path: root/build.js

const fs = require('fs')
const path = require('path')
const yaml = require('js-yaml')
const marked = require('marked')
const fixWS = require('fix-whitespace')
const mjAPI = require('mathjax-node')
const cheerio = require('cheerio')

const { promisify } = require('util')

const ncp = promisify(require('ncp').ncp)
const writeFile = promisify(fs.writeFile)
const readFile = promisify(fs.readFile)
const readDir = promisify(fs.readdir)
const stat = promisify(fs.stat)

const mkdir = function(dir) {
  return promisify(fs.mkdir)(dir)
    .catch(error => {
      if (error.code === 'EEXIST') {
        return
      } else {
        throw error
      }
    })
}

const SITE_ORIGIN = (process.env.BLOG_ORIGIN || 'http://localhost:8000/')
const AUTHOR = (process.env.BLOG_AUTHOR || 'Florrie')

mjAPI.config({
  MathJax: {}
})

mjAPI.start()

const mathjaxTypeset = (math, format) => new Promise((resolve, reject) => {
  mjAPI.typeset({
    math, format,
    svg: true
  }, function(data) {
    if (data.errors) {
      reject(data.errors)
    } else {
      resolve(data.svg)
    }
  })
})

async function build() {
  console.log('Site origin:', getSiteOrigin())

  console.log('\x1b[1mSetting up things.\x1b[0m') // ----------------------------------------------

  process.stdout.write('Setting up directories...') // ............................................

  await mkdir('site')
  await Promise.all([
    mkdir('site/posts'),
    mkdir('site/archive'),
    mkdir('site/static').then(() => mkdir('site/static/math-svg'))
  ])

  console.log(' Created necessary folders.')

  process.stdout.write('Getting category data...') // .............................................

  const categoryData = await getCategoryData()

  console.log(` Gotten data for ${Object.keys(categoryData).length} categories.`)

  console.log('\x1b[1mBuilding posts.\x1b[0m') // -------------------------------------------------

  process.stdout.write('Getting post filenames...') // ............................................

  const postFiles = (await readDir('site/posts'))
    .filter(item => item.endsWith('.md'))

  console.log(` Found ${postFiles.length} markdown files.`)

  console.log('Parsing post texts...') // .........................................................

  let parsedPostCount = 0
  const posts = await Promise.all(postFiles.map(
    f => readFile('site/posts/' + f, 'utf-8').then(parsePostText)
      .then(post => {
        if (post.config.permalink) {
          console.error(`\n\x1b[1mWarning!\x1b[0m ${f} still has a manually-set permalink.`)
        }

        post.config.permalink = path.basename(f, '.md')
        parsedPostCount++
        process.stdout.write(`\r${Math.floor(parsedPostCount / postFiles.length * 100)}%`)
        return Object.assign(post, {sourceFile: f})
      })
  ))

  console.log('\rParsed.')

  /*
  process.stdout.write('Filtering just the updated posts...') // ..................................

  const updatedPostFiles = (await Promise.all(postFiles.map(f =>
    Promise.all([
      stat('site/posts/' + f).then(s => s.mtime),
      stat('site/posts/' + path.basename(f, '.md') + '.html').then(
        s => s.mtime,
        e => 0
      )
    ]).then(([ mtimeSrc, mtimeHTML ]) => {
      return mtimeSrc > mtimeHTML ? f : false
    })
  ))).filter(Boolean)

  const updatedPosts = posts.filter(p => updatedPostFiles.includes(p.sourceFile))

  console.log(` Found ${updatedPostFiles.length} .md files that have been modified more recently than their respective .html files.`)
  */

  // For now just rebuild EVERYTHING - only rebuilding updated posts is kind of not good when
  // you consider that like categories and "next post in category" and blah blah blah those
  // kinds of links and also like uhhhhhhhhhhhhhhhhhhhhhhhh blog origin changes it's all a
  // mess and I don't want to figure out how to deal with it without rebuilding everything lul.
  const updatedPosts = posts

  process.stdout.write('Writing post pages...') // ................................................
  await Promise.all(updatedPosts.map(post => {
    writeFile('site/' + getPostPath(post), generatePostPage(post, categoryData, posts))
  }))
  console.log(' Written.')

  process.stdout.write('Writing index (latest post) page...') // ..................................
  await writeFile('site/index.html', generatePostPage(getLatestPost(posts), categoryData, posts))
  await writeFile('site/index.md', await readFile('site/posts/' + getLatestPost(posts).sourceFile))
  console.log(' Written.')

