ax-backend/app/services/document_service.py

import re
from datetime import datetime, timezone

from sqlalchemy import select
from sqlalchemy.ext.asyncio import AsyncSession

from app.core.exceptions import ContentTooLargeError, DocumentNotFoundError
from app.models.document import Document
from app.schemas.document import CreateDocumentRequest, UpdateDocumentRequest

CONTENT_MAX_BYTES = 200_000  # 200KB


class DocumentService:
    def __init__(self, db: AsyncSession) -> None:
        self.db = db

    # ------------------------------------------------------------------ #
    # CREATE
    # ------------------------------------------------------------------ #
    async def create_document(
        self,
        data: CreateDocumentRequest,
        user_id: str | None = None,
    ) -> Document:
        if len(data.content.encode("utf-8")) > CONTENT_MAX_BYTES:
            raise ContentTooLargeError(len(data.content.encode("utf-8")))

        doc = Document(
            title=data.title,
            content=data.content,
            format=data.format,
            session_id=data.session_id,
            template_id=data.template_id,
            created_by=user_id,
        )
        self.db.add(doc)
        await self.db.commit()
        await self.db.refresh(doc)
        return doc

    # ------------------------------------------------------------------ #
    # READ
    # ------------------------------------------------------------------ #
    async def get_document(self, document_id: str) -> Document:
        result = await self.db.execute(
            select(Document).where(Document.id == document_id)
        )
        doc = result.scalar_one_or_none()
        if doc is None:
            raise DocumentNotFoundError(document_id)
        return doc

    async def list_documents(
        self,
        page: int = 1,
        page_size: int = 20,
        session_id: str | None = None,
        sort_by: str = "updated_at",
        sort_order: str = "desc",
    ) -> tuple[list[Document], int]:
        query = select(Document)

        if session_id:
            query = query.where(Document.session_id == session_id)

        # 排序
        sort_col = getattr(Document, sort_by, Document.updated_at)
        if sort_order == "asc":
            query = query.order_by(sort_col.asc())
        else:
            query = query.order_by(sort_col.desc())

        # 总数
        count_result = await self.db.execute(
            query.with_only_columns(Document.id)
        )
        total = len(count_result.all())

        # 分页
        offset = (page - 1) * page_size
        result = await self.db.execute(query.offset(offset).limit(page_size))
        docs = list(result.scalars().all())
        return docs, total

    # ------------------------------------------------------------------ #
    # UPDATE
    # ------------------------------------------------------------------ #
    async def update_document(
        self, document_id: str, data: UpdateDocumentRequest
    ) -> Document:
        doc = await self.get_document(document_id)

        if data.title is not None:
            doc.title = data.title

        if data.content is not None:
            if len(data.content.encode("utf-8")) > CONTENT_MAX_BYTES:
                raise ContentTooLargeError(len(data.content.encode("utf-8")))
            doc.content = data.content

        if data.blocks is not None:
            doc.content = self._apply_block_updates(doc.content, data.blocks)

        doc.updated_at = datetime.now(timezone.utc)
        await self.db.commit()
        await self.db.refresh(doc)
        return doc

    # ------------------------------------------------------------------ #
    # DELETE
    # ------------------------------------------------------------------ #
    async def delete_document(self, document_id: str) -> None:
        doc = await self.get_document(document_id)
        await self.db.delete(doc)
        await self.db.commit()

    # ------------------------------------------------------------------ #
    # 局部块更新：按 level + index 定位并替换对应的标题块
    # ------------------------------------------------------------------ #
    @staticmethod
    def _apply_block_updates(content: str, blocks: list) -> str:
        """
        将文档按标题行拆分成若干块，按 level+index 替换对应块后重新拼接。

        块的定义：以标题行（# / ## / ...）为分割点，
                  每个标题及其下属正文构成一个块。
        """
        lines = content.split("\n")
        # 找出所有标题行的位置
        heading_pattern = re.compile(r"^(#{1,6})\s+")
        heading_positions: list[tuple[int, int]] = []  # (line_index, level)
        for i, line in enumerate(lines):
            m = heading_pattern.match(line)
            if m:
                level = len(m.group(1))
                heading_positions.append((i, level))

        # 统计每个 level 已出现的次数，得到 index
        level_counter: dict[int, int] = {}
        heading_info: list[tuple[int, int, int]] = []  # (line_idx, level, index)
        for line_idx, level in heading_positions:
            idx = level_counter.get(level, 0)
            heading_info.append((line_idx, level, idx))
            level_counter[level] = idx + 1

        # 构建查找字典 (level, index) → line_idx
        heading_map: dict[tuple[int, int], int] = {
            (level, index): line_idx
            for line_idx, level, index in heading_info
        }

        # 将 lines 拆成块
        # 块边界 = 各标题行的 line_idx
        split_points = sorted({line_idx for line_idx, _, _ in heading_info})
        split_points.append(len(lines))  # 末尾哨兵

        # 前置正文（首个标题之前的内容）
        pre_content_end = split_points[0] if split_points else len(lines)
        chunks: list[str] = []
        chunks.append("\n".join(lines[:pre_content_end]))

        # 各标题块
        block_keys: list[tuple[int, int] | None] = [None]  # 对应 pre_content
        for i, sp in enumerate(split_points[:-1]):
            end = split_points[i + 1]
            chunks.append("\n".join(lines[sp:end]))
            _, level, index = heading_info[i]
            block_keys.append((level, index))

        # 执行替换
        for block_update in blocks:
            key = (block_update.level, block_update.index)
            if key in heading_map:
                chunk_idx = block_keys.index(key)
                chunks[chunk_idx] = block_update.content

        return "\n".join(chunks).strip() + "\n"