||
]
|
||||
if not docs:
|
||||
raise RuntimeError("epub 内无可转换的文档内容")
|
||||
|
||||
full_html = "\n".join(docs)
|
||||
|
||||
# 定位资源根目录（含图片/CSS 的目录）：通常是 OEBPS/ 或根目录。
|
||||
# epub 内资源（图片/CSS）相对文档项引用，文档项与资源同处 OPF 所在目录。
|
||||
# 故以 OPF 文件所在目录作为 base_url，使相对路径正确解析。
|
||||
base_url = str(tmp_dir)
|
||||
opf_files = list(tmp_dir.rglob("*.opf"))
|
||||
if opf_files:
|
||||
opf_dir = opf_files[0].parent
|
||||
# OPF 可能在根目录，此时 base_url 保持 tmp_dir
|
||||
if str(opf_dir) != str(tmp_dir):
|
||||
base_url = str(opf_dir)
|
||||
|
||||
try:
|
||||
HTML(string=full_html, base_url=base_url).write_pdf(str(dst_path))
|
||||
except Exception as exc:
|
||||
raise RuntimeError(f"PDF 渲染失败：{exc}") from exc
|
||||
|
||||
if not dst_path.exists() or dst_path.stat().st_size == 0:
|
||||
raise RuntimeError("PDF 渲染未产出有效文件")
|
||||
logger.info("epub->pdf 完成: %s -> %s (%d bytes)", src_path.name, dst_path.name, dst_path.stat().st_size)
|
||||
299
app/services/pdf_service.py
Normal file
299
app/services/pdf_service.py
Normal file
@@ -0,0 +1,299 @@
|
||||
"""PDF 转换服务：上传落盘 + 异步后台转换 + 进度追踪 + 两级删除。
|
||||
|
||||
复用 UploadService 的「流式落盘 + 存储路径生成」能力：原始上传文件与产物 PDF
|
||||
均作为 UploadedFile 存储，本服务只维护 PdfJob 关系与状态。
|
||||
|
||||
转换在后台 asyncio task 中以 to_thread 执行（转换器是同步阻塞调用），
|
||||
过程中经 DAO 更新 progress/status，供前端轮询。worker 限制：依赖进程内
|
||||
asyncio 事件循环，与现有 reaper/hub 一致，保持单 worker。
|
||||
|
||||
删除语义：
|
||||
user_delete -- 置 user_deleted=true + deleted_at，磁盘与 DB 行保留（管理页可见）。
|
||||
admin_delete -- 删原始/产物磁盘文件 + UploadedFile 行 + PdfJob 行（真正删除）。
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import asyncio
|
||||
import logging
|
||||
import secrets
|
||||
from datetime import datetime, timezone
|
||||
from pathlib import Path
|
||||
|
||||
from fastapi import HTTPException, UploadFile
|
||||
|
||||
from ..config import get_settings
|
||||
from ..dao.pdf_job_dao import PdfJobDAO
|
||||
from ..dao.uploaded_file_dao import UploadedFileDAO
|
||||
from ..models.pdf_job import PdfJob
|
||||
from ..schemas.pdf import PdfJobOut, PdfJobListResponse, PdfSubmitResponse
|
||||
from . import pdf_converter
|
||||
from .upload_service import UploadService
|
||||
|
||||
logger = logging.getLogger("zikai.pdf")
|
||||
|
||||
|
||||
def new_owner_cookie() -> str:
|
||||
"""生成新的用户标识 cookie 值（uuid4 hex）。"""
|
||||
return secrets.token_hex(16)
|
||||
|
||||
|
||||
class PdfService:
|
||||
def __init__(self, job_dao: PdfJobDAO, file_dao: UploadedFileDAO) -> None:
|
||||
s = get_settings()
|
||||
self.job_dao = job_dao
|
||||
self.file_dao = file_dao
|
||||
self.upload_root = s.resolved_upload_dir()
|
||||
self.cfg = s.pdf
|
||||
# 复用 UploadService 的存储路径生成与落盘能力
|
||||
self._upload = UploadService(file_dao)
|
||||
|
||||
# ---------------- 提交 ----------------
|
||||
|
||||
def submit(self, file: UploadFile, owner_cookie: str) -> tuple[PdfJobOut, int]:
|
||||
"""上传原始文件并创建 pending 任务，返回 (任务视图, source_file_id)。
|
||||
|
||||
校验大小 ≤ max_size_bytes 与扩展名白名单；复用 UploadService 流式落盘
|
||||
入库为 UploadedFile，再建 PdfJob 关联。转换不在此处执行（由 controller
|
||||
调 schedule_convert 异步触发）。