  console.log('\x1b[1mWriting archive and category pages.\x1b[0m') // -----------------------------

  process.stdout.write('Writing archive index page...') // ........................................
  await writeFile('site/archive.html', generateArchiveCategoriesPage(categoryData))
  console.log(' Written.')

  process.stdout.write('Writing every-post archive page...') // ...................................
  await writeFile('site/archive/all.html', generateArchivePage(posts))
  console.log(' Written.')

  process.stdout.write('Writing each category\'s own page...') // .................................
  await writeCategoryPages(posts, categoryData)
  console.log(' Written.')

  console.log('\x1b[1mWriting other miscellaneous pages.\x1b[0m') // ------------------------------
  // TODO: Don't hard-code these, lol

  process.stdout.write('Writing "About" page...') // ..............................................
  await writeFile('site/about.html', generateAboutPage(await readFile('site/about.md', 'utf-8')))
  console.log(' Written.')
}

const generateStaticPage = (md, head = '') => (
  generateSitePage(head, marked(md))
)

const generateAdjacentPostLinks = (post, categoryData, allPosts) => {
  const postDate = getDate(post)

  const dateReduce = fn => (choice, cur) => {
    const curDate = getDate(cur)
    const choiceDate = choice ? getDate(choice) : 0
    if (fn(curDate, choiceDate)) {
      return cur
    } else {
      return choice
    }
  }

  const reduceNextByDate = dateReduce(
    (curDate, choiceDate) => curDate > postDate && (!choiceDate || curDate < choiceDate)
  )

  const reducePreviousByDate = dateReduce(
    (curDate, choiceDate) => curDate < postDate && (!choiceDate || curDate > choiceDate)
  )

  const nextByDate = allPosts.reduce(reduceNextByDate, null)
  const previousByDate = allPosts.reduce(reducePreviousByDate, null)

  const filterPostsWithSameCategory = cat => p => {
    return (p.config.categories || []).includes(cat)
  }

  const nextByCategories = (post.config.categories || []).map(cat => {
    return [cat,
      allPosts
        .filter(filterPostsWithSameCategory(cat))
        .reduce(reduceNextByDate, null)
    ]
  }).filter(([ cat, post ]) => !!post)
    .filter(([ cat, post ]) => post !== nextByDate)

  const previousByCategories = (post.config.categories || []).map(cat => {
    return [cat,
      allPosts
        .filter(filterPostsWithSameCategory(cat))
        .reduce(reducePreviousByDate, null)
    ]
  }).filter(([ cat, post ]) => !!post)
    .filter(([ cat, post ]) => post !== previousByDate)

  const makeAdjacentInCategoriesLinks = postsByCategories => {
    if (postsByCategories.length) {
      return fixWS`
        (by date; in ${postsByCategories
          .map(([ cat, post ]) => fixWS`
            <a href='${getPostPermalink(post)}'>${categoryData[cat].title}</a>
          `)
          .join(', ')
        })
      `
    } else {
      return ''
    }
  }

  // TODO: fix-whitespace hates this string, for some reason??
  return fixWS`
    ${previousByDate
      ? fixWS`
        <a href='${getPostPermalink(previousByDate)}'>Previous post</a>
        ${makeAdjacentInCategoriesLinks(previousByCategories)}
      `
      : ''
    }
    ${nextByDate ? '<br>' : ''}
    ${nextByDate
      ? fixWS`
        <a href='${getPostPermalink(nextByDate)}'>Next post</a>
        ${makeAdjacentInCategoriesLinks(nextByCategories)}
      `
      : ''
    }
  `
}

const generatePostPage = (post, categoryData, allPosts) => {
  const categories = post.config.categories

  const categoryLinks = (
    (categories && categories.length > 0)
    ? categories.map(id => fixWS`
      <a href='${getCategoryPath(id)}'>${categoryData[id].title}</a>
    `)
    : null
  )

  const categoryLinkText = (
    categoryLinks
    ? 'categories: ' + categoryLinks.join(', ')
    : 'no categories'
  )

  return generateSitePage(
    // No description-meta here.. yet. In theory there should be, but Google
    // does a pretty good job at guessing descriptions from the content of
    // the post.
    fixWS`
      <title>${post.config.title}</title>
      ${generateMetaHead({
        'Description': getOneSentence(getPostDescription(post)),
        'twitter:card': post.config.presentArt ? 'summary_image' : 'summary',
        'twitter:site': '@towerofnix',
        'twitter:title': post.config.title,
        'twitter:description': getPostDescription(post),
        'twitter:image': (
          post.config.thumbnail
          ? (getSiteOrigin() + post.config.thumbnail)
          : null
        ),
        'twitter:image:alt': null // </3 TODO: fix this?
      })}
    `,

