Before I wrote a cron script to check website is down, but it can't check multiple endpoint and can't alert if ssl certificates is about to expired.

So, I created this bash snippet function to check multiple endpoints (and their SSL certificates)

#!/bin/bash

function checkEndpoint() {
    URL="$1"
    FIND="$2"

    RESPONSE=$(curl -m 10 -s "$URL")
    CONTENT=$(echo $RESPONSE | grep "$FIND")

    if [ $? -ne 0 ]; then
        echo "Error: $URL same down. Please check!"
        return 1
    fi

    # Check SSL certificate expiration
	HOST=$(echo "$URL" | sed -E 's|https?://([^/]+).*|\1|')
    EXPIRY_DATE=$(echo | openssl s_client -connect "$HOST:443" -servername "$HOST" 2>/dev/null | openssl x509 -noout -enddate | cut -d= -f2)
    EXPIRY_TIMESTAMP=$(date -d "$EXPIRY_DATE" +%s)
    CURRENT_TIMESTAMP=$(date +%s)
    DAYS_LEFT=$(( (EXPIRY_TIMESTAMP - CURRENT_TIMESTAMP) / 86400 ))

    if [ "$DAYS_LEFT" -le 7 ]; then
        echo "Error: $URL - SSL certificate will expire in $DAYS_LEFT days."
        return 1
    fi

    return 0
}

function checkAndAlert() {
	CHAT_ID="telegram_chat_id"
	BASIC_AUTH_BOT_PARAM="bot*****:*********"
    ALERT_MSG=$(checkEndpoint "$1" "$2")

    if [ $? -ne 0 ]; then
		curl -s -m 10 --get --data-urlencode "chat_id=$CHAT_ID" --data-urlencode "text=$ALERT_MSG" "https://api.telegram.org/$BASIC_AUTH_BOT_PARAM/sendMessage"
        return 1
    fi
}

checkAndAlert "https://google.com/" "google"
checkAndAlert "https://news.ycombinator.com" "Hacker News"

In PHP, when you pass variables into another function arguments, it

1. Copy the variable memory if the argument is scalar data type (int, float, bool, ...), because it's cheap operations
2. Copy a shadow clone jutsu variable zval struct (24-32 bytes), then point the real data pointer of dynamically data type (array, string, $object)

Below testing will say more than me

<?php

class A {
  public $arr;
  public function __construct() {
    $this->arr = array_fill(0, 10000, "Hello world");
  }

  public function push(){
    $this->arr[] = "ok";
  }
}

function haha() {
  $a = func_get_args()[0];
  mem("inside func haha, after func_get_args");
  return $a;
}

function a($b, $c) {
  $a = new A();
  $str = str_repeat("Helo", 100000);
  mem("after create an object A");
  $x = haha($a, $str, $b, $c);
  mem("after call func_get_args");
  $x->arr[] = "there";
  var_dump($a->arr[count($a->arr) - 1]);
  var_dump($x->arr[count($x->arr) - 1]);
  $x->push();
  var_dump($a->arr[count($a->arr) - 1]);
  var_dump($x->arr[count($x->arr) - 1]);
  mem("after modify arr property");
  // var_dump($a === $x);
  return $b+$c;
}

function mem($msg = null) {
  if ($msg) print_r($msg . " : ");
  echo (memory_get_usage(false) . " bytes\n");
}

mem("start");
var_dump(a(1,2));
mem("end");

This is the result

start : 469464 bytes
after create an object A : 1137200 bytes
inside func haha, after func_get_args : 1137200 bytes
after call func_get_args : 1137200 bytes
string(5) "there"
string(5) "there"
string(2) "ok"
string(2) "ok"
after modify arr property : 1137200 bytes
int(3)
end : 469464 bytes

Surprise ! :D

Vite server has a middleware named viteHMRPingMiddleware in this source code

https://github.com/vitejs/vite/blob/81327eb980c308474a586a9cb9c0c5fff10eba34/packages/vite/src/node/server/index.ts#L876

So we have a simple way to check if Vite HMR server is running by calling a ping http request to the url endpoint (with timeout 1s)

function checkViteHMR() {
	fetch('http://localhost:5173/_ping', {
		method: 'HEAD',
		headers: {
			'Accept': 'text/x-vite-ping'
		}
	})
	.then(response => {
		if (response.status === 204) {
			// Do something here
			console.log('Vite HMR is running');
		} else {
			console.log('Received response with status:', response.status);
		}
	})
	.catch(error => {
		console.error('Error making request:', error);
	});
}

By default, all script tags will be turned off when you try to insert to a element using innerHTML. So the tricks to make it run to create the new script tag using document.createElement('script') then copy the original scripts to new scripts. After then append the scripts to body or the parent element.