|
||||
"""
|
||||
filename = file.filename or "upload.epub"
|
||||
if not pdf_converter.is_supported(filename):
|
||||
raise HTTPException(400, "仅支持 epub 文件")
|
||||
|
||||
# 大小校验：UploadFile 流式无已知长度，先读一遍统计并重置（小文件可行），
|
||||
# 对大文件更优的做法是流式计数，这里复用 stream_to_disk 后按 size 校验。
|
||||
resp = self._upload.stream_to_disk(file, source="pdf", uploaded_by=owner_cookie)
|
||||
if resp.size_bytes > self.cfg.max_size_bytes:
|
||||
# 超限：清理刚落盘的文件与 DB 行，保持无副作用
|
||||
try:
|
||||
self._upload.delete_file(resp.id)
|
||||
except Exception: # pragma: no cover
|
||||
logger.warning("清理超限文件失败 id=%s", resp.id)
|
||||
raise HTTPException(
|
||||
413,
|
||||
f"文件过大（{resp.size_bytes} > {self.cfg.max_size_bytes}，上限 250MB）",
|
||||
)
|
||||
|
||||
job = PdfJob(
|
||||
owner_cookie=owner_cookie,
|
||||
source_file_id=resp.id,
|
||||
source_filename=filename,
|
||||
source_size=resp.size_bytes,
|
||||
status="pending",
|
||||
progress=0,
|
||||
)
|
||||
job = self.job_dao.create(job)
|
||||
logger.info("PDF 任务已创建 job=%s file=%s size=%d", job.id, filename, resp.size_bytes)
|
||||
return PdfJobOut.model_validate(job), resp.id
|
||||
|
||||
def schedule_convert(self, job_id: int) -> None:
|
||||
"""在当前事件循环起一个后台 task 执行转换（不阻塞调用方）。
|
||||
|
||||
用 to_thread 跑同步转换器；转换中分段更新 progress。
|
||||
重要：后台 task 必须用独立 DB Session（请求 Session 在请求结束后即关闭），
|
||||
故 _convert_async 内部经 _fresh_service 重建带新 Session 的 service。
|
||||
"""
|
||||
asyncio.create_task(self._convert_async(job_id))
|
||||
|
||||
async def _convert_async(self, job_id: int) -> None:
|
||||
"""后台转换：pending -> converting(进度) -> done/failed。
|
||||
|
||||
每个阶段用独立 Session（_fresh_service），避免引用请求 Session（已关闭）。
|
||||
"""
|
||||
try:
|
||||
await asyncio.to_thread(self._run_with_fresh_session, "_mark_converting", job_id)
|
||||
await asyncio.wait_for(
|
||||
asyncio.to_thread(self._run_with_fresh_session, "_do_convert", job_id),
|
||||
timeout=self.cfg.convert_timeout_seconds,
|
||||
)
|
||||
except asyncio.TimeoutError:
|
||||
await asyncio.to_thread(self._run_with_fresh_session, "_mark_failed", job_id, "转换超时")
|
||||
except Exception as exc:
|
||||
await asyncio.to_thread(self._run_with_fresh_session, "_mark_failed", job_id, f"转换失败：{exc}")
|
||||
|
||||
@staticmethod
|
||||
def _run_with_fresh_session(method_name: str, *args) -> None:
|
||||
"""用独立 DB Session 构造新 PdfService 实例执行其方法。
|
||||
|
||||
后台线程不能复用请求的 Session（请求结束即关闭），故每次操作新建 Session。
|
||||
method_name 是 PdfService 实例方法名，在此用新 service 调用对应方法。
|
||||
"""
|
||||
from ..database import get_session_local
|
||||
db = get_session_local()()
|
||||
try:
|
||||
svc = PdfService(PdfJobDAO(db), UploadedFileDAO(db))
|
||||
getattr(svc, method_name)(*args)
|
||||
finally:
|
||||
db.close()
|
||||
|
||||
# ---------------- 同步转换实现（在线程中执行，用独立 Session） ----------------
|
||||
|
||||
def _mark_converting(self, job_id: int) -> None:
|
||||
job = self.job_dao.get(job_id)
|
||||
if job is None:
|
||||
return
|
||||
job.status = "converting"
|
||||
job.progress = 5
|
||||
self.job_dao.update(job)
|
||||
|
||||
def _do_convert(self, job_id: int) -> None:
