hot_comment_radar/app/services/report_service.py

import json
from collections import Counter, defaultdict
from collections.abc import Callable

from sqlalchemy import func, select
from sqlalchemy.orm import Session

from app.models import Comment, ContentItem, Hotspot, Report


DEFAULT_SUMMARY = "总结生成失败，请查看上方统计数据。"


def _labels(value: str | None) -> list[str]:
    try:
        parsed = json.loads(value or "[]")
    except json.JSONDecodeError:
        return []
    return [str(label) for label in parsed] if isinstance(parsed, list) else []


def build_comment_metrics(comments: list[Comment]) -> dict:
    sentiment_counts = Counter(comment.sentiment or "unknown" for comment in comments)
    label_counts = Counter(label for comment in comments for label in _labels(comment.labels))
    total = len(comments)

    def sentiment_entry(name: str) -> dict:
        count = sentiment_counts.get(name, 0)
        return {"count": count, "pct": round((count / total * 100) if total else 0, 2)}

    return {
        "sample_count": total,
        "sentiment": {
            "positive": sentiment_entry("positive"),
            "negative": sentiment_entry("negative"),
            "neutral": sentiment_entry("neutral"),
            "unknown": sentiment_entry("unknown"),
        },
        "top_labels": [{"name": name, "count": count} for name, count in label_counts.most_common(5)],
    }


def select_typical_comments(comments: list[Comment], *, per_sentiment: int = 2) -> dict[str, list[dict]]:
    buckets: dict[str, list[Comment]] = defaultdict(list)
    for comment in comments:
        buckets[comment.sentiment or "unknown"].append(comment)

    result = {}
    for sentiment, values in buckets.items():
        sorted_values = sorted(values, key=lambda c: (c.like_count or 0, c.comment_time or c.created_at), reverse=True)
        result[sentiment] = [
            {"content": comment.content, "like_count": comment.like_count or 0, "comment_id": comment.source_comment_id}
            for comment in sorted_values[:per_sentiment]
        ]
    return result


def generate_item_report(
    session: Session,
    content_item_id: str,
    *,
    summary_provider: Callable[[dict, dict], str] | None = None,
) -> Report:
    item = session.get(ContentItem, content_item_id)
    if item is None:
        raise ValueError("content item not found")
    hotspot = session.get(Hotspot, item.hotspot_id)
    comments = list(session.scalars(select(Comment).where(Comment.content_item_id == item.id)))
    metrics = build_comment_metrics(comments)
    typical = select_typical_comments(comments)
    summary = _safe_summary(summary_provider, metrics, typical, limit=200)
    markdown = _build_markdown(title=item.title or item.source_item_id, metrics=metrics, typical=typical, summary=summary, hotspot_title=hotspot.title if hotspot else "")
    report = Report(
        task_id=item.task_id,
        hotspot_id=item.hotspot_id,
        content_item_id=item.id,
        report_type="item",
        title=item.title or item.source_item_id,
        data=json.dumps(metrics, ensure_ascii=False),
        markdown=markdown,
        metrics_json=json.dumps(metrics, ensure_ascii=False),
        typical_comments_json=json.dumps(typical, ensure_ascii=False),
        summary=summary,
        markdown_content=markdown,
    )
    session.add(report)
    session.commit()
    session.refresh(report)
    return report


def generate_hotspot_report(
    session: Session,
    hotspot_id: str,
    *,
    summary_provider: Callable[[dict, dict], str] | None = None,
) -> Report:
    hotspot = session.get(Hotspot, hotspot_id)
    if hotspot is None:
        raise ValueError("hotspot not found")
    comments = list(session.scalars(select(Comment).where(Comment.hotspot_id == hotspot.id)))
    item_count = session.scalar(select(func.count(ContentItem.id)).where(ContentItem.hotspot_id == hotspot.id)) or 0
    metrics = build_comment_metrics(comments)
    metrics["item_count"] = item_count
    typical = select_typical_comments(comments)
    summary = _safe_summary(summary_provider, metrics, typical, limit=300)
    markdown = _build_markdown(title=hotspot.title, metrics=metrics, typical=typical, summary=summary)
    report = Report(
        task_id=hotspot.task_id,
        hotspot_id=hotspot.id,
        content_item_id=None,
        report_type="hotspot",
        title=hotspot.title,
        data=json.dumps(metrics, ensure_ascii=False),
        markdown=markdown,
        metrics_json=json.dumps(metrics, ensure_ascii=False),
        typical_comments_json=json.dumps(typical, ensure_ascii=False),
        summary=summary,
        markdown_content=markdown,
    )
    session.add(report)
    session.commit()
    session.refresh(report)
    return report


def _safe_summary(summary_provider, metrics: dict, typical: dict, *, limit: int) -> str:
    if summary_provider is None:
        return DEFAULT_SUMMARY
    try:
        summary = summary_provider(metrics, typical)
    except Exception:
        return DEFAULT_SUMMARY
    return str(summary)[:limit]


def _build_markdown(*, title: str, metrics: dict, typical: dict, summary: str, hotspot_title: str = "") -> str:
    lines = [
        f"# {title}",
        "",
    ]
    if hotspot_title:
        lines.extend([f"- 所属热点：{hotspot_title}", ""])
    lines.extend(
        [
            f"- 样本评论数量：{metrics['sample_count']}",
            "",
            "## 情绪分布",
        ]
    )
    for name in ("positive", "neutral", "negative", "unknown"):
        item = metrics["sentiment"][name]
        lines.append(f"- {name}: {item['count']} ({item['pct']}%)")
    lines.extend(["", "## Top 标签"])
    for label in metrics["top_labels"]:
        lines.append(f"- {label['name']}: {label['count']}")
    lines.extend(["", "## 典型评论"])
    for sentiment, comments in typical.items():
        for comment in comments:
            lines.append(f"- [{sentiment}] {comment['content']}")
    lines.extend(["", "## 总结", summary])
    return "\n".join(lines)