【XPath】ページの指定箇所の情報を取得する方法【GASスクレイピング】

function getTag(str, offset)
{
  var res = {};

  res.begin = str.indexOf('<', offset);
  res.end = str.indexOf('>', res.begin) + 1;
  res.tag = str.slice(res.begin, res.end);
  res.name = res.tag.replace('/','').replace('<','').replace('>','').split(' ')[0];
  res.isClose = res.tag.indexOf('</') >= 0;
  res.isComment = res.tag.indexOf('<!') >= 0;
  res.isSingle = res.tag.indexOf('/>') >= 0;

  return res;
}

function searchEnd(content, begin)
{
  var end = begin;
  var list = [];
  var offset = begin;
  while (true)
  {
    var res = getTag(content, offset);
    if (res.begin < 0) break;
    offset = res.end;

    if (res.isSingle) continue;
    if (res.isComment) continue;

    if (res.isClose)
    {
      while (list.length > 0)
      {
        var name = list.shift();
        if (name == res.name) break;
      }
    }
    else
    {
      list.unshift(res.name);
    }

    if (list.length <= 0)
    {
      end = offset;
      break;
    }
  }
  return end;
}

function removeParentTag(content)
{
  var begin = content.indexOf('>') + 1;
  var end = content.lastIndexOf('<');
  return content.slice(begin, end);
}

function searchId(content, hierarchy)
{
  var id = hierarchy.replace('*[@','').replace(']','');
  var index = content.indexOf(id);
  return content.lastIndexOf('<', index);
}

function searchTag(content, hierarchy)
{
  var array = hierarchy.replace('[', ',').replace(']', '').split(',');

  var tag = array[0];
  var count = 1;
  if (array.length > 1)
  {
    count = parseInt(array[1]);
  }

  var begin = 0;
  var list = [];
  var offset = 0;
  while (true)
  {
    var res = getTag(content, offset);
    if (res.begin < 0) break;
    offset = res.end;

    if (res.isSingle) continue;
    if (res.isComment) continue;

    if (res.isClose)
    {
      while (list.length > 0)
      {
        var name = list.shift();
        if (name == res.name) break;
      }
    }
    else
    {
      if (list.length <= 0)
      {
        if (tag == res.name)
        {
          --count;
        }
      }
      list.unshift(res.name);
    }

    if (count <= 0)
    {
      begin = res.begin;
      break;
    }
  }
  return begin;
}

function searchBegin(content, hierarchy)
{
  var index = hierarchy.indexOf('@');
  if (index < 0)
  {
    return searchTag(content, hierarchy);
  }
  return searchId(content, hierarchy);
}

function extractText(content)
{
  var offset = 0;
  var res = getTag(content, offset);
  if (res.begin < 0)
  {
    return content;
  }
  var begin = searchTag(content, res.name);
  var end = searchEnd(content, begin);
  var remove = content.slice(begin, end);
  return content.replace(remove, '');
}

function narrow(content, hierarchy)
{
  if (hierarchy.length <= 0) return content;

  if (hierarchy.indexOf('text()') >= 0)
  {
    return extractText(content);
  }

  var begin = searchBegin(content, hierarchy);
  var end = searchEnd(content, begin);
  var target = content.slice(begin, end);
  return removeParentTag(target);
}

function searchXPath(content, xpath)
{
  var result = [];
  for (var i in xpath)
  {
    var target = content;
    var array = xpath[i].split('/');
    for (var j in array)
    {
      if (array[j].length <= 0) continue;
      target = narrow(target, array[j]);
    }
    result.push(target);
  }
  return result;
}

function searchXPathFromURL(url, xpath)
{
  var content = UrlFetchApp.fetch(url).getContentText('UTF-8');
  return searchXPath(content, xpath);
}

function Test_01()
{
  var code = XXXX;
  var url = 'https://XXXX/stock/?code=' + code;

  var market = '//*[@id="stockinfo_i1"]/div[1]/span';                       // market
  var name = '//*[@id="stockinfo_i1"]/div[1]/h2/text()';                    // name
  var industry = '//*[@id="stockinfo_i2"]/div/a';                           // industry
  var price = '//*[@id="stockinfo_i1"]/div[2]/span[2]';                     // price
  var eps = '//*[@id="kobetsu_right"]/div[3]/table/tbody/tr[3]/td[4]';      // eps
  var dividend = '//*[@id="kobetsu_right"]/div[3]/table/tbody/tr[3]/td[5]'; // dividend
  var yield = '//*[@id="stockinfo_i3"]/table/tbody/tr[1]/td[3]/text()';     // yield
  var xpath = [market, name, industry, price, eps, dividend, yield];

  var content = UrlFetchApp.fetch(url).getContentText('UTF-8');
  var result = searchXPath(content, xpath);

  for (var i in xpath)
  {
    console.log(xpath[i] + ' > ' + result[i]);
  }
}