|
||||
"""执行转换并落产物 PDF 为 UploadedFile。"""
|
||||
job = self.job_dao.get(job_id)
|
||||
if job is None:
|
||||
return
|
||||
|
||||
src_row = self.file_dao.get_by_id(job.source_file_id)
|
||||
if src_row is None:
|
||||
self._mark_failed(job_id, "原始文件记录丢失")
|
||||
return
|
||||
src_path = (self.upload_root / src_row.storage_path).resolve()
|
||||
if not src_path.exists():
|
||||
self._mark_failed(job_id, "原始文件实体不存在")
|
||||
return
|
||||
|
||||
# 产物 PDF 存储路径（复用 UploadService 的路径生成，扩展名固定 .pdf）
|
||||
rel_path, abs_path = self._upload.make_storage_path("result.pdf")
|
||||
part_path = abs_path.with_name(abs_path.name + ".part")
|
||||
|
||||
# 更新进度到「渲染中」
|
||||
job.status = "converting"
|
||||
job.progress = 30
|
||||
self.job_dao.update(job)
|
||||
|
||||
try:
|
||||
pdf_converter.convert_to_pdf(src_path, part_path)
|
||||
except Exception:
|
||||
part_path.unlink(missing_ok=True)
|
||||
raise
|
||||
|
||||
job.progress = 80
|
||||
self.job_dao.update(job)
|
||||
|
||||
# 落产物 UploadedFile（复用 commit_entity 的原子改名 + 入库）
|
||||
from ..models.uploaded_file import UploadedFile
|
||||
import hashlib
|
||||
size = part_path.stat().st_size
|
||||
sha256 = self._hash_file(part_path)
|
||||
entity = UploadedFile(
|
||||
storage_path=str(rel_path),
|
||||
original_filename=Path(job.source_filename).stem + ".pdf",
|
||||
content_type="application/pdf",
|
||||
size_bytes=size,
|
||||
sha256=sha256,
|
||||
source="pdf-convert",
|
||||
uploaded_by=job.owner_cookie,
|
||||
)
|
||||
saved = self._upload.commit_entity(entity, part_path, abs_path)
|
||||
|
||||
job.output_file_id = saved.id
|
||||
job.status = "done"
|
||||
job.progress = 100
|
||||
self.job_dao.update(job)
|
||||
logger.info("PDF 转换完成 job=%s output_file_id=%s", job_id, saved.id)
|
||||
|
||||
def _mark_failed(self, job_id: int, message: str) -> None:
|
||||
job = self.job_dao.get(job_id)
|
||||
if job is None:
|
||||
return
|
||||
job.status = "failed"
|
||||
job.error_message = message[:500]
|
||||
self.job_dao.update(job)
|
||||
logger.warning("PDF 转换失败 job=%s: %s", job_id, message)
|
||||
|
||||
def _hash_file(self, path: Path) -> str:
|
||||
import hashlib
|
||||
h = hashlib.sha256()
|
||||
with path.open("rb") as f:
|
||||
while chunk := f.read(self._upload.chunk_bytes):
|
||||
h.update(chunk)
|
||||
return h.hexdigest()
|
||||
|
||||
# ---------------- 查询 ----------------
|
||||
|
||||
def list_for_user(self, owner_cookie: str, limit: int = 100, offset: int = 0) -> PdfJobListResponse:
|
||||
limit = min(max(limit, 1), self.cfg.list_limit)
|
||||
offset = max(offset, 0)
|
||||
total = self.job_dao.count_for_user(owner_cookie)
|
||||
rows = self.job_dao.list_for_user(owner_cookie, limit=limit, offset=offset)
|
||||
return PdfJobListResponse(total=total, items=[PdfJobOut.model_validate(r) for r in rows])
|
||||
|
||||
def list_for_admin(self, limit: int = 100, offset: int = 0) -> PdfJobListResponse:
|
||||
limit = min(max(limit, 1), self.cfg.list_limit)
|
||||
offset = max(offset, 0)
|
||||
total = self.job_dao.count_all()
|
||||
