Datasets:

Qwen
/

RecreationBench

/* _chart_settle.js -- climatewatch.org /countries/<ISO>/ pages only.
 *
 * WHY
 * The flagship "What are <country>'s greenhouse gas emissions and emissions targets?"
 * chart is mounted only once /api/v1/emissions?gas=553&location=<ISO>&source=274 lands in
 * the Redux store. The app dispatches that request LAST, behind ~20 other API JSONs, and
 * the benchmark's static file server (serving/site_server.py -> http.server.HTTPServer)
 * is single-threaded, so that burst is answered one response at a time. Measured on
 * /countries/BRA/: the burst is issued at t=2.23s and drains until t=4.9s; the emissions
 * request only goes out at t=4.89s and the chart mounts at t=5.6-9.6s, with ~3s of
 * run-to-run jitter. recharts then animates every series in from a zero-width <clipPath>
 * (animationBegin 500ms + animationDuration 1500ms), so for a further ~2.0s the chart
 * paints its axes, tick labels and target dots but NONE of its data series.
 *
 * Playwright's `networkidle` (0 connections for 500ms) is meanwhile governed by things
 * that have nothing to do with the chart -- third-party beacons when the cache is cold
 * (clarity c.gif redirect chain, engage.wri.org page-event, GA/GTM 404s), and NOTHING at
 * all when the cache is warm, because then the whole API waterfall is served from disk.
 * Measured: networkidle fires anywhere between 1.0s and 9.6s. Every capture recipe in this
 * repo is "networkidle + N ms", so the captured DOM lands, at random, in one of three
 * states:
 *   1. before the mount           -> no .recharts-wrapper in the DOM at all
 *   2. inside the ~2.0s animation -> an empty axis grid with no data series
 *      (this is the state the shipped countries_BRA desktop GT was frozen in)
 *   3. after the animation        -> the correct chart
 *
 * WHAT THIS DOES
 * Nothing to the DOM, the styles or the app state. It only keeps one 1-byte request in
 * flight every 200ms -- below the 500ms networkidle threshold -- until the flagship chart
 * has both mounted and finished animating. That makes `networkidle` mean "the charts are
 * painted", which is what every capture recipe here already assumes it means, so captures
 * are always taken in state 3.
 *
 * The pinging is done from a Worker, not from this thread: parsing and booting the 10MB
 * app bundle blocks the main thread for ~3.5s, and a main-thread setInterval cannot fire
 * during that block -- which is exactly the window in which a warm-cache load goes quiet
 * and networkidle slips through. A Worker keeps pinging straight through it (verified).
 *
 * This file is loaded from <head> WITHOUT defer so the Worker is started before the app
 * bundle begins executing. Readiness detection is read-only (Resource Timing + a DOM
 * query); nothing is monkey-patched. Every wait is hard-capped, so a page that never
 * mounts a chart still reaches networkidle well inside goto()'s 30s timeout.
 *
 * State 2 is ALSO fixed independently, and has to be: the full-page screenshot resizes the
 * layout viewport, the scrollbar disappears, every ResponsiveContainer gets ~15px wider and
 * recharts restarts the animation from zero AFTER networkidle, where no amount of waiting
 * can help. `repair.disable_chart_animations` therefore forces isAnimationActive false in
 * packs/main-*.js. This shim owns state 1 (capture before the mount); the ANIM_MS tail below
 * is now only margin.
 */
(function () {
  'use strict';

  // Only the country profiles carry the late-mounting flagship chart.
  if (!/^\/countries\/[A-Za-z]{3}\/?$/.test(location.pathname)) return;

  var POLL_MS = 100;        // main-thread readiness poll
  var STABLE_MS = 400;      // the mounted chart must hold this long before we start the anim wait
  var ANIM_MS = 2400;       // recharts entry animation: 500ms begin + 1500ms duration + margin
  var DATA_CAP_MS = 12000;  // stop waiting for the emissions payload to appear
  var HARD_CAP_MS = 17000;  // absolute ceiling on this thread; the Worker has its own

