A better Trader has been a work-in-progress project of mine for a very long time. The web UI was done in vanilla javascript, with old school imports, and PHP-style server-generated html, then added jQuery, then bootstrap, then started removing jQuery, then moved towards more-static html with JSON payloads for the data, then pined for better node imports, then then then. You get the point.

I went too long without a rewrite, so I recently became a weekend warrior skeletoning up the next gen web ui. Parts include:

  • Vite due to its blazing hot-loading, tree-shaking, polyfilling goodness
  • React because it’s good enough and gets the job done
  • Bootstrap because I won’t have time to finish the mobile apps for a while

For my UIs, D3 is the most important library there is. So the skeleton is based on responsively displaying a handful of some of Mike Bostock’s greatest hits (imho). This was quite important to me because D3 examples have been somewhat obfuscated when they were migrated to Observables notebooks. The skeleton makes it much easier for me to quickly get working D3 examples by copying Mike’s Observables code into a cozy little container with all the bells and whistles in place.

Let me know what you think of it.

Reddit | Live Demo | GitLab | GitHub | Wiki

In c++, all a sort operation needs is a lessThan function. Complex sorts are EASY:

struct AnalyzerJobs_lessthan
{
    bool operator()(const AnalyzerJob* left, const AnalyzerJob* right) const
    {
        assert(left ->psq_->second && left ->psq_->second->p_shared_aps_ && left ->psq_->second->p_shared_aps_->p_ad_);
        assert(right->psq_->second && right->psq_->second->p_shared_aps_ && right->psq_->second->p_shared_aps_->p_ad_);

        // Primary sort
        // Critical jobs first
        // If only one is critical, it wins.
        if (left ->psq_->second->p_shared_aps_->bCriticalAnalysisNeeded() != right->psq_->second->p_shared_aps_->bCriticalAnalysisNeeded())
            return left ->psq_->second->p_shared_aps_->bCriticalAnalysisNeeded();

        // Secondary sort
        // If only one has never been analyzed, it wins.
        if (left->psq_->second->p_shared_aps_->bAnalyzed() != right->psq_->second->p_shared_aps_->bAnalyzed())
            return right->psq_->second->p_shared_aps_->bAnalyzed();

        // Tertiary sort by reverse timestamp of last analysis.
        double lefttime  = left ->psq_->second->p_shared_aps_->analysis_finished_timestamp_;
        double righttime = right->psq_->second->p_shared_aps_->analysis_finished_timestamp_;
        return (lefttime > righttime);
    }
};

You’d think JavaScript would keep that battle-tested methodology, but they decided it would be better to expect a result from the sort function of [ -1, 0, 1 ]. I’m assuming so they can eek out a bit of performance when sorting arrays that have a large number of “equal” items. Seems an outlier case to optimize for, but it probably puts it on some slightly higher score on some manic benchmark.

So for even the simplest sort, you have to do this stupidity:

myArray.sort(( a, b ) => ( a.value > b.value ? 1 : -1 ));

It seems to work as long as the sign is right, so this is also possible:

myArray.sort(( a, b ) => ( a.value - b.value ));

Also, remember that JavaScript provides all kinds of “helpful” dangerous ugly default conversions. Cest la vie. I have to remember though – I’ve used c++ style comparisons in JavaScript sort functions more than once, and that shit don’t fly. ok byeeeee!

Browsers are here to own our desktops. But they will never deliver the absolute horsepower of a natively-coded world-class desktop app or game.

These days, when you try to run one, you will find it is crippled by the browser’s 100 open tabs (admit it, you have that many open…). And lord knows you don’t want to close them all.

I am the user. I am the king. Not the last 20 sites I happened to browse, each of which is allowed to take a huge chunk of our system resources for whatever task they deem important. Once you’ve rabbit-holed down a few holes, you have given your entire system resources to them.

Don’t put up with this. OWN YOUR OWN SYSTEM.

To really take it back when you need to play a game, work in your DAW, or play a 4k video at high frame rate, you need that browser to give up ALL its resources. And in linux, that is super easy.

  • Find the parent browser process pid:
ps ax --forest| grep "[0-9][\_ ]*/usr/lib/firefox/firefox$"
  • “STOP” it. This is the magic that tells linux to just stop serving the parent process and all its children, stripping all CPU allocation from the tasks, without actually closing anything (especially your precious tabs scattered across all your virtual desktops). It’s amazing, it will freeze your browser in its tracks in an instant:
kill -STOP #pid-from-previous-step#
  • Start your powerhouse desktop app and do your thing.
  • When you’re done, you can “CONTINUE” the stopped process, kicking off the CPU to immediately start servicing all that javascript again:
kill -CONT #pid-from-previous-step#

I’ve scripted this up so I can press a media key to “stop” firefox and start up Kodi with beautiful performance. Here’s my javascript “continue” script, which uses my rad-scripts tooling:

#!/usr/bin/env node
import * as rs from 'rad-scripts';
const ffPs = rs.run_command_sync('ps ax --forest| grep "[0-9][\_ ]*/usr/lib/firefox/firefox$"');
const ffProcId = +( ffPs.trim().split(' ')[ 0 ] );
console.log( `Parent firefox pid = ${ffProcId}` );
rs.run_command_sync_to_console(`kill -CONT ${ffProcId}`);

Boost 1.77 is working well with it, and I’ve fully migrated to boost::json, which is working out great. The straw that broke the camel’s back was map.contains(), lolz. I was ready! I feel like I’ve had a hot bath. Nice.

TL;DR add this config to dnsmasq.conf to expand your LAN single-label hosts into FQDNs that systemd won’t hate:

domain=mydomain.lan
local=/mydomain.lan/
expand-hosts

For decades, DNS resolution was fine, and then systemd arrived and it was not. For the past couple years, for me and many others, systemd-resolved has been quite a headache, failing basic DNS resolution on any fresh Ubuntu (and other) installs. A real joy when you just want to get things going ASAP. After years of getting around it with shitty hacks, I finally nailed down what was going on.

Poettering and his crew decided that “single-label” host names on your LAN should never be allowed to resolve, even though it always used to in what he calls “classic DNS”. The systemd-resolved DNS handler in a vanilla install of Ubuntu 20.04 will drop your name resolution requests simply because the name does not have a dot in it. End of story. How nice. This awful default behavior follows some 2013 guidelines that were never previously enforced, at least not on me.

I use the simplest possible solutions that work, and for me that is dnsmasq. You can use one config file to configure static IPs for the MACs of all your LAN boxes. I just needed to reconfigure it to expand single-label names into longer “fake” FQDNs, with some dnsmasq hackery, and DNS on new boxes now works out of the gate. See the top of post (or this SO answer) for the configuration details.

There are supposed to be other solutions, like flipping on the LLMNR flag of systemd-resolved. That didn’t work for me, and I don’t want to have to futz with every single new LAN box and VM anyway. Hopefully this nightmare is behind me now. I hope for you, as well.