<div>
  <h1>Hello from Test</h1>
  <script data-sign="hello">
    console.log(document.currentScript);
    const parent = document.currentScript.parentElement;
    parent.style.color = 'blue';
  </script>
</div>

<!DOCTYPE html>
<html lang="en">
  <body>
    <div id="test"></div>
  </body>
  <script>
    const ele = document.getElementById('test');
    fetch('test.html')
      .then((res) => res.text())
      .then((content) => {
        ele.innerHTML = content;
        ele.querySelectorAll('script').forEach((script) => {
          console.log(script.dataset.sign || 'No signature'); // Your can use this to check the signature and minimize XSS attacks
          const newScript = document.createElement('script');
          newScript.textContent = script.textContent;
          ele.appendChild(newScript);
        });
      });
  </script>
</html>

HTMX Library use the same way to do the trick, to turn off using htmx.config.allowScriptTags = false

This is shell script using curl and grep to send request then check if the keyword (FIND) include in the response text body. If failed, try to alert

URL="https://news.ycombinator.com/"
FIND="Hacker News"

response=$(curl -m 10 -s "$URL")
content=$(echo $response | grep "$FIND")

if [ $? -eq 0 ]; then
	echo "OK!"
	exit 0
fi

# Do watever you want to alert
CHAT_ID="telegram_chat_id"
TEXT="The web is down !"
BASIC_AUTH_BOT_PARAM="bot*****:*********"

curl -s -m 10 --get --data-urlencode "chat_id=$CHAT_ID" --data-urlencode "text=$TEXT" "https://api.telegram.org/$BASIC_AUTH_BOT_PARAM/sendMessage"
exit 1

XCode (the main IDE of Apple), but lacks much common features like auto formating code.
So we have to install a external tool (lol it's Apple tool itself) by brew install swift-format

Then I have to use a trick in Build Phases settings in XCode, we create a bash script run format tool on every build time (before source code has been built, because sometimes your code buidling will failed)

Here is the bash script, make sure the Run script below "Run Build Tool Plug-ins" step

echo "Formating all source codes"
cd $SOURCE_ROOT
/opt/homebrew/bin/swift-format --ignore-unparsable-files -i -p -r .
echo "Formated all source codes."

Important : You have to turn off Build Options named User Script Sandboxing (this diffenrence of App Sandboxing), to allow your shell scripts can affected real filesystem instead of shadowing sandbox files, below is the screenshot

Some IDEs has good LSP or Intellisense can show link of class methods, so we can using this mechanism to listing all the routes to faster searching route and click method link to navigate actual code.

I make this Laravel command to generate a stub file into routes directory so the IDE can index

So sometimes when you need to re-generate stubs, run this command

$ php artisan routing_stubs

<?php
namespace App\Console\Commands;

use Illuminate\Console\Command;
use Illuminate\Support\Facades\Route;

class RoutingStubs extends Command
{
    /**
     * The name and signature of the console command.
     *
     * @var string
     */
    protected $signature = 'routing_stubs';

    /**
     * The console command description.
     *
     * @var string
     */
    protected $description = 'Generate routing stubs';

    public function handle()
    {
        $routes = collect(Route::getRoutes())
            ->map(fn (\Illuminate\Routing\Route $route) => explode('@', $route->getActionName()))
            ->mapToGroups(fn ($params) => [$params[0] => $params[1] ?? ''])
            ->filter(fn ($methods, $class) => class_exists($class))
            ->map(fn ($methods, $class) => $methods->map(
                fn ($method) => sprintf("* @uses %s::%s", $class, $method)
            )->join("\n"))
            ->flatten()
            ->join("\n");

        file_put_contents(base_path('routes/stubs.php'), "<?php\n/**\n" . $routes . "\n*/\n");

        $this->info('Done');
    }
}

Learned from https://www.phpied.com/minimum-viable-no-image-image-src/

You can use this for lazy load images in pages

<img 
  src="data:image/svg+xml,<svg xmlns='http://www.w3.org/2000/svg'/>" 
  ...
>

Hmm, when I inspected his demo image, I saw new img attribute named decoding="async", so interesting! So we can tell browser not to wait process this dummy image for later and priority the rest higher. (Good for LCP vital metric)

To transform a array to stdClass in PHP, try to cast type using (object) before the array variable.

