EPUB to TXT via NodeJS

README.md

epub-to-text

epub-to-text.js

'use strict'

var fs = require('fs');
var EPub = require('epub');
var htmlToText = require('html-to-text');
var path = require('path');
var htmlParser = require('node-html-parser');

class EPUBToText {
  extract(sourceFile, callback, initialCallback) {
    var epub = new EPub(sourceFile);
    var klass = this;

    epub.on('end', function() {
      epub.flow.forEach(function(chapter, sequence) {
        epub.getChapter(chapter.id, function(err, html) {
          var txt = '';
          if (html) {
            txt = htmlToText.fromString(html.toString(), {ignoreHref: true});
          };
          var meta = {};
          meta.id = chapter.id;
          meta.excerpt = txt.trim().slice(0, 250);
          meta.size = txt.length
          meta.sequence_number = sequence
          if (chapter.title) {
            meta.title = chapter.title
          } else {
            meta.title = klass.getTitleFromHtml(html);
          }
          callback(err, txt, sequence, meta);
        });
      });
    });

    epub.on('end', function() {
      if (initialCallback) {
        initialCallback(null, epub.flow.length);
      };
    });

    epub.parse();
  }

  extractTo(sourceFile, destFolder, callback) {
    var totalCount;
    var processedCount = 0;

    this.extract(sourceFile, (err, txt, sequence) => {
      var destFile = destFolder + '/' + sequence + '-' + path.basename(sourceFile) + '.txt'
      fs.writeFileSync(destFile, txt);
      processedCount += 1;
      if (processedCount >= totalCount) {
        callback(null);
      }
    }, (err, numberOfChapters) => {
      totalCount = numberOfChapters
    });
  }

  getTitleFromHtml(html) {
    const root = htmlParser.parse(html);
    var title = root.querySelector('h1');
    if (title == null) {
      title = root.querySelector('title');
      if (title == null) {
        return '';
      };
    };
    return title.structuredText.replace("\n", " ");
  }
}

module.exports = EPUBToText;