hn job query search

hn_seach.js

// Usage:
//   Copy and paste all of this into a debug console window of the "Who is Hiring?" comment thread 
//   then use as follows:
//
//   query(term | [term, term, ...], term | [term, term, ...], ...)
// 
// When arguments are in an array then that means an "or" and when they are seperate that means "and"
//
// Term is of the format:
//    ((-)text/RegExp)  ( '-' means negation )
//
// A first argument of '+' signifies an additional pass on the filtered data as opposed to
// resetting everything.
//
// Example: Let's look for jobs in california that involve rust or python and not crypto:
//
//   > query('ca', '-crypto', ['rust', 'python']);
//   {filtered: '98.57%', query: 'ca AND NOT crypto AND (rust OR python)'}
//
// Then you see, "oh right, I don't care about blockchain either":
//
//   > query('+', '-blockchain');
//   {filtered: '98.57%', query: 'ca AND NOT crypto AND (rust OR python) AND NOT blockchain'}
//
// Another example:
//   > query(['ca', 'sf', 'san jose', 'mountan view'])
//   {filtered: '90.61%', query: '(ca OR sf OR san jose OR mountan view)'}
//
// COVID killed Silicon Valley. Quod Erat Demonstrandum!
//
// Changelog for 2022-08-02
// 
//  ADDED
//
//    * Negation via '-'
//
//    * Multi-pass querying via first argument being '+'
//
//    * Debugging query string added in the response
//
//  CHANGED
//
//    * "or" and "and" works the opposite of how it did previously. 
//      This form seems to be more useful.
//
//    * Whole word matching is default
//
//    * Terms such as "c++" are properly escaped
//
//  UPDATED
//
//    * Rewrote as an absurd implementation. 
//      I had a fun afternoon writing this.
//
function query(...queryList) {
  // HN is done with very unsemantic classes.
  let jobList = [...document.querySelectorAll('.c5a,.cae,.c00,.c9c,.cdd,.c73,.c88')],
      // Traverses up the dom stack trying to find a match of a specific class
      upto    = (node, klass) => node.classList.contains(klass) ? node : upto(node.parentNode, klass),
      display = (node, what)  => upto(node, 'athing').style.display = what,
      hide    = node => { display(node, 'none');  node.show = false},
      show    = node => { display(node, 'block'); node.show = true},

      // Use RegExp as is. Otherwise make it a case insensitive RegExp
      destring = what => [
        what[0] === '-',
        what.test ? what : new RegExp([
          '\\b',
          what.toString()
            .replace(/^-/,'')
            .replace(/[.*+?^${}()|[\]\\]/g, '\\$&'),
          '\\b'
        ].join(''), 'i'), what
      ];

  // This is our grand reset
  if(queryList[0] !== '+') { 
    jobList.forEach(show);

    // Have fun with that.
    query.hidden = +!( query.fn = [] );
  } else {
    queryList.shift();
  }

  // The AND is an artifact of the design. It's just iterative napped subsets
  query.fn = query.fn.concat(queryList.map(arg => {
    // Make it an array if it isn't one and pass it through our destring
    let orList = Array.of(arg).flat().map(destring);

    // If we're showing the job, then go through the list of terms
    // If all of them do not match, hide it, then return the length.
    query.hidden += jobList.filter(node => node.show 
        && orList.every(([neg, r]) => neg ^ !(node.innerHTML.search(r) + 1))
      ).map(hide).length;

    // You're on your own here - this is just the construction of
    // the debug string. There's far more reasonable ways to do this
    // But what fun would that be?!
    return (
      ' ('[+!!(orList.length - 1)] +
      orList.map(([neg, ig, r]) => ['', 'NOT '][+neg] + r.slice(+neg)).join(' OR ') + 
      ' )'[+!!(orList.length - 1)] 
    ).trim();
  }));
  
  return {
    filtered: (100 * query.hidden / jobList.length).toFixed(2) + '%',
    query: query.fn.join(' AND ')
  };
}

Thanks @kristopolous and @meiamsome, browser search functionality is definitely not enough to search hundreds of job positions!

Because I wanted a mix of both scripts (i.e. nested criterias AND regular expressions being first-class objects), and because it was fun to write, I ended up creating just another version which looks like this:

// Non-Angular Javascript contract positions in London or remote
hn.filter(
  hn.or(/(javascript|typescript)/i, /ES\d/, 'JS'),
  hn.not(/angular/i),
  /contract/i,
  hn.or(hn.and('ONSITE', /london/i), 'REMOTE')
);

Details at https://gist.github.com/frosas/4cadd8392a3c4af82ef640cbedea3027

This script loads all pages via AJAX; you may execute it before this one so you search on all pages instead of just first one

;(function ajaxLoadNextPage () {
    var more = document.querySelector('.comment-tree > tbody > tr:last-child a');
    if (more && more.innerHTML === "More") {    
        var httpRequest = new XMLHttpRequest();
        httpRequest.onreadystatechange = function () {
            if (httpRequest.readyState === XMLHttpRequest.DONE) {
                if (httpRequest.status === 200) {
                  more.remove();
                  var div = document.createElement('div');
                  div.innerHTML = httpRequest.responseText;
                  var nextHTML = div.querySelector('.comment-tree > tbody').innerHTML;
                  document.querySelector('.comment-tree > tbody').innerHTML += nextHTML;
                  ajaxLoadNextPage();
                } else {
                  alert('There was a problem with the request to ' + more.href);
                }
            }
        };
        httpRequest.open('GET', more.href);
        httpRequest.send();
    }
})();

Any plans to package this as an extension?

I was revisiting this this month ... I think what I really want these days is exclusion more than inclusion. For instance, I don't care about healthcare, remote e-learning or fintech (I find them to be huxsters trying to arbitrage broken markets with snake oil tech) but anyway ... a blacklist seems really useful ... I should do that instead.

This also works, replace the id with whatever you want.

curl 'https://hacker-news.firebaseio.com/v0/item/31947297.json?print=pretty' | jq '.kids' | grep -Po '[0-9]*' | xargs -n 1 -P 20 -I %% wget https://hacker-news.firebaseio.com/v0/item/%%.json\?print=pretty

Then you can grep that.

ok I updated it to implement all the things I've been musing about for 7 years and to hopefully make you laugh out loud while reading it.

It is extremely silly but hopefully not stupid and still legible

Damn it

I'm still in bed. I'll look when I'm at my office

You're right. I was so careful in this. damn it. That's extremely disappointing. I apparently foolishly introduced the bug here when I was just using a phone and their textbox interface: https://gist.github.com/kristopolous/19260ae54967c2219da8/revisions#diff-63e9a5e5dead19d4e7a3ee13c24221089b165a04e534e6e675d491e9422576d1

Fixed. Thanks for the report @nemanjam

Thank you.

Updated pagination code using fetch and async/await, and while instead of recursion. Forked from @Ivanca

It'd be nice to merge this into query().

(async function () {
    var more;
    while (more = document.querySelector('a.morelink')) {
        const r = await fetch(more.href);
        more.remove();
        const div = document.createElement('div');
        div.innerHTML = await r.text();
        document.querySelector('.comment-tree > tbody').innerHTML += div.querySelector('.comment-tree > tbody').innerHTML;
    }
})();

Also, maybe a bookmarklet? You can drag/drop or copy/paste to your boomarks toolbar. eg:

javascript:
/* /Say hello# */
(function () {
  alert('Hello, HN');
})();