#!/usr/bin/env bash
# 构建 PRISM 出版级 PDF 全书
#   1. 预处理 14 个 markdown(剥离 emoji、改链接)
#   2. pandoc 转 LaTeX body
#   3. 注入模板 → XeLaTeX 编译两次(目录)
#   4. 用 pypdf 合并 cover + body → PRISM_Whole.pdf
#
# 用法:
#   bash plans/PRISM/tools/build_book.sh
#
# 输出:
#   plans/PRISM/PRISM_Book.pdf    正文部分
#   plans/PRISM/PRISM_Whole.pdf   封面 + 正文 完整版

set -euo pipefail

ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/../../.." && pwd)"
PRISM_DIR="$ROOT/plans/PRISM"
TOOLS_DIR="$PRISM_DIR/tools"
BUILD_DIR="$PRISM_DIR/.build"
mkdir -p "$BUILD_DIR"

# ── 版本与构建日期(注入封面/页眉/章节正文)──────────────────────────
if [[ -f "$PRISM_DIR/VERSION" ]]; then
  VERSION="$(tr -d '[:space:]' < "$PRISM_DIR/VERSION")"
else
  VERSION="0.0.0"
fi
BUILD_DATE="$(date +'%Y-%m-%d')"
echo "==> Building PRISM Book v${VERSION} (${BUILD_DATE})"

echo "==> 1a) 生成统计章节 16_stats.md"
python3 "$TOOLS_DIR/gen_stats.py" "$PRISM_DIR" "$PRISM_DIR/16_stats.md"

echo "==> 1a') 生成变更日志章节 17_changelog.md (从 CHANGELOG.md)"
python3 "$TOOLS_DIR/gen_changelog_chapter.py" \
  "$PRISM_DIR/CHANGELOG.md" \
  "$PRISM_DIR/17_changelog.md" \
  "$VERSION" "$BUILD_DATE"

echo "==> 1) 预处理 markdown 文件"
# 章节顺序
CHAPTERS=(
  "README.md"
  "00_overview.md"
  "01_capability_decomposition.md"
  "02_architecture.md"
  "03_data_schema.md"
  "04_pipeline_A_iphone_offline.md"
  "05_pipeline_B_relocalization.md"
  "06_pipeline_C_online_perception.md"
  "07_pipeline_D_consolidation.md"
  "08_runtime_timeline.md"
  "09_roadmap.md"
  "10_tech_stack.md"
  "11_world_model_bridge.md"
  "12_risks.md"
  "13_evaluation.md"
  "14_mvp.md"
  "15_glossary.md"
  "16_stats.md"
  "18_lyra_inspirations.md"
  "17_changelog.md"
)

# Python 预处理:把每个 md 的一级标题改成 LaTeX \chapter,
# 移除 emoji / 修复链接 / 删除 raw HTML
python3 - <<'PYEOF'
import re, sys, pathlib
PRISM_DIR = pathlib.Path("plans/PRISM")
OUT = PRISM_DIR / ".build" / "merged.md"

# 章节顺序
order = [
  "README.md",
  "00_overview.md",
  "01_capability_decomposition.md",
  "02_architecture.md",
  "03_data_schema.md",
  "04_pipeline_A_iphone_offline.md",
  "05_pipeline_B_relocalization.md",
  "06_pipeline_C_online_perception.md",
  "07_pipeline_D_consolidation.md",
  "08_runtime_timeline.md",
  "09_roadmap.md",
  "10_tech_stack.md",
  "11_world_model_bridge.md",
  "12_risks.md",
  "13_evaluation.md",
  "14_mvp.md",
  "15_glossary.md",
  "16_stats.md",
  "18_lyra_inspirations.md",
  "17_changelog.md",
]