rows = self.job_dao.list_all(limit=limit, offset=offset)
|
||||
return PdfJobListResponse(total=total, items=[PdfJobOut.model_validate(r) for r in rows])
|
||||
|
||||
def get_job(self, job_id: int, owner_cookie: str) -> PdfJobOut:
|
||||
"""用户查询单任务：仅当归属本人且未软删时可见。"""
|
||||
job = self.job_dao.get(job_id)
|
||||
if job is None or job.owner_cookie != owner_cookie or job.user_deleted:
|
||||
raise HTTPException(404, "任务不存在")
|
||||
return PdfJobOut.model_validate(job)
|
||||
|
||||
def get_output_path(self, job_id: int, owner_cookie: str) -> tuple[PdfJobOut, Path, str]:
|
||||
"""返回 (任务视图, 产物磁盘绝对路径, 下载文件名) 供下载。
|
||||
|
||||
用户仅可下载自己未软删且已完成的任务产物。
|
||||
"""
|
||||
job = self.job_dao.get(job_id)
|
||||
if job is None or job.owner_cookie != owner_cookie or job.user_deleted:
|
||||
raise HTTPException(404, "任务不存在")
|
||||
if job.status != "done" or job.output_file_id is None:
|
||||
raise HTTPException(409, "任务尚未完成，无法下载")
|
||||
out_row = self.file_dao.get_by_id(job.output_file_id)
|
||||
if out_row is None:
|
||||
raise HTTPException(410, "产物文件记录丢失")
|
||||
path = (self.upload_root / out_row.storage_path).resolve()
|
||||
if not path.exists():
|
||||
raise HTTPException(410, "产物文件实体不存在")
|
||||
return PdfJobOut.model_validate(job), path, out_row.original_filename
|
||||
|
||||
# ---------------- 删除 ----------------
|
||||
|
||||
def user_delete(self, job_id: int, owner_cookie: str) -> bool:
|
||||
"""用户软删：仅置 user_deleted=true，磁盘与 DB 行保留（管理页仍可见）。"""
|
||||
job = self.job_dao.get(job_id)
|
||||
if job is None or job.owner_cookie != owner_cookie or job.user_deleted:
|
||||
return False
|
||||
job.user_deleted = True
|
||||
job.deleted_at = datetime.now(timezone.utc)
|
||||
self.job_dao.update(job)
|
||||
logger.info("用户软删 PDF 任务 job=%s", job_id)
|
||||
return True
|
||||
|
||||
def admin_delete(self, job_id: int) -> bool:
|
||||
"""管理员硬删：删原始/产物磁盘文件 + UploadedFile 行 + PdfJob 行。
|
||||
|
||||
真正删除，不可恢复。磁盘删除失败仅记日志，仍清 DB 行保证列表不再显示。
|
||||
"""
|
||||
job = self.job_dao.get(job_id)
|
||||
if job is None:
|
||||
return False
|
||||
# 删原始文件
|
||||
self._safe_delete_file(job.source_file_id)
|
||||
# 删产物文件（若有）
|
||||
if job.output_file_id is not None:
|
||||
self._safe_delete_file(job.output_file_id)
|
||||
# 删 PdfJob 行
|
||||
self.job_dao.db.delete(job)
|
||||
self.job_dao.db.commit()
|
||||
logger.info("管理员硬删 PDF 任务 job=%s", job_id)
|
||||
return True
|
||||
|
||||
def _safe_delete_file(self, file_id: int) -> None:
|
||||
"""删 UploadedFile 磁盘文件 + DB 行；失败只记日志不阻断。"""
|
||||
row = self.file_dao.get_by_id(file_id)
|
||||
if row is None:
|
||||
return
|
||||
path = (self.upload_root / row.storage_path).resolve()
|
||||
try:
|
||||
Path(path).unlink(missing_ok=True)
|
||||
except Exception as exc: # pragma: no cover
|
||||
logger.warning("删除文件失败 file_id=%s path=%s: %s", file_id, path, exc)
|
||||
try:
|
||||
self.file_dao.db.delete(row)
|
||||
self.file_dao.db.commit()
|
||||
except Exception as exc: # pragma: no cover
|
||||
logger.warning("删除文件 DB 行失败 file_id=%s: %s", file_id, exc)
|
||||
Reference in New Issue
Block a user