From 260bdde2df8a87e77a77bd450fef9a2a0693d373 Mon Sep 17 00:00:00 2001 From: albert4719 Date: Sat, 19 Sep 2026 09:41:13 +0800 Subject: [PATCH] =?UTF-8?q?=E5=8A=A0=20PDF=20=E6=9E=84=E5=BB=BA=EF=BC=9Apa?= =?UTF-8?q?ndoc=20=E8=BD=AC=20typst=20=E6=8E=92=E7=89=88=EF=BC=8CA4=20?= =?UTF-8?q?=E6=8C=89=E4=B9=A6=E6=8E=92=EF=BC=8C=E7=9B=AE=E5=BD=95=E5=B8=A6?= =?UTF-8?q?=E9=A1=B5=E7=A0=81=E3=80=81=E4=B9=A6=E7=AD=BE=E5=88=B0=E6=9D=A1?= =?UTF-8?q?=E7=9B=AE=E7=BA=A7?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- tools/pdf/build.mjs | 97 ++++++++++++++++++++++++++++++++++++++++++ tools/pdf/template.typ | 74 ++++++++++++++++++++++++++++++++ 2 files changed, 171 insertions(+) create mode 100644 tools/pdf/build.mjs create mode 100644 tools/pdf/template.typ diff --git a/tools/pdf/build.mjs b/tools/pdf/build.mjs new file mode 100644 index 0000000..1e5f38f --- /dev/null +++ b/tools/pdf/build.mjs @@ -0,0 +1,97 @@ +// 把 README + book/*.md + docs/*.md 排成一本 PDF:pandoc 把 Markdown 转成 typst,typst 排版。 +// 用法:node tools/pdf/build.mjs [输出路径] 默认输出 dist/HowToLiveBetter.pdf +// 需要 pandoc(≥3.1,要有 typst 输出)和 typst(≥0.13)在 PATH 上,或用环境变量 PANDOC、TYPST 指路径。 +// 版面在 tools/pdf/template.typ 里;正文一个字都不改,只做三件事: +// 去掉「← 回总目录」、给每节的标题挂上锚点、把仓库内的链接改成书内跳转或 GitHub 网址。 +import { writeFileSync, mkdirSync, statSync } from 'node:fs'; +import { resolve, dirname, posix, basename } from 'node:path'; +import { execFileSync } from 'node:child_process'; +import { ROOT, REPO, SITE, TITLE, read, readBook, gitCommit, stripBackLink } from '../lib/book.mjs'; + +const OUT = resolve(ROOT, process.argv[2] ?? 'dist/HowToLiveBetter.pdf'); +const WORK = resolve(ROOT, 'dist/pdf-build.md'); +const PANDOC = process.env.PANDOC ?? 'pandoc'; +const TYPST = process.env.TYPST ?? 'typst'; +const DATE = new Date().toISOString().slice(0, 10); +const COMMIT = gitCommit(); + +const { description, frontMd, contentsMd, bookFiles, docFiles } = readBook(); + +// ---------- 页(每页一个一级标题,一级标题在 typst 里另起一页) ---------- +const anchorOf = new Map(); +bookFiles.forEach(f => anchorOf.set(f, 'sec-' + (basename(f).match(/^\d+/)?.[0] ?? anchorOf.size + 1))); +docFiles.forEach((f, i) => anchorOf.set(f, `doc-${i + 1}`)); + +const pages = [ + { src: 'README.md', md: `# 前言\n\n${description}\n\n${frontMd}`, anchor: 'front' }, + { src: 'README.md', md: contentsMd.replace(/^## 目录/, '# 各节简介'), anchor: 'contents' }, + ...[...bookFiles, ...docFiles].map(src => ({ src, md: stripBackLink(read(src)), anchor: anchorOf.get(src) })), + { src: 'README.md', md: aboutMd(), anchor: 'about' }, +]; + +function aboutMd() { + const commitLine = COMMIT ? `- 对应提交:${COMMIT.slice(0, 7)}\n` : ''; + return `# 版本说明 + +这本 PDF 由仓库里的 Markdown 正文自动排版,正文一改就重新排一本。手里这本的版本: + +- 生成日期:${DATE} +${commitLine}- 最新版下载、在线检索、提意见:${REPO} +- 在线检索页(按关键词、章节、证据等级和成本筛选,也能存成单文件离线看):${SITE} + +正文里指向书内其他节的链接已改成书内跳转;指向核实记录、许可证这类没排进书的文件的链接改成了 GitHub 网址。 + +全书以 Unlicense 发布,属于公有领域,可以随意复制、修改、分发。`; +} + +// ---------- 链接:书内的改成锚点,书外的改成绝对网址 ---------- +function rewriteLinks(md, src) { + return md.replace(/\]\(([^)\s]+)(\s+"[^"]*")?\)/g, (all, href, title) => { + if (/^(https?:|mailto:)/.test(href)) return all; + // README 里指向自身小节的锚点(#目录 这种)在书里不一定有,指回 GitHub 上的 README + if (href.startsWith('#')) return `](${REPO}/blob/main/README.md${href}${title ?? ''})`; + const [path] = href.split('#'); + const target = posix.normalize(posix.join(posix.dirname(src), path)); + const anchor = anchorOf.get(target); + if (anchor) return `](#${anchor}${title ?? ''})`; + const kind = target.endsWith('/') ? 'tree' : 'blob'; + return `](${REPO}/${kind}/main/${target}${title ?? ''})`; + }); +} + +const body = pages.map(p => { + const md = rewriteLinks(p.md, p.src) + .replace(//g, '') // 成本标签这类 HTML 注释不进 PDF + .replace(/^(# .+?)\s*$/m, `$1 {#${p.anchor}}`); // 给这一页的一级标题挂锚点 + if (!md.includes(`{#${p.anchor}}`)) throw new Error(`${p.src} 里没找到一级标题,挂不上锚点`); + return md.trim(); +}).join('\n\n'); + +mkdirSync(dirname(OUT), { recursive: true }); +writeFileSync(WORK, body); + +// ---------- pandoc → typst → pdf ---------- +const run = (cmd, args) => { + try { + return execFileSync(cmd, args, { cwd: ROOT, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'] }); + } catch (err) { + if (err.code === 'ENOENT') throw new Error(`找不到 ${cmd},装上它或用环境变量 ${cmd === PANDOC ? 'PANDOC' : 'TYPST'} 指到可执行文件`); + throw new Error(`${cmd} 失败:\n${err.stderr || err.stdout || err.message}`); + } +}; + +const typFile = resolve(ROOT, 'dist/pdf-build.typ'); +run(PANDOC, [ + '--from=gfm+attributes', '--to=typst', '--wrap=none', + `--template=${resolve(ROOT, 'tools/pdf/template.typ')}`, + '-V', `booktitle=${TITLE}`, '-V', `subtitle=${description}`, + '-V', `builddate=${DATE}`, '-V', `commit=${COMMIT.slice(0, 7) || '未知'}`, + '-V', `site=${SITE}`, '-V', `repo=${REPO}`, + '-o', typFile, WORK, +]); +const log = run(TYPST, ['compile', typFile, OUT, '--root', ROOT]); +if (log.trim()) console.log(log.trim()); + +const entries = pages.filter(p => bookFiles.includes(p.src)) + .reduce((n, p) => n + p.md.split('\n').filter(l => l.startsWith('### ')).length, 0); +console.log(`已生成 ${OUT}:${bookFiles.length} 节 ${entries} 条,附录 ${docFiles.length} 篇,${(statSync(OUT).size / 1048576).toFixed(1)} MB`); diff --git a/tools/pdf/template.typ b/tools/pdf/template.typ new file mode 100644 index 0000000..20993b2 --- /dev/null +++ b/tools/pdf/template.typ @@ -0,0 +1,74 @@ +$-- +$-- pandoc 的 typst 模板(只认 $body$ 和几个 -V 变量),不用 pandoc 自带的 conf(): +$-- 自带模板把页面设置锁在 conf() 里,改不了页眉页脚,所以这里自己排。 +$-- 开头到 divider 那段是 pandoc 生成的正文要用的辅助定义,照抄自 `pandoc -D typst`,别删。 +$-- +#set terms(hanging-indent: 1.5em) + +#set table(inset: 6pt, stroke: none) +// pandoc 把表格塞进 align(center) 里,单元格会跟着居中,中文表格左对齐才好读 +#show table.cell: it => align(left, it) + +#let horizontalRule = line(start: (25%, 0%), end: (75%, 0%)) +#let divider = if "divider" in std { divider } else { horizontalRule } + +#show figure.where(kind: table): set figure.caption(position: top) +#show figure.where(kind: image): set figure.caption(position: bottom) +// 长表格要能跨页,否则整块挤不下就留一页白 +#show figure: set block(breakable: true) +#set smartquote(enabled: false) + +// ---------- 版面 ---------- +#set document(title: "$booktitle$", author: "eternity4719") +#set text( + // 西文用 typst 自带的 Libertinus,中文按可用性往后找:CI 上是 Noto,本机是雅黑 + font: ("Libertinus Serif", "Noto Serif CJK SC", "Noto Serif SC", "Source Han Serif SC", "Noto Sans CJK SC", "Microsoft YaHei", "SimSun"), + size: 10.5pt, lang: "zh", region: "cn", +) +#set par(justify: false, leading: 0.78em, spacing: 0.9em) +#set list(indent: 0.6em, spacing: 0.75em) +#show raw: set text(font: ("DejaVu Sans Mono", "Noto Sans Mono CJK SC", "Consolas"), size: 9pt) +#show link: set text(fill: rgb("#1a4fb4")) +#show heading: set block(sticky: true, above: 1.5em, below: 0.65em) +#show heading.where(level: 1): set text(19pt) +#show heading.where(level: 2): set text(14pt) +#show heading.where(level: 3): set text(11.5pt) +// 每节另起一页;weak 保证前一页正好排满时不多出一张空页 +#show heading.where(level: 1): it => { pagebreak(weak: true); it } + +// 页眉:左边书名,右边当前节名;一节的头一页不打页眉 +#let running-head = context { + let next = query(selector(heading.where(level: 1)).after(here())).at(0, default: none) + if next != none and next.location().page() == here().page() { return } + let seen = query(selector(heading.where(level: 1)).before(here())) + if seen.len() == 0 { return } + set text(8.5pt, fill: luma(120)) + grid(columns: (1fr, auto), align(left)[$booktitle$], align(right)[#seen.last().body]) + v(-7pt) + line(length: 100%, stroke: 0.4pt + luma(215)) +} + +// ---------- 封面 ---------- +#set page(paper: "a4", margin: (x: 2.2cm, top: 2.2cm, bottom: 2cm), header: none, footer: none) +#align(center + horizon)[ + #image("/og.png", width: 100%) + #v(1.2cm) + #block(width: 80%)[#text(11.5pt, fill: luma(60))[$subtitle$]] + #v(2cm) + #text(10pt, fill: luma(90))[ + 生成于 $builddate$ · 正文提交 $commit$ \ + 正文每天都在改,以在线版为准:$site$ \ + 在线检索、EPUB 与本 PDF 的最新版都在 $repo$ + ] +] + +// ---------- 目录 ---------- +#pagebreak() +#outline(title: [目录], depth: 1, indent: 1em) + +// ---------- 正文 ---------- +#pagebreak(weak: true) +#set page(header: running-head, footer: context align(center, text(8.5pt, fill: luma(120))[#counter(page).at(here()).first() / #counter(page).final().first()])) +#counter(page).update(1) + +$body$