# 章节显示名(替换原一级标题,使目录整齐)
chapter_titles = {
  "README.md":                          "项目总览 (README)",
  "00_overview.md":                     "第 0 章 总览与命名由来",
  "01_capability_decomposition.md":     "第 1 章 两方案能力解构",
  "02_architecture.md":                 "第 2 章 四层空间记忆架构",
  "03_data_schema.md":                  "第 3 章 统一数据模型",
  "04_pipeline_A_iphone_offline.md":    "第 4 章 管线 A:iPhone 离线建图",
  "05_pipeline_B_relocalization.md":    "第 5 章 管线 B:ZED 重定位握手",
  "06_pipeline_C_online_perception.md": "第 6 章 管线 C:在线感知与差异检测",
  "07_pipeline_D_consolidation.md":     "第 7 章 管线 D:记忆巩固",
  "08_runtime_timeline.md":             "第 8 章 端到端运行时序",
  "09_roadmap.md":                      "第 9 章 实施路线图",
  "10_tech_stack.md":                   "第 10 章 技术栈选型",
  "11_world_model_bridge.md":           "第 11 章 与世界模型衔接",
  "12_risks.md":                        "第 12 章 风险与对策",
  "13_evaluation.md":                   "第 13 章 评测指标",
  "14_mvp.md":                          "第 14 章 最小可复现 demo",
  "15_glossary.md":                     "第 15 章 缩略语表",
  "16_stats.md":                        "第 16 章 文档与代码统计",
  "18_lyra_inspirations.md":            "第 18 章 Lyra 2.0 启发的设计原则",
  "17_changelog.md":                    "第 17 章 变更日志",
}

# 常见 emoji / 装饰符号 → 文字或删除
EMOJI_MAP = {
    "🌈":"", "📱":"", "📷":"", "🤖":"", "🚀":"", "💻":"", "📊":"", "📋":"",
    "📚":"", "🎓":"", "🎬":"", "🏗️":"", "🏨":"", "💰":"", "📐":"", "🌟":"",
    "⚠️":"[!]", "✅":"[OK]", "❌":"[X]", "🟢":"[L]", "🟡":"[M]", "🔴":"[H]",
    "🎯":"", "📈":"", "📞":"", "📍":"", "🗺️":"", "📦":"", "🔧":"", "🤝":"",
    "📝":"", "🌐":"", "📄":"", "📧":"", "💡":"", "💬":"", "📤":"", "📥":"",
    "▶":">", "◀":"<", "⭐":"★",
    "↑":"↑", "↓":"↓",
    # 第二轮补充:xelatex_pass2.log 中报告的缺失字符
    "🔬":"", "🧑":"", "💼":"", "🎤":"", "📹":"",
    "→":"->", "↔":"<->",
    "①":"(1)", "②":"(2)", "③":"(3)", "④":"(4)",
    "✓":"[v]",
    "・":"·",
}

def preprocess(text: str, source_name: str) -> str:
    # 1) emoji → 替换
    for k, v in EMOJI_MAP.items():
        text = text.replace(k, v)
    # 2) 移除起首 yaml/front matter(不会有,跳过)
    # 3) 删除 raw HTML 注释 / details
    text = re.sub(r"<!--.*?-->", "", text, flags=re.S)
    text = re.sub(r"</?details[^>]*>", "", text)
    text = re.sub(r"</?summary[^>]*>", "", text)
    # 4) 替换文件名链接 [`xx.md`](path) → \emph{xx.md} 等;
    #    Pandoc 默认会把这些保留为超链接(指向不存在的文件),
    #    我们把 .md 链接收敛为只显示文本,避免大量"红色断链"。
    # 直接降级为反引号代码标记,让 pandoc 自己处理转义(它会输出 \texttt{}+下划线转义)
    def md_link_replace(m):
        token = m.group(1)
        # 保留为 inline-code: `xxx`
        return f"`{token}`"
    text = re.sub(r"\[`([^`\]]+)`\]\([^)]+\)", md_link_replace, text)
    # 普通 [text](xxx.md...) 不带反引号的:仅保留文本(纯文本无需转义)
    text = re.sub(r"\[([^\]]+?)\]\([^)]*\.md[^)]*\)",
                  lambda m: m.group(1), text)
    # 5) 把第一行 H1 替换为指定章节标题
    title = chapter_titles.get(source_name, source_name)
    lines = text.splitlines()
    for i, l in enumerate(lines):
        if l.startswith("# "):
            lines[i] = f"# {title}"
            break
    else:
        lines.insert(0, f"# {title}")
    # 6) 降级:把 ## 起的所有标题保留;一级 # 后面紧跟的 "## —— xxx" 类副标题
    #    保持原样,但避免 pandoc 把它认成另一个 chapter
    out = "\n".join(lines)
    # 7) 把 horizontal rule 上下的"章节版本/估计阅读时间/关键收获"等元信息
    #    保留(它们是有用尾页内容)
    return out