  // This page requests far more than the 250-entry default, and we read Resource Timing
  // to know when the emissions payload landed. Raise the buffer before anything else loads.
  try { performance.setResourceTimingBufferSize(1000); } catch (e) { /* older engine */ }

  var worker = null;
  try { worker = new Worker('/_settle_worker.js'); } catch (e) { worker = null; }

  function stop() {
    if (!worker) return;
    // Ask it to stop; never terminate(). A hard terminate can strand an in-flight ping with no
    // finished/failed event, which leaves Playwright's networkidle counter above zero forever.
    try { worker.postMessage('stop'); } catch (e) { /* already gone */ }
    worker = null;
  }
  if (!worker) return;

  function emissionsSettled() {
    // Resource Timing lists cache hits too (transferSize 0), so this works warm or cold.
    var es;
    try { es = performance.getEntriesByType('resource'); } catch (e) { return false; }
    for (var i = 0; i < es.length; i++) {
      if (es[i].name.indexOf('/api/v1/emissions__') !== -1 && es[i].responseEnd > 0) return true;
    }
    return false;
  }

  function ghgMounted() {
    // recharts' ResponsiveContainer renders null until it has measured itself, so a
    // container with no .recharts-wrapper inside means "still mounting".
    var c = document.querySelector('[class*="country-ghg-emissions-styles__container"]');
    if (!c) return false;
    var rc = c.querySelectorAll('.recharts-responsive-container');
    if (!rc.length) return false;
    for (var i = 0; i < rc.length; i++) {
      if (!rc[i].querySelector('.recharts-wrapper')) return false;
    }
    return true;
  }

  var t0 = Date.now(), dataAt = 0, stableSince = 0, mountedAt = 0;
  var id = setInterval(function () {
    var now = Date.now(), age = now - t0;

    if (!dataAt && (emissionsSettled() || age >= DATA_CAP_MS)) dataAt = now;

    // Only trust the DOM once the payload has landed: before React's first render the DOM
    // still holds the frozen snapshot's own (already-rendered) chart markup.
    if (dataAt && !mountedAt) {
      if (ghgMounted()) {
        if (!stableSince) stableSince = now;
        if (now - stableSince >= STABLE_MS) mountedAt = now;
      } else {
        stableSince = 0;
      }
    }

    if ((mountedAt && now - mountedAt >= ANIM_MS) || age >= HARD_CAP_MS) {
      clearInterval(id);
      stop();
    }
  }, POLL_MS);
})();

	/* _chart_settle.js -- climatewatch.org /countries/<ISO>/ pages only.
	*
	* WHY
	* The flagship "What are <country>'s greenhouse gas emissions and emissions targets?"
	* chart is mounted only once /api/v1/emissions?gas=553&location=<ISO>&source=274 lands in
	* the Redux store. The app dispatches that request LAST, behind ~20 other API JSONs, and
	* the benchmark's static file server (serving/site_server.py -> http.server.HTTPServer)
	* is single-threaded, so that burst is answered one response at a time. Measured on
	* /countries/BRA/: the burst is issued at t=2.23s and drains until t=4.9s; the emissions
	* request only goes out at t=4.89s and the chart mounts at t=5.6-9.6s, with ~3s of
	* run-to-run jitter. recharts then animates every series in from a zero-width <clipPath>
	* (animationBegin 500ms + animationDuration 1500ms), so for a further ~2.0s the chart
	* paints its axes, tick labels and target dots but NONE of its data series.
	*
	* Playwright's `networkidle` (0 connections for 500ms) is meanwhile governed by things
	* that have nothing to do with the chart -- third-party beacons when the cache is cold
	* (clarity c.gif redirect chain, engage.wri.org page-event, GA/GTM 404s), and NOTHING at
	* all when the cache is warm, because then the whole API waterfall is served from disk.
	* Measured: networkidle fires anywhere between 1.0s and 9.6s. Every capture recipe in this
	* repo is "networkidle + N ms", so the captured DOM lands, at random, in one of three
	* states:
	* 1. before the mount -> no .recharts-wrapper in the DOM at all
	* 2. inside the ~2.0s animation -> an empty axis grid with no data series
	* (this is the state the shipped countries_BRA desktop GT was frozen in)
	* 3. after the animation -> the correct chart
	*
	* WHAT THIS DOES
	* Nothing to the DOM, the styles or the app state. It only keeps one 1-byte request in
	* flight every 200ms -- below the 500ms networkidle threshold -- until the flagship chart
	* has both mounted and finished animating. That makes `networkidle` mean "the charts are
	* painted", which is what every capture recipe here already assumes it means, so captures
	* are always taken in state 3.
	*
	* The pinging is done from a Worker, not from this thread: parsing and booting the 10MB
	* app bundle blocks the main thread for ~3.5s, and a main-thread setInterval cannot fire
	* during that block -- which is exactly the window in which a warm-cache load goes quiet
	* and networkidle slips through. A Worker keeps pinging straight through it (verified).
	*
	* This file is loaded from <head> WITHOUT defer so the Worker is started before the app
	* bundle begins executing. Readiness detection is read-only (Resource Timing + a DOM
	* query); nothing is monkey-patched. Every wait is hard-capped, so a page that never
	* mounts a chart still reaches networkidle well inside goto()'s 30s timeout.
	*
	* State 2 is ALSO fixed independently, and has to be: the full-page screenshot resizes the
	* layout viewport, the scrollbar disappears, every ResponsiveContainer gets ~15px wider and
	* recharts restarts the animation from zero AFTER networkidle, where no amount of waiting
	* can help. `repair.disable_chart_animations` therefore forces isAnimationActive false in
	* packs/main-*.js. This shim owns state 1 (capture before the mount); the ANIM_MS tail below
	* is now only margin.
	*/
	(function () {
	'use strict';

	// Only the country profiles carry the late-mounting flagship chart.
	if (!/^\/countries\/[A-Za-z]{3}\/?$/.test(location.pathname)) return;

	var POLL_MS = 100; // main-thread readiness poll
	var STABLE_MS = 400; // the mounted chart must hold this long before we start the anim wait
	var ANIM_MS = 2400; // recharts entry animation: 500ms begin + 1500ms duration + margin
	var DATA_CAP_MS = 12000; // stop waiting for the emissions payload to appear
	var HARD_CAP_MS = 17000; // absolute ceiling on this thread; the Worker has its own

	// This page requests far more than the 250-entry default, and we read Resource Timing
	// to know when the emissions payload landed. Raise the buffer before anything else loads.
	try { performance.setResourceTimingBufferSize(1000); } catch (e) { /* older engine */ }

	var worker = null;
	try { worker = new Worker('/_settle_worker.js'); } catch (e) { worker = null; }

	function stop() {
	if (!worker) return;
	// Ask it to stop; never terminate(). A hard terminate can strand an in-flight ping with no
	// finished/failed event, which leaves Playwright's networkidle counter above zero forever.
	try { worker.postMessage('stop'); } catch (e) { /* already gone */ }
	worker = null;
	}
	if (!worker) return;

	function emissionsSettled() {
	// Resource Timing lists cache hits too (transferSize 0), so this works warm or cold.
	var es;
	try { es = performance.getEntriesByType('resource'); } catch (e) { return false; }
	for (var i = 0; i < es.length; i++) {
	if (es[i].name.indexOf('/api/v1/emissions__') !== -1 && es[i].responseEnd > 0) return true;
	}
	return false;
	}

	function ghgMounted() {
	// recharts' ResponsiveContainer renders null until it has measured itself, so a
	// container with no .recharts-wrapper inside means "still mounting".
	var c = document.querySelector('[class*="country-ghg-emissions-styles__container"]');
	if (!c) return false;
	var rc = c.querySelectorAll('.recharts-responsive-container');
	if (!rc.length) return false;
	for (var i = 0; i < rc.length; i++) {
	if (!rc[i].querySelector('.recharts-wrapper')) return false;
	}
	return true;
	}

	var t0 = Date.now(), dataAt = 0, stableSince = 0, mountedAt = 0;
	var id = setInterval(function () {
	var now = Date.now(), age = now - t0;

	if (!dataAt && (emissionsSettled() \|\| age >= DATA_CAP_MS)) dataAt = now;

	// Only trust the DOM once the payload has landed: before React's first render the DOM
	// still holds the frozen snapshot's own (already-rendered) chart markup.
	if (dataAt && !mountedAt) {
	if (ghgMounted()) {
	if (!stableSince) stableSince = now;
	if (now - stableSince >= STABLE_MS) mountedAt = now;
	} else {
	stableSince = 0;
	}
	}

	if ((mountedAt && now - mountedAt >= ANIM_MS) \|\| age >= HARD_CAP_MS) {
	clearInterval(id);
	stop();
	}
	}, POLL_MS);
	})();