parcer/mx-interpreter/browserSide/scraper.js

/* eslint-disable @typescript-eslint/no-unused-vars */

const area = (element) => element.offsetHeight * element.offsetWidth;

function getBiggestElement(selector) {
  const elements = Array.from(document.querySelectorAll(selector));
  const biggest = elements.reduce(
    (max, elem) => (
      area(elem) > area(max) ? elem : max),
    { offsetHeight: 0, offsetWidth: 0 },
  );
  return biggest;
}

/**
 * Generates structural selector (describing element by its DOM tree location).
 *
 * **The generated selector is not guaranteed to be unique!** (In fact, this is
 *    the desired behaviour in here.)
 * @param {HTMLElement} element Element being described.
 * @returns {string} CSS-compliant selector describing the element's location in the DOM tree.
 */
function GetSelectorStructural(element) {
  // Base conditions for the recursive approach.
  if (element.tagName === 'BODY') {
    return 'BODY';
  }
  const selector = element.tagName;
  if (element.parentElement) {
    return `${GetSelectorStructural(element.parentElement)} > ${selector}`;
  }

  return selector;
}

/**
 * Heuristic method to find collections of "interesting" items on the page.
 * @returns {Array<HTMLElement>} A collection of interesting DOM nodes
 *  (online store products, plane tickets, list items... and many more?)
 */
function scrapableHeuristics(maxCountPerPage = 50, minArea = 20000, scrolls = 3, metricType = 'size_deviation') {
  const restoreScroll = (() => {
    const { scrollX, scrollY } = window;
    return () => {
      window.scrollTo(scrollX, scrollY);
    };
  })();

  /**
* @typedef {Array<{x: number, y: number}>} Grid
*/

  /**
 * Returns an array of grid-aligned {x,y} points.
 * @param {number} [granularity=0.005] sets the number of generated points
 *  (the higher the granularity, the more points).
 * @returns {Grid} Array of {x, y} objects.
 */
  function getGrid(startX = 0, startY = 0, granularity = 0.005) {
    const width = window.innerWidth;
    const height = window.innerHeight;

    const out = [];
    for (let x = 0; x < width; x += 1 / granularity) {
      for (let y = 0; y < height; y += 1 / granularity) {
        out.push({ x: startX + x, y: startY + y });
      }
    }
    return out;
  }

  let maxSelector = { selector: 'body', metric: 0 };

  const updateMaximumWithPoint = (point) => {
    const currentElement = document.elementFromPoint(point.x, point.y);
    const selector = GetSelectorStructural(currentElement);

    const elements = Array.from(document.querySelectorAll(selector))
      .filter((element) => area(element) > minArea);

    // If the current selector targets less than three elements,
    // we consider it not interesting (would be a very underwhelming scraper)
    if (elements.length < 3) {
      return;
    }

    let metric = null;

    if (metricType === 'total_area') {
      metric = elements
        .reduce((p, x) => p + area(x), 0);
    } else if (metricType === 'size_deviation') {
      // This could use a proper "statistics" approach... but meh, so far so good!
      const sizes = elements
        .map((element) => area(element));

      metric = (1 - (Math.max(...sizes) - Math.min(...sizes)) / Math.max(...sizes));
    }

    // console.debug(`Total ${metricType} is ${metric}.`)
    if (metric > maxSelector.metric && elements.length < maxCountPerPage) {
      maxSelector = { selector, metric };
    }
  };

  for (let scroll = 0; scroll < scrolls; scroll += 1) {
    window.scrollTo(0, scroll * window.innerHeight);

    const grid = getGrid();

    grid.forEach(updateMaximumWithPoint);
  }

  restoreScroll();

  let out = Array.from(document.querySelectorAll(maxSelector.selector));

  const different = (x, i, a) => a.findIndex((e) => e === x) === i;
  // as long as we don't merge any two elements by substituing them for their parents,
  // we substitute.
  while (out.map((x) => x.parentElement).every(different)
  && out.forEach((x) => x.parentElement !== null)) {
    out = out.map((x) => x.parentElement ?? x);
  }

  return out;
}

/**
 * Returns a "scrape" result from the current page.
 * @returns {Array<Object>} *Curated* array of scraped information (with sparse rows removed)
 */
function scrape(selector = null) {
  /**
   * **crudeRecords** contains uncurated rundowns of "scrapable" elements
   * @type {Array<Object>}
   */
  const crudeRecords = (selector
    ? Array.from(document.querySelectorAll(selector))
    : scrapableHeuristics())
    .map((record) => ({
      ...Array.from(record.querySelectorAll('img'))
        .reduce((p, x, i) => {
          let url = null;
          if (x.srcset) {
            const urls = x.srcset.split(', ');
            [url] = urls[urls.length - 1].split(' ');
          }

          /**
             * Contains the largest elements from `srcset` - if `srcset` is not present, contains
             * URL from the `src` attribute
             *
             * If the `src` attribute contains a data url, imgUrl contains `undefined`.
             */
          let imgUrl;
          if (x.srcset) {
            imgUrl = url;
          } else if (x.src.indexOf('data:') === -1) {
            imgUrl = x.src;
          }

          return ({
            ...p,
            ...(imgUrl ? { [`img_${i}`]: imgUrl } : {}),
          });
        }, {}),
      ...record.innerText.split('\n')
        .reduce((p, x, i) => ({
          ...p,
          [`record_${String(i).padStart(4, '0')}`]: x.trim(),
        }), {}),
    }));

  return crudeRecords;
}