$arr = ['a' => 1, 'b' => 5];
$obj = (object) $arr;
echo $obj->b;

window.btoa is built-in function to encode string to base64 string, but it can't process any Unicode characters (each char must be fit in 1 byte, but Unicode is multi-byte char)

So solution is transform the value to Latin character using some pre-processing function like

const new_btoa = (val) => btoa(unescape(encodeURIComponent(val)))
const result = new_btoa('Hế lô xin chào')

So you have to pair the atob too

const new_atob = (val) => decodeURIComponent(escape(atob(val)))
console.log(new_atob(result))

Or using 3rd-lib like https://www.npmjs.com/package/js-base64 (I recommend use this way)

Old fashion way

$groupTrue = $groupFalse = [];
foreach ($items as $item) {
	if (some_logic_function($item)) {
		$groupTrue[] = $item;
	} else {
		$groupFalse[] = $item;
	}
}

New way

list($groupTrue, $groupFalse) = collect($items)
	->partition(fn ($item) => some_logic_function($item))
	->map(fn ($group) => $group->values());

tip: Using the last map then values to reset indexing of 2 groups after spliting.

Svelte uses it compiler to detect the reactive dependencies, to trigger the reactive updater function. So our work need is make sure all dependent variables appear in the line.

I use the below logic trick like this (!0 always true so it won't run init array [x,z,y]), but compiler always see the x,z,y in dependencies. :D

result = (!0 || [x,y,z]) ? calc() : null;

OK, here is sample Svelve 4 code (sure Svelte 5 Runes is the saver)

<script>
	let x = 1;
	let y = 1;	
	let z = 1;	
	let result = 0;

	function calc() {
			return x * y * z;
	}

	$: result = (!0 || [x,y,z]) ? calc() : null;

	function handleClickX() {
		x += 1;
	}
	
	function handleClickY() {
		y += 1;
	}

	function handleClickZ() {
		z += 1;
	}
</script>

<button on:click={handleClickX}>
	x = {x}
</button>
<button on:click={handleClickY}>
	y = {y}
</button>
<button on:click={handleClickZ}>
	z = {z}
</button>

<p>Result = {result}</p>

Never ever upgrade OS without :

• Backup important files (better sync it to cloud every work hour)
• Unplug all devices or accessories from all USB ports
• Plugged in power cable

Yesterday, I upgraded macOS to another version when plugged to the USB-C hub (HDMI external screen, USB keyboard, USB for LED light). Then it restarted few times and hang with black screen. After waiting 10 mins nothing changes, I power off the macbook hard way (by hold power button). Then I never can boot it again (even to Recovery mode).

After researching whole Internet (Google, MacForum, Youtube) and trying 42 times, I failed to revive it! :(

But, keep searching I found my life saver's blog post , now I can restore the Macbook and reinstall clean MacOS Sonoma.

Thanks GOD :D

Btw, never stupid like me ! And if the thing works fine, keep it works ! If not, carefully upgrade it and always have a plan B.

I created a small util Cache storage using redis server to cache seriable data, it stores serialzied data string in text then unserialize to wake the data up.

<?php

namespace App\Library;

use Predis\Client;
use Predis\PredisException;

class CacheStorage
{
    public function __construct(private Client $redis)
    {
    }

    public function remember($key, $ttl, $dataFn)
    {
        try {
            $cached_str = $this->redis->get($key);
            $cached = is_null($cached_str) ? false : unserialize($cached_str);

            if ($cached) {
                return $cached;
            }

            $freshData = $dataFn();
            $this->redis->set($key, serialize($freshData), "ex", $ttl);

            return $freshData;
        } catch (PredisException $e) {
            return $dataFn();
        }
    }
}

function getData(range, columnMap) {
  const dataValues = range.getValues();
  const mapEntries = Object.entries(columnMap)

  return dataValues.map((row) => Object.fromEntries(mapEntries.map((colMap) => [colMap[0], row[colMap[1]]])))
}

Example:

You have a sheet with data from A1:D*** (*** is last row), with the field mappings :

• first col is product_code
• second col is total
• fourth col is rank1

So the code of main function is

function main(){
  // Get the sheet you want
  var sheet = SpreadsheetApp.getActiveSpreadsheet().getSheets()[0];
  var lastRow = sheet.getLastRow();
  var dataRange = sheet.getRange("A1:D" + (lastRow));
  
  var data = getData(dataRange, {
    "product_code": 0,
    "total": 1,
    // skip third column
    "rank1": 3,
  })

  console.log(data)
}

The DOM event handling go through 2 phrases : Capture and Bubble (with support of preventDefault, stopPropagation, stopImmediatePropagation methods to modify behavior)

Try the demo here : https://svelte.dev/repl/99d1edbf7bd9426384be2c7763d2e872?version=4.2.0

