xiaoxia-saas/apps/worker/worker_app/tasks/compose_video.py

"""视频合成 Celery 任务 — Phase 8 任务 2.10.

使用 JobService 管理任务生命周期，通过 RenderAdapter 调用 UnifiedRenderService 执行合成。
"""

from __future__ import annotations

import os
import subprocess  # pragma: no cover
import tempfile
from pathlib import Path

from celery.exceptions import SoftTimeLimitExceeded  # pragma: no cover
from celery.utils.log import get_task_logger
from worker_app.celery_app import celery_app
from worker_app.db import SessionLocal

logger = get_task_logger(__name__)

# 任务超时时间（秒）：超过此时间 Celery 会抛出 SoftTimeLimitExceeded
RENDER_TASK_SOFT_TIME_LIMIT = 600  # pragma: no cover  # 10 分钟
# 硬超时：超过此时间进程会被强制 kill
RENDER_TASK_TIME_LIMIT = 660  # pragma: no cover  # 10 分钟 + 1 分钟清理缓冲


def _get_job_service():
    """延迟导入 JobService，避免循环依赖。"""
    from apps.api.app.services.job_service import JobService
    from packages.adapters.sqlalchemy_impl.job_repository import SQLAlchemyJobRepository

    db = SessionLocal()
    repo = SQLAlchemyJobRepository(db)
    return JobService(repo), db


@celery_app.task(
    name="worker.compose_video",
    bind=True,
    max_retries=3,
    default_retry_delay=60,
    soft_time_limit=RENDER_TASK_SOFT_TIME_LIMIT,
    time_limit=RENDER_TASK_TIME_LIMIT,
)
def compose_video(self, job_id: str, **kwargs):  # pragma: no cover
    """视频合成任务。

    使用 UnifiedRenderService（图层架构）进行渲染。

    Args:
        job_id: JobService 中的任务 ID
        **kwargs: 来自 Job.payload 的额外参数（plan_id, output_path 等）
    """
    job_service, db = _get_job_service()

    try:
        job = job_service.get_job(job_id)
        if job is None:
            logger.error("Job not found: %s", job_id)
            return {"status": "error", "message": f"Job {job_id} not found"}

        plan_id = job.payload.get("plan_id", "")
        if not plan_id:
            job_service.fail_job(job_id, "Missing plan_id in job payload")
            return {"status": "error", "message": "Missing plan_id"}

        # 使用 unified 渲染引擎
        return _compose_with_unified_engine(self, job_service, job, plan_id, db)

    except SoftTimeLimitExceeded:
        # Celery 软超时：任务执行超过 soft_time_limit
        error_msg = f"渲染任务超时（超过 {RENDER_TASK_SOFT_TIME_LIMIT // 60} 分钟）"
        logger.error("视频合成超时: job_id=%s", job_id)
        try:
            job_service.fail_job(job_id, error_msg)
        except Exception:
            logger.exception("更新 Job 超时失败状态时出错")
        # 超时不重试
        return {"status": "error", "message": error_msg, "error_type": "timeout"}

    except subprocess.TimeoutExpired as exc:
        # FFmpeg 子进程超时
        error_msg = f"FFmpeg 渲染超时（{exc.timeout}s）"
        logger.error("视频合成 FFmpeg 超时: job_id=%s timeout=%s", job_id, exc.timeout)
        try:
            job_service.fail_job(job_id, error_msg)
        except Exception:
            logger.exception("更新 Job 超时失败状态时出错")
        # 超时不重试
        return {"status": "error", "message": error_msg, "error_type": "ffmpeg_timeout"}

    except self.retry_exc as exc:
        logger.warning("视频合成重试中: job_id=%s, exc=%s", job_id, exc)
        raise

    except subprocess.CalledProcessError as exc:
        # FFmpeg 执行失败，提取有意义的错误信息
        from video_processing.video_validation import get_exit_code_message

        exit_msg = get_exit_code_message(exc.returncode)
        stderr_text = (exc.stderr or "").strip()
        stderr_tail = stderr_text[-300:] if len(stderr_text) > 300 else stderr_text
        error_msg = f"渲染失败: {exit_msg}"
        if stderr_tail:
            error_msg += f" | {stderr_tail[:200]}"

        logger.error("视频合成 FFmpeg 失败: job_id=%s %s", job_id, exit_msg)
        try:
            job_service.fail_job(job_id, error_msg[:500])
        except Exception:
            logger.exception("更新 Job 失败状态时出错")
        # FFmpeg 错误不重试（通常是素材或配置问题）
        return {"status": "error", "message": error_msg, "error_type": "ffmpeg_error", "exit_code": exc.returncode}

    except Exception as exc:
        logger.exception("视频合成异常: job_id=%s", job_id)
        try:
            job_service.fail_job(job_id, str(exc)[:500])
        except Exception:
            logger.exception("更新 Job 失败状态时出错")
        raise self.retry(exc=exc, countdown=60) from exc
    finally:
        db.close()


def _compose_with_unified_engine(task, job_service, job, plan_id: str, db) -> dict:  # pragma: no cover
    """新引擎渲染路径（UnifiedRenderService + RenderAdapter）。"""
    job_id = job.id

    # 标记为 running
    job_service.update_progress(job_id, progress=10.0, current_stage="初始化统一渲染引擎")

    from video_processing.render_adapter import RenderAdapter

    adapter = RenderAdapter(db)

    # 校验合成条件
    job_service.update_progress(job_id, progress=15.0, current_stage="校验合成条件")
    valid, errors, warnings, ready_count, total_count = adapter.validate_plan(plan_id)
    if not valid:
        error_msg = "; ".join(errors)
        job_service.fail_job(job_id, f"合成校验失败: {error_msg}")
        return {"status": "error", "message": error_msg}

    # 进度回调
    def progress_cb(progress: float, stage: str) -> None:
        try:
            job_service.update_progress(job_id, progress=progress, current_stage=stage)
        except Exception:
            logger.exception("更新进度失败")

    # 执行渲染
    job_service.update_progress(job_id, progress=20.0, current_stage="开始渲染")
    logger.info("统一渲染引擎开始: job_id=%s plan_id=%s", job_id, plan_id)

    result = adapter.render_plan(
        plan_id=plan_id,
        job_id=job_id,
        progress_cb=progress_cb,
    )

    if not result.success:
        error_msg = f"渲染失败: {result.error_message}"
        job_service.fail_job(job_id, error_msg[:500])
        raise RuntimeError(result.error_message)

    # 更新 Job 状态为完成
    result_data = {
        "plan_id": plan_id,
        "output_path": str(result.output_path) if result.output_path else "",
        "storage_key": f"rendered/{plan_id}/{job_id}.mp4",
        "output_url": result.output_url,
        "estimated_duration": result.duration,
        "clip_count": result.clip_count,
        "engine": "unified",
        "width": result.width,
        "height": result.height,
        "file_size": result.file_size,
    }
    job_service.complete_job(job_id, result=result_data)

    logger.info(
        "视频合成完成(unified): job_id=%s plan_id=%s duration=%.2fs",
        job_id,
        plan_id,
        result.duration,
    )
    return {"status": "completed", "job_id": job_id, "result": result_data}


def _cleanup_output(job_id: str) -> None:
    """清理临时输出文件。"""
    try:
        _output_dir = os.environ.get("VIDEO_OUTPUT_DIR", os.path.join(tempfile.gettempdir(), "video_output"))
        output_path = os.path.join(_output_dir, f"{job_id}.mp4")
        if Path(output_path).exists():
            Path(output_path).unlink()
    except Exception as e:
        logger.warning(f"清理输出文件失败: {e}", exc_info=True)