import JSZip from 'jszip'

const resultTablePattern = /<w:tbl(?:\s[^>]*)?>[\s\S]*?<\/w:tbl>/g
const floatingPositionPattern = /<w:tblpPr\b[^>]*\/>/
const tableIndentPattern = /<w:tblInd\b[^>]*\/>/
const runFontsPattern = /<w:rFonts\b[^>]*\/>/g

function alignReportHeader(xml, reportCode) {
  // 旧检验报告把页码放在浮动文本框中；SuperDoc 重新分页时会丢失 Word 的横向锚点。
  // 保留报告编号的所有 Word runs（模板填充值可能被拆到下一 run），只重建页码段落。
  if (!xml.includes('报告编号') || !/\bPAGE\b/.test(xml) || !/\bNUMPAGES\b/.test(xml)) return null
  const wordNamespace = 'http://schemas.openxmlformats.org/wordprocessingml/2006/main'
  const document = new DOMParser().parseFromString(xml, 'application/xml')
  const header = document.documentElement
  if (header.localName !== 'hdr' || document.getElementsByTagName('parsererror').length) return null
  const children = node => Array.from(node.childNodes).filter(child => child.nodeType === 1)
  const paragraphs = children(header).filter(node => node.localName === 'p')
  const reportParagraph = paragraphs.find(paragraph =>
    children(paragraph).some(run => run.localName === 'r' && run.textContent.includes('报告编号')))
  if (!reportParagraph) return null
  const reportRuns = children(reportParagraph).filter(node => node.localName === 'r')
  const labelIndex = reportRuns.findIndex(run => run.textContent.includes('报告编号'))
  if (labelIndex < 0) return null
  const reportText = reportRuns.slice(labelIndex).map(run => run.textContent).join('')
  const missingCode = !reportText.replace(/^.*?报告编号\s*[:：]\s*/, '').trim() && reportCode?.trim()

  const alreadyAligned = paragraphs.length === 2 && paragraphs[0] === reportParagraph
    && !xml.includes('w:txbxContent')
    && paragraphs[1].textContent.includes('第')
  if (alreadyAligned && !missingCode) return null

  const element = name => document.createElementNS(wordNamespace, `w:${name}`)
  const appendRun = (paragraph, name, value, preserveSpace = false) => {
    const run = element('r')
    const text = element(name)
    if (preserveSpace) text.setAttribute('xml:space', 'preserve')
    text.textContent = value
    run.appendChild(text)
    paragraph.appendChild(run)
  }
  const appendField = (paragraph, name, fallback) => {
    for (const kind of ['begin', 'separate', 'end']) {
      const run = element('r')
      const marker = element('fldChar')
      marker.setAttributeNS(wordNamespace, 'w:fldCharType', kind)
      run.appendChild(marker)
      paragraph.appendChild(run)
      if (kind === 'begin') appendRun(paragraph, 'instrText', ` ${name} `, true)
      if (kind === 'separate') appendRun(paragraph, 't', fallback)
    }
  }
  const alignedReport = element('p')
  const reportProperties = element('pPr')
  const reportJustification = element('jc')
  reportJustification.setAttributeNS(wordNamespace, 'w:val', 'right')
  reportProperties.appendChild(reportJustification)
  alignedReport.appendChild(reportProperties)
  for (const run of reportRuns.slice(labelIndex)) alignedReport.appendChild(run.cloneNode(true))
  if (missingCode) appendRun(alignedReport, 't', ` ${reportCode.trim()}`, true)

  const pageParagraph = element('p')
  const pageProperties = reportProperties.cloneNode(true)
  const indent = element('ind')
  indent.setAttributeNS(wordNamespace, 'w:right', '300')
  pageProperties.appendChild(indent)
  pageParagraph.appendChild(pageProperties)
  appendRun(pageParagraph, 't', '第 ', true)
  appendField(pageParagraph, 'PAGE', '1')
  appendRun(pageParagraph, 't', ' 页 共 ', true)
  appendField(pageParagraph, 'NUMPAGES', '4')
  appendRun(pageParagraph, 't', ' 页', true)

  while (header.firstChild) header.removeChild(header.firstChild)
  header.appendChild(alignedReport)
  header.appendChild(pageParagraph)
  return new XMLSerializer().serializeToString(document)
}

function chineseThemeFont(themeXml, kind) {
  const fontGroup = themeXml.match(new RegExp(`<a:${kind}Font\\b[^>]*>[\\s\\S]*?<\\/a:${kind}Font>`))?.[0]
  if (!fontGroup) return null
  // Word 按中文语言区域选 Hans；主题中的 a:ea 常为空，不能把它当成实际字体。
  const hansFont = [...fontGroup.matchAll(/<a:font\b[^>]*\/>/g)]
    .find(([tag]) => /\bscript="Hans"/.test(tag))?.[0]
  return hansFont?.match(/\btypeface="([^"]+)"/)?.[1]
    || fontGroup.match(/<a:ea\b[^>]*\btypeface="([^"]+)"/)?.[1]
    || null
}

function materializeChineseThemeFonts(xml, themeFonts) {
  let changed = false
  const normalized = xml.replace(runFontsPattern, tag => {
    let result = tag
    for (const script of ['ascii', 'hAnsi', 'eastAsia', 'cs']) {
      const themeAttribute = new RegExp(`\\s+w:${script}Theme="(majorEastAsia|minorEastAsia)"`)
      const theme = result.match(themeAttribute)?.[1]
      const font = theme === 'majorEastAsia' ? themeFonts.major : themeFonts.minor
      if (!font || !theme) continue
      // OOXML 中主题字体优先于同脚本的显式字体；解析成 Word 实际使用的字体。
      result = result.replace(themeAttribute, '')
      const directAttribute = new RegExp(`\\s+w:${script}="[^"]*"`)
      result = directAttribute.test(result)
        ? result.replace(directAttribute, ` w:${script}="${font}"`)
        : result.replace(/\s*\/>$/, ` w:${script}="${font}"/>`)
      changed = true
    }
    return result
  })
  return changed ? normalized : null
}

function alignResultTable(xml) {
  let changed = false
  const documentXml = xml.replace(resultTablePattern, (table, offset) => {
    if (!table.includes('标准要求值') || !table.includes('检测结果')) return table
    const position = table.match(floatingPositionPattern)?.[0]
    if (!position?.includes('w:horzAnchor="page"')) return table

    // 旧版结果表使用相对页面的固定坐标；沿用前一张正文表的缩进才能保持同一左边线。
    const precedingTables = [...xml.slice(0, offset).matchAll(resultTablePattern)]
    const precedingIndent = precedingTables.reverse()
      .map(match => match[0].match(tableIndentPattern)?.[0]?.match(/w:w="(-?\d+)"/)?.[1])
      .find(value => value != null)
    const indent = precedingIndent ?? '-670'
    changed = true
    return table.replace(position, '').replace(tableIndentPattern,
      `<w:tblInd w:w="${indent}" w:type="dxa"/>`)
  })
  return changed ? documentXml : null
}

export async function normalizeReportDocxForEditor(blob, reportCode) {
  const zip = await JSZip.loadAsync(blob)
  const documentPart = zip.file('word/document.xml')
  if (!documentPart) return blob
  const documentXml = await documentPart.async('string')
  const themePart = zip.file('word/theme/theme1.xml')
  const themeXml = themePart ? await themePart.async('string') : ''
  const themeFonts = { major: chineseThemeFont(themeXml, 'major'), minor: chineseThemeFont(themeXml, 'minor') }
  let normalizedXml = alignResultTable(documentXml) || documentXml
  normalizedXml = materializeChineseThemeFonts(normalizedXml, themeFonts) || normalizedXml
  let changed = normalizedXml !== documentXml
  if (changed) zip.file('word/document.xml', normalizedXml)
  const stylesPart = zip.file('word/styles.xml')
  if (stylesPart) {
    const stylesXml = await stylesPart.async('string')
    const normalizedStyles = materializeChineseThemeFonts(stylesXml, themeFonts)
    if (normalizedStyles) {
      zip.file('word/styles.xml', normalizedStyles)
      changed = true
    }
  }
  const headerNames = Object.keys(zip.files).filter(name => /^word\/header\d+\.xml$/.test(name))
  for (const headerName of headerNames) {
    const headerXml = await zip.file(headerName).async('string')
    const normalizedHeader = alignReportHeader(headerXml, reportCode)
    if (normalizedHeader) {
      zip.file(headerName, normalizedHeader)
      changed = true
    }
  }
  if (!changed) return blob
  return zip.generateAsync({ type: 'blob', compression: 'DEFLATE' })
}