fixing gender parsing

koronavirus.gov.hu site data extractor

class koronaVirusGovHuParser
{
  constructor() {
    this.cellSeparator = "\t";
    this.newLine = "\n";
    this.deathCasePageSelector = '.view-elhunytak .views-table > tbody:nth-child(2)';

    this.knownCases = new Map();
    this.__initTextOutput();
  }

  __initTextOutput()
  {
    this.textOutput = '';
  }

  __getFormattedDateTime(with_time = false, query_date_time = new Date())
  {
    var query_date_time_str =
      query_date_time.getFullYear() + '-' +
      (query_date_time.getMonth() + 1 < 10 ? '0' : '') + (query_date_time.getMonth()+1) + '-' +
      (query_date_time.getDate() < 10 ? '0' : '') + query_date_time.getDate();

    if (with_time === true)
    {
      query_date_time_str +=
        ' ' + (query_date_time.getHours() < 10 ? '0' : '') + query_date_time.getHours() +
        ':' + (query_date_time.getMinutes() < 10 ? '0' : '') + query_date_time.getMinutes();
    }

    return query_date_time_str;
  }

  __normalizeGender(gender, id)
  {
    if (gender == 'Férfi' || gender == 'férfi') return 'man';
    if (gender == 'Nő' || gender == 'nő' || gender == 'Nõ' || gender == 'nõ') return 'woman';

    console.error('unknown gender "' + gender + '" @' + id);
    return '???';
  }

  __normalizeDiseases(diseases_list)
  {
    return '';
  }

  __insertToTextOutput(row)
  {
    this.textOutput = row + this.newLine + this.textOutput;
  }

  __addToTextOutput(row)
  {
    this.textOutput += row + this.newLine;
  }

  __parseDeathCasePage(last_known_case_id)
  {
    this.__initTextOutput();

    var last_added_id = 0;
    var data_row;

    for (data_row of document.querySelector(this.deathCasePageSelector).rows)
    {
      var id = parseInt(data_row.children.item(0).innerText);

      if (id <= last_known_case_id) break;

      var case_data = {
        'id' : id,
        'gender' : this.__normalizeGender(data_row.children.item(1).innerText.trim(), id),
        'age' : data_row.children.item(2).innerText.trim(),
        'original_diseases' : data_row.children.item(3).innerText.trim(),
        'normalized_diseases' : this.__normalizeDiseases(data_row.children.item(3).innerText.trim()),
      }

      this.__insertToTextOutput(
        case_data.id + this.cellSeparator +
        this.__getFormattedDateTime() + this.cellSeparator +
        case_data.gender + this.cellSeparator +
        case_data.age + this.cellSeparator +
        '"' + case_data.original_diseases + '"'
        // + this.cellSeparator
        // + case_data.normalized_diseases + '"'
      );

      this.knownCases[id] = case_data;
      last_added_id = id;
    }

    return last_added_id ; //last parsed id
  }

  __parseMainPage()
  {
    this.__initTextOutput();
    var selectors =
    {
      '#content-fertozott-pest' : 'fertozott bp',
      '#content-fertozott-videk' : 'fertozott videk',
      '#content-gyogyult-pest' : 'gyogyult bp',
      '#content-gyogyult-videk' : 'gyogyult videk',
      '#content-elhunyt-pest' : 'elhunyt bp',
      '#content-elhunyt-videk' : 'elhunyt videk',
      '#content-karantenban' : 'karanten',
      '#content-mintavetel' : 'elvegzett teszt',
      "#content-beoltottak" : "oltottak szama",
    };

    this.__addToTextOutput('query time' + this.cellSeparator + this.__getFormattedDateTime(true));
    this.__addToTextOutput('last update' + this.cellSeparator + document.querySelector('#block-block-1 div p').innerHTML.substring(28).replace(/\./g,'-').replace(/\- /,' '));

    for (var key in selectors)
    {
      var label = selectors[key];
      var value = document.querySelector(key).innerHTML.replace(/\s/g, '');
      this.__addToTextOutput(label + this.cellSeparator + value);
    }

    for (var slided_content of document.querySelector("#flexslider-1 ul").children)
    {
      var slided_content_text = slided_content.innerText;
      var found = slided_content_text.match(/\s*(([0-9]+\s)*[0-9]+) koronavírusos beteget ápolnak kórházban, közülük (([0-9]+\s)*[0-9]+)-[a|e]n vannak lélegeztetőgépen\./i); //https://regexr.com/
      console.log(found);
      if (found != null)
      {
        this.__addToTextOutput('korhazban' + this.cellSeparator + found[1].replace(/\s/g, ''));
        this.__addToTextOutput('lelegezteton' + this.cellSeparator + found[3].replace(/\s/g, ''));
      }
    }
  }

  getTextOutput()
  {
    return this.textOutput;
  }

  getInfoByPage(last_known_case_id = 1)
  {
    var url = new URL(window.location);

    if (url.pathname == '/')
    {
      this.__parseMainPage();
    }
    else if (url.pathname == '/elhunytak')
    {
      var last_added_id = this.__parseDeathCasePage(last_known_case_id);
      console.info('last_added_id:', last_added_id);
    }

    return this.getTextOutput();
  }

}

var kvghp = new koronaVirusGovHuParser()
last_known_case_id = 19224;
console.log(kvghp.getInfoByPage(last_known_case_id));