    fixWS`
      ${post.html}
      <p class='post-meta'>
        (-${AUTHOR},
          <a href='${getPostPermalink(post)}'>${getTimeElement(post)}</a>
          (<a href='${getPostPermalink(post, 'md')}'>markdown</a>);
          ${categoryLinkText})
      </p>
      <p class='post-meta'>
        ${generateAdjacentPostLinks(post, categoryData, allPosts)}
      </p>
    `,

    fixWS`
      <div class='post-nav'>
        ${generateAdjacentPostLinks(post, categoryData, allPosts)}
      </div>
    `
  )
}

const generateAboutPage = md => {
  const description = (
    "The (slightly) extended description of towerofnix's blog."
  )

  return generateStaticPage(
    md,
    fixWS`
      <title>Blog</title>
      ${generateMetaHead({
        'Description': description,
        'twitter:card': 'summary',
        'twitter:site': '@towerofnix',
        'twitter:title': "About",
        'twitter:description': description,
      })}
    `
  )
}

const generateSitePage = (head, body, extraNav = '') => {
  if (!head.includes('twitter')) {
    // TODO: It would be really nice to be able to log the title of the page
    // here! - But virtually impossible without being smart and parsing the
    // head text.
    console.warn('No twitter meta tags..?\n' + head + '\n\n')
  }

  if (!head.includes('Description')) {
    console.warn('No Description meta tag..?\n' + head + '\n\n')
  }

  return fixWS`
    <!DOCTYPE html>
    <html>
      <head>
        <base href='${getSiteOrigin()}'>
        <meta charset='utf-8'>
        ${head}

        <link rel='stylesheet' href='static/site.css'>
      </head>
      <body>
        <div id='main' class='top-block'>
          <div id='nav'>
            <a href='index.html'>(Front.)</a>
            <a href='about.html'>(About!)</a>
            <a href='archive.html'>(Archive.)</a>${
              // TODO: fix-whitespace doesn't like whitespace-only lines??
              extraNav
            }
          </div>
          <div id='content'>
            ${body}
          </div>
        </div>
        <footer class='top-block'>
          <p>My posts and art are licensed under
             <a href='https://creativecommons.org/licenses/by-sa/4.0/'>
             CC BY-SA 4.0</a>, and this blog's source code is available
             <a href='https://git.ed1.club/florrie/blog.git'>online</a>!</p>
        </footer>
      </body>
    </html>
  `
}

const generateMetaHead = dict => {
  return Object.entries(dict)
    .filter(([ k, v ]) => {
      if (!v) {
        // It's normal to have a null value (that means "there's nothing in
        // this field"), but for there to be an undefined value (or
        // otherwise)..? That's strange!
        if (v !== null) {
          console.warn('Unset ' + k + '?', dict)
        }

        return false
      }

      return true
    })

    // fix-whitespace is broken??????? yikes.jpg
    // .map(([ k, v ]) => fixWS`
    //   <meta name='${k}' content="${
    //     v.replace(/"/g, '&quot;').replace(/\n/g, '__-\n-__')
    //   }">
    // `)

    .map(([ k, v ]) => (
      `<meta name='${k}' content="${v.replace(/"/g, '&quot;')}">`
    ))
    .join('\n')
}

const generateArchivePage = posts => (
  generateArchiveCategoryPage(
    '', // just use the default title, Archive (not "Archive - Archive"!)
    fixWS`
      A quick (read: long) table of all of the posts I've published here.
      See also the <a href='archive.html'>category-based archive</a>.
    `,
    posts
  )
)

const generateArchiveCategoriesPage = (categoryData) => {
  const description = (
    "A list of categories the posts on the blog are organized into."
  )

  return generateSitePage(
    fixWS`
      <title>Archive</title>
      ${generateMetaHead({
        'Description': description,
        'twitter:card': 'summary',
        'twitter:site': '@towerofnix',
        'twitter:title': 'Archive',
        'twitter:description': description
      })}
    `,

    fixWS`
      <h1>Archive</h1>
      <p>A list of the categories the posts on the blog are organized
      into. See also the <a href='archive/all.html'>archive of all
      posts</a>.</p>