# 拼接
parts = []
for fn in order:
    p = PRISM_DIR / fn
    if not p.exists():
        print(f"[WARN] missing {p}")
        continue
    body = p.read_text(encoding="utf-8")
    body = preprocess(body, fn)
    # 每章前换页
    parts.append("\n\n\\newpage\n\n" + body + "\n")

OUT.write_text("\n".join(parts), encoding="utf-8")
print(f"[ok] merged → {OUT} ({OUT.stat().st_size/1024:.1f} KB)")
PYEOF

echo "==> 1b) 渲染 mermaid 图块 (mmdc → PNG)"
FIG_DIR="$BUILD_DIR/figs"
if [[ ! -x "$TOOLS_DIR/node_modules/.bin/mmdc" ]]; then
  echo "[ERROR] mmdc 未安装,请先执行:"
  echo "        cd $TOOLS_DIR && npm install @mermaid-js/mermaid-cli"
  exit 1
fi
python3 "$TOOLS_DIR/render_mermaid.py" \
  "$BUILD_DIR/merged.md" \
  "$BUILD_DIR/merged.md" \
  "$FIG_DIR"

echo "==> 2) pandoc → LaTeX body"
pandoc "$BUILD_DIR/merged.md" \
  --from=markdown+pipe_tables+grid_tables+raw_tex+task_lists+yaml_metadata_block \
  --to=latex \
  --top-level-division=chapter \
  --listings=false \
  --wrap=preserve \
  --highlight-style=tango \
  -o "$BUILD_DIR/body.tex"

# 把 body 注入模板,同时把 __PRISM_VERSION__ / __PRISM_BUILD_DATE__ 占位符替换为真实值;
# 并对 pandoc 生成的"自然宽度 longtable"做自适应换行后处理。
export PRISM_VERSION="$VERSION"
export PRISM_BUILD_DATE="$BUILD_DATE"
python3 - <<'PYEOF'
import os, pathlib, re
tpl  = pathlib.Path("plans/PRISM/tools/book_template.tex").read_text(encoding="utf-8")
body = pathlib.Path("plans/PRISM/.build/body.tex").read_text(encoding="utf-8")

version    = os.environ.get("PRISM_VERSION", "0.0.0")
build_date = os.environ.get("PRISM_BUILD_DATE", "unknown")
tpl = tpl.replace("__PRISM_VERSION__", version)
tpl = tpl.replace("__PRISM_BUILD_DATE__", build_date)

