zachleat / performance-leaderboard Goto Github PK

Per feedback here it would be nice to see a full log of all of the runs, rather than throwing non-median runs away (should later analysis be needed)

https://twitter.com/reubenlillie/status/1241866530591301632

cc @reubenlillie

Problem

If you run the script multiple times the report is overwritten, because the report folder name is the current date.

Started getting these when running tests this week. Only two sites, both of which do have some Axe errors (that I'm working on) - but unsure why those errors would cause timeouts. Any ideas here? Results in this file:

leaderboard.txt

When I run performance-leaderboard to test only one of my 11ty sites, first without any number of pass console.log( await PerfLeaderboard(urls) );
and second with 5 passes console.log( await PerfLeaderboard(urls, 5) );, I get a big difference (43%) between the first one (3 passes) and the second one (5 passes).

Here is the log:

Testing 1 sites:
(Site 1 of 1, run 1 of 3): https://corsa.netlify.app/posts/unfolding.html
(Site 1 of 1, run 2 of 3): https://corsa.netlify.app/posts/unfolding.html
(Site 1 of 1, run 3 of 3): https://corsa.netlify.app/posts/unfolding.html
[
  {
    url: 'https://corsa.netlify.app/posts/unfolding.html',
    finalUrl: 'https://corsa.netlify.app/posts/unfolding.html',
    lighthouseScore: 1,
    firstContentfulPaint: 1156.5130000000001,
    firstMeaningfulPaint: 1156.5130000000001,
    speedIndex: 1395.6202760075153,
    rank: 1
  }
]
Testing 1 sites:
(Site 1 of 1, run 1 of 5): https://corsa.netlify.app/posts/unfolding.html
(Site 1 of 1, run 2 of 5): https://corsa.netlify.app/posts/unfolding.html
(Site 1 of 1, run 3 of 5): https://corsa.netlify.app/posts/unfolding.html
(Site 1 of 1, run 4 of 5): https://corsa.netlify.app/posts/unfolding.html
(Site 1 of 1, run 5 of 5): https://corsa.netlify.app/posts/unfolding.html
[
  {
    url: 'https://corsa.netlify.app/posts/unfolding.html',
    finalUrl: 'https://corsa.netlify.app/posts/unfolding.html',
    lighthouseScore: 1,
    firstContentfulPaint: 969.73,
    firstMeaningfulPaint: 991.73,
    speedIndex: 969.73,
    rank: 1
  }
]

At the same time, DevTools Lighthouse gives me a result similar to the second one (speedIndex: 0.9s).

Is there any explanation?

https://dev.to/philw_/show-off-your-lighthouse-scores-in-eleventy-with-the-pagespeed-insights-api-1cpp

https://developers.google.com/speed/docs/insights/v5/get-started

Also switch our versioning to make the major version of performance-leaderboard always match the major version of lighthouse.

It would be very useful to be able to run a setup script for each URL in Chrome before lighthouse runs.

For example:

• To log into a secured area
• To set cookies
• Even (and I don't know if this is possible - I've been spoiled by WPT) to change where a domain resolves

How?

I don't know. This issue is really a discussion starter to see if it's something you are interested in adding, and to think about how/if it's feasible.

Using Puppeteer to script interactions seems sensible. I've done some Googling for ideas.

I've seen this project uses chrome-launcher to set up Chrome rather than Puppeteer directly so that could be a complication.

Resources

Running Lighthouse on Authenticated Pages with Puppeteer
[2 ways to] Using Puppeteer with Lighthouse
chrome-launcher README
Connecting Puppeteer to a Chrome-Launcher instance.
chromelauncher_puppeteer.js

cc @dgrammatiko

I don’t know if this is an official API or not but I think we may need to remove the carbon testing part, unfortunately.

Hi,

I have the following error when using this package inside a Docker container.
(Site 1 of 1, run 1 of 1): https://google.nl Logged an error with https://google.nl: TypeError: Cannot read property 'score' of undefined at ResultLogger.mapResult (/app/node_modules/performance-leaderboard/src/ResultLogger.js:237:56) at runLighthouse (/app/node_modules/performance-leaderboard/performance-leaderboard.js:88:37) at runMicrotasks (<anonymous>) at processTicksAndRejections (internal/process/task_queues.js:95:5) at async /app/src/index.js:113:30 [ { url: 'https://google.nl', error: '{}', sidequests: { '-weight.total': 1, '+weight.total': 1, '-weight.document': 1, '+weight.document': 1, '-weight.script': 1, '+weight.script': 1, '-weight.image': 1, '+weight.image': 1, '-weight.font': 1, '+weight.font': 1, '+weight.fontCount': 1, '-timeToFirstByte': 1, '-totalBlockingTime': 1, '-largestContentfulPaint': 1 } } ]

My code is:
`(async function() {

        let urls = [
            "https://google.nl",
        ];

        const options = {
            //axePuppeteerTimeout: 120000, // 120 seconds
            writeLogs: false, // Store audit data
            readFromLogDirectory: false, // Skip tests with existing logs
            onlyCategories: ["performance"],
            chromeFlags: [
                '--headless',
                '--disable-gpu',
                '--disable-dev-shm-usage',
                '--no-first-run'
            ],
        }

        try {
            const data = await PerfLeaderboard(urls, 1, options);
            console.log(data);
        } catch (error) {
            console.log(error);
        }

    })();`

What can this be?

Would there be an appetite to add web page carbon estimates back to the results (removed in #16) without depending on an external API?

The calculations outlined in this post https://sustainablewebdesign.org/calculating-digital-emissions/ could be used. They are the same as is used by the Website Carbon API which was previously used.

I'd be happy to contribute a PR (with some guidance).

Top level lighthouseVersion in lighthouse results.

https://developer.chrome.com/blog/lighthouse-10-0/