      ${generateCategoryList(categoryData)}
    `
  )
}

const writeCategoryPages = (posts, categoryData) => (
  Object.entries(categoryData).map(
    ([id, cat]) => writeFile(
      'site/' + getCategoryPath(id),
      generateArchiveCategoryPage(
        cat.title, cat.description,
        posts.filter(post => post.config.categories.includes(id))
      )
    )
  )
)

const generateArchiveCategoryPage = (title, description, posts) => {
  const fullTitle = 'Archive' + (title ? ` - ${title}` : '')

  return generateSitePage(
    fixWS`
      <title>${fullTitle}</title>
      ${generateMetaHead({
        'Description': getOneSentence(getHTMLDescription(description)),
        'twitter:card': 'summary',
        'twitter:site': '@towerofnix',
        'twitter:title': fullTitle,
        'twitter:description': getHTMLDescription(description)
      })}
    `,

    fixWS`
      <h1>${title || 'Archive'}</h1>
      <p>${description}</p>

      ${generateArchiveTable(posts.sort((a, b) => getDate(a) - getDate(b)))}
    `
  )
}

const generateArchiveTable = posts => (
  fixWS`
    <table>
      <colgroup>
        <col>
        <col class="date-col">
      </colgroup>
      <tbody>
        ${
          posts.filter(post => !post.config.unlisted).map(
            post => {
              const link = getPostPath(post)

              return fixWS`
                <tr>
                  <td><a href='${link}'>${post.config.title}</a></td>
                  <td>${getTimeElement(post)}</td>
                </tr>
              `
            }
          ).join('\n')
        }
      </tbody>
    </table>
  `
)

const generateCategoryList = (categoryData) => (
  fixWS`
    <ul>
      ${
        Object.entries(categoryData).map(
        ([id, category]) => fixWS`
          <li>
            <a href='archive/${id}.html'>${category.title}:</a>
            ${category.description}
          </li>
        `
        ).join('\n')
      }
    </ul>
  `
)

const parsePostText = async (text) => {
  // Parses the text contents of a post .md file.
  //
  // YAML configuration is stored before a '---' separator. Markdown content
  // comes after the separator.
  //
  // It's recommended to indent the configuration code before '---' as a code
  // block, so that it's valid Markdown code.

  const separator = '\n---\n'

  const separatorIndex = text.indexOf(separator)
  const code = text.slice(0, separatorIndex)
  const markdown = text.slice(separatorIndex + separator.length).trim()

  let config

  try {
    config = yaml.safeLoad(code)
  } catch (err) {
    throw new Error('Invalid YAML: ' + err.message)
  }

  // We SHOULD use a custom Marked renderer, but MathJax runs asynchronously,
  // while marked expects a synchronous value. To be fair you could hack around
  // that (e.g. by storing a key to each math element, and then rendering those
  // after running marked), but that's a little yucky, and we can get inline
  // math if we deal with MathJax and Marked separately.

  let processedMarkdown = markdown

  let processedMath = []

  const handleMatch = (texType, applyTemplateFunc) => match => {
    const reference = match[1]
    const mathText = match[2]

    const directoryName = `static/math-svg/${config.permalink}`
    const svgFile = `${directoryName}/${reference}.svg`

    processedMath.push({reference, mathText})

    return mkdir('site/' + directoryName)
      .catch(err => {
        if (err.code !== 'EEXIST') {
          throw err
        }
      })
      .then(() => mathjaxTypeset(mathText, texType))
      // TODO: Sanitize this.. I'm going to inevitably write a slash in a
      // math ID, at some point!
      .then(math => writeFile('site/' + svgFile, math))
      .then(math => applyTemplateFunc(reference, svgFile))
  }

  processedMarkdown = await processMarkdown(processedMarkdown,
    /<pre class='math' id='([^']*)'>([\s\S]+?)<\/pre>/g,
    handleMatch('TeX', (reference, svgFile) => fixWS`
      <p class='math' id='${reference}'>
        <img src='${svgFile}'>
      </p>
    `)
  )

  processedMarkdown = await processMarkdown(processedMarkdown,
    /<code class='math' id='([^']*)'>([\s\S]+?)<\/code>/g,

    handleMatch('inline-TeX', (reference, svgFile) => fixWS`
      <span class='inline-math' id='${reference}'><img src='${svgFile}'></span>
    `)
  )

  processedMarkdown = await processMarkdown(processedMarkdown,
    /<art([^>]*)>([^<]*)<\/art>/g,
    ([ _, attrText = '', filename ]) => {
      const attrs = attrText.split(' ')
        .filter(Boolean)
        .map(entry => entry.includes('=') ? entry.split('=') : [entry, entry])
        .map(([ k, v ]) => [k, v.startsWith('"') && v.endsWith('"') ? v.slice(1, -1) : v])
        .map(([ k, v ]) => [k, v === 'true' ? true : v === 'false' ? false : v === k ? true : v])
        .reduce((acc, [ k, v ]) => Object.assign({}, acc, {[k]: v}), {})

      const options = Object.assign({
        ext: 'kra',
        noext: false
      }, attrs)

      return fixWS`
        <div class="art-container${options.fullwidth ? ' full-width' : ''}"><div>
          <a href="static/media/${encodeURIComponent(filename)}.png"><img alt="${filename.match(/^[0-9a-z]+-(.*)$/)[1]}" src="static/media/${encodeURIComponent(filename)}.png"></a>
          ${options.noext ? '' : fixWS`
            <br>(<a href="static/media/${encodeURIComponent(filename)}.${options.ext}">${filename.match(/^[0-9a-z]+-(.*)$/)[1]}.${options.ext}</a>)
          `}
        </div></div>
      `
    }
  )

  return {
    config: yaml.safeLoad(code),
    html: marked(processedMarkdown),
    markdown: markdown // Explicit is usually better than implicit
    // (The meaning of the above comment is implied)
  }
}

const getPostPath = (post, extension = 'html') => (
  `posts/${post.config.permalink}.${extension}`
)

const getPostPermalink = (post, extension = undefined) => (
  getSiteOrigin() + getPostPath(post, extension)
)

const getPostDescription = post => {
  if (post.config.description) {
    return post.config.description
  }

  return getHTMLDescription(post.html)
}

const getHTMLDescription = html => {
  const $ = cheerio.load(html)

  // Some paragraphs can only include images; we don't want to use those for
  // getting the description text.
  for (let p of $('p').get()) {
    const text = $(p).text()

    if (text) {
      return text
    }
  }

  return $.root().text()
}

const getOneSentence = text => {
  let i = 0

  while (i < text.length && text[i - 1] !== '.') {
    i++
  }

  const textRange = text.slice(0, i)
  const sentence = textRange.replace(/\n/g, ' ')

  return sentence
}

const getCategoryPath = id => (
  `archive/${id}.html`
)

const getCategoryData = () => (
  readFile('categories.yaml', 'utf-8')
    .then(text => yaml.safeLoad(text))
    .then(
      yamlObj => {
        const ret = {}

        for (let entry of yamlObj.categories) {
          ret[entry.id] = entry
        }

        return ret
      },
      err => {
        throw new Error("YAML Error: " + err.message, err)
      }
    )
)

const getLatestPost = posts => {
  let latestPost = posts.find(post => !post.config.unlisted)
  let latestDate = 0

  for (let curPost of posts) {
    // If it's unlisted, skip it.
    if (curPost.config.unlisted) {
      continue
    }

    const curDate = getDate(curPost)
    if (curDate > latestDate) {
      latestPost = curPost
      latestDate = curDate
    }
  }

  return latestPost
}

const getDate = post => {
  if (!('date' in post.config)) {
    console.warn(`Post ${post.config.title} has no date!`)
    return null
  }

  const { y, m, d } = post.config.date

  return new Date(y, m - 1, d)
}

const getTimeElement = post => {
  const date = getDate(post)

  const datetime = !date ? '' : asDatetime(date)

  const humanDatetime = (
    !date ? '(No date)' : asHumanReadableDate(date)
  )

  return `<time datetime='${datetime}'>${humanDatetime}</time>`
}

const asDatetime = date => (
  `${date.getFullYear()}-${date.getMonth() + 1}-${date.getDate()}`
)

const asHumanReadableDate = date => {
  const months = [
    'January', 'February', 'March', 'April', 'May', 'June', 'July', 'August',
    'September', 'October', 'November', 'December'
  ]

  const monthName = months[date.getMonth()]

  return `${monthName} ${date.getDate()}, ${date.getFullYear()}`
}

const getSiteOrigin = () => SITE_ORIGIN

const processMarkdown = async (markdown, regex, matchFunction) => {
  let processedMarkdown = ''
  let lastIndex = 0

  while (true) {
    const match = regex.exec(markdown)

    if (match) {
      processedMarkdown += markdown.slice(lastIndex, match.index)
      processedMarkdown += await matchFunction(match)
      lastIndex = regex.lastIndex
    } else {
      break
    }
  }

  processedMarkdown += markdown.slice(lastIndex)

  return processedMarkdown
}

if (require.main === module) {
  build()
    .catch(error => console.error(
      `${error.name} - ${error.reason}\n${error.stack}`
    ))
}