# pandoc 默认生成 \begin{longtable}[]{@{}lll@{}} 这种自然宽列规范,
# 长内容(URL/中文长句/代码)会冲出页面。把所有形如 @{}[lcr]+@{} 的
# 自然宽列规范替换为等宽 p{w} 列,启用单元格内换行。
LT_HEAD_RE = re.compile(
    r"\\begin\{longtable\}(\[[^\]]*\])?\{@\{\}([lcr]+)@\{\}\}"
)

def _replace_lt(m):
    opts = m.group(1) or ""
    cols = m.group(2)
    n    = len(cols)
    col_spec = " ".join(
        r">{\raggedright\arraybackslash}p{(\linewidth - "
        f"{2*n}" + r"\tabcolsep)/" + f"{n}" + r"}"
        for _ in range(n)
    )
    return f"\\begin{{longtable}}{opts}{{@{{}}{col_spec}@{{}}}}"

n_fixed = len(LT_HEAD_RE.findall(body))
body = LT_HEAD_RE.sub(_replace_lt, body)
print(f"[longtable] auto-wrapped {n_fixed} natural-width tables -> p columns")

out = tpl.replace("__PRISM_BODY_PLACEHOLDER__", body)
pathlib.Path("plans/PRISM/.build/book.tex").write_text(out, encoding="utf-8")
print("[ok] injected -> plans/PRISM/.build/book.tex")
PYEOF

echo "==> 3) XeLaTeX 编译 (两次,生成目录)"
cd "$BUILD_DIR"
# 用 -interaction=nonstopmode 跑两遍;第二遍解决 \tableofcontents 引用
for i in 1 2; do
  echo "    pass $i ..."
  xelatex -interaction=nonstopmode -halt-on-error -file-line-error \
          book.tex > "xelatex_pass${i}.log" 2>&1 || {
    echo "[ERROR] XeLaTeX pass $i failed; tail of log:"
    tail -60 "xelatex_pass${i}.log"
    exit 1
  }
done
cd - >/dev/null

cp "$BUILD_DIR/book.pdf" "$PRISM_DIR/PRISM_Book.pdf"
echo "[ok] body PDF → $PRISM_DIR/PRISM_Book.pdf ($(du -h "$PRISM_DIR/PRISM_Book.pdf" | cut -f1))"

echo "==> 4) 合并 cover + body → PRISM_Whole.pdf"
python3 - <<PYEOF
from pypdf import PdfWriter, PdfReader
import pathlib
prism = pathlib.Path("plans/PRISM")
cover = prism / "PRISM_Cover.pdf"
body  = prism / "PRISM_Book.pdf"
out   = prism / "PRISM_Whole.pdf"

w = PdfWriter()
if cover.exists():
    for p in PdfReader(str(cover)).pages:
        w.add_page(p)
    print(f"  + cover: {len(PdfReader(str(cover)).pages)} pages")
else:
    print(f"  [WARN] cover not found: {cover}")

# 从正文 PDF 继承 metadata(Title/Author/Subject/Keywords),
# 这样合并后的 Whole.pdf 也带 v__VERSION__ 标识。
body_reader = PdfReader(str(body))
for p in body_reader.pages:
    w.add_page(p)
print(f"  + body:  {len(body_reader.pages)} pages")

src_meta = body_reader.metadata or {}
w.add_metadata({
    "/Title":    src_meta.get("/Title",    "PRISM"),
    "/Author":   src_meta.get("/Author",   "Project Planning Team"),
    "/Subject":  src_meta.get("/Subject",  "PRISM"),
    "/Keywords": src_meta.get("/Keywords", "PRISM"),
    "/Creator":  src_meta.get("/Creator",  "LaTeX + pypdf"),
})

with open(out, "wb") as f:
    w.write(f)
print(f"[ok] merged → {out} ({out.stat().st_size/1024/1024:.2f} MB)")
print(f"     Title:   {src_meta.get('/Title', '<none>')}")
print(f"     Subject: {src_meta.get('/Subject', '<none>')}")
PYEOF

echo ""
echo "============================================================"
echo " ✅ PRISM 全书构建完成"
echo "------------------------------------------------------------"
echo "  封面:        $PRISM_DIR/PRISM_Cover.pdf"
echo "  正文:        $PRISM_DIR/PRISM_Book.pdf"
echo "  整合全本:    $PRISM_DIR/PRISM_Whole.pdf"
echo "============================================================"