<script>
	const onClickEle = (e) => {
		console.log((new Date()).toISOString() + " : Clicked " + e.currentTarget.tagName)
	}
</script>

<h5>Default (bubble)</h5>
<div on:click={onClickEle}>
	<button on:click={onClickEle}>Click here!</button>
</div>

<h5>Default (bubble with stopPropagation)</h5>
<div on:click={onClickEle}>
	<button on:click|stopPropagation={onClickEle}>Click here!</button>
</div>

<h5>Parent capture</h5>
<div on:click|capture={onClickEle}>
	<button on:click={onClickEle}>Click here!</button>
</div>

<h5>Parent capture stopPropagation</h5>
<div on:click|capture|stopPropagation={onClickEle}>
	<button on:click={onClickEle}>Click here!</button>
</div>

<h5>Self (ele = event target)</h5>
<div on:click|self={onClickEle}>
	<button on:click={onClickEle}>Click here!</button>
</div>

<style>
		div { padding: 10px 30px; background: gray; cursor: pointer; display: inline }
</style>

moduleA.js

export const message = "Hello from moduleA!";

moduleB.js

import { message } from './moduleA.js';
console.log(message);

app.html

<script type="module" src="moduleB.js"></script>

// Print console log "Hello from moduleA!"

So moduleB.js acts like a module entry-point, it loads all dependencies deeply to run it-self, so you only need add script tag point to it.

Reactivity in Svelte works via assignment of variable, so we have this as example

let todos = [];

// Don't work
todos.push({id: 1, name: 'cleaning'})
todos.push({id: 2, name: 'coding'})

// Work
todos = [...todos, {id: 1, name: 'cleaning'}]

// or this even work as tricky way
todos.push({id: 1, name: 'cleaning'})
todos = todos

Why ? I don't know in internal system how it implement, but I think because Svelte works as a compiler, so it triggers the update when it see the assignment variable line.

Different with another frameworks, they work based on proxy object

Object.prototype.groupBy = function(cb) {
  const groupByCategory = this.reduce((group, item) => {
    const category = cb(item);
    group[category] = group[category] ?? [];
    group[category].push(item);
    return group;
  }, {});

  return groupByCategory;
}

// Example
const people = [{"name":"John","age":28},{"name":"Alice","age":24},{"name":"Michael","age":32},{"name":"Emily","age":29},{"name":"David","age":27},{"name":"Sophia","age":38}];

const by_first_age_period = people.groupBy((person) => Math.floor(person.age / 10) * 10 + '+');

console.log(by_first_age_period);

[abc*="something"] means select element has attribute abc, and its value contains "something" string

Example :

/* Warning all admin link is red text color */
a[href*="/admin/"] {
	color: red;
}

• HTTP 307 : Temporary redirect, keep METHOD and BODY
• HTTP 308 : Permanent redirect, keep METHOD and BODY

vs

• HTTP 301 : Permanent redirect, fallback to GET method
• HTTP 302 : Temporary redirect, fallback to GET method

Many examples, articles and repository use insecure config in port mapping Docker.

This is example in official document of Docker

https://docs.docker.com/engine/reference/run/#detached--d

$ docker run -d -p 80:80 my_image nginx -g 'daemon off;'

or

from 12.2k stars repo in Github

https://github.com/chatwoot/chatwoot/blob/develop/docker-compose.yaml#L89

    ports:
      - '5432:5432'

This mapping config means you will proxy all traffic from the container port to the docker host (your dev machine) with binding 0.0.0.0. It means if you don't have a firewall or accidently open all ports, your contaier data will be public in the Local Network (so danger if you use in public places like coffee shops)

So remember to add the IP, which you wanted to bind in docker host (or have a network firewall ON)

    ports:
      - '127.0.0.1:5432:5432'
      - '172.17.0.1:5432:5432'

By default, all docker container using DNS server from docker host, so if you want to overwrite the specific hostnames ip address, try the flag --add-host [hostname]:[ip]

$ docker run -it --add-host db1:1.2.3.4 --add-host db2:2.3.4.5 alpine cat /etc/hosts
127.0.0.1       localhost
::1     localhost ip6-localhost ip6-loopback
fe00::0 ip6-localnet
ff00::0 ip6-mcastprefix
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters
1.2.3.4 db1
2.3.4.5 db2
172.17.0.2      0bdcf2fb2216

If you use docker-compose, then add the extra-hosts param below the service you want

    extra_hosts:
      - "crm.dev:host-gateway"
      - "db1:1.2.3.4"

Bonus : host-gateway mean the ip address of the gateway of network container in, often be the docker host!

Last day I updated my Laravel app from version 8 to 9, all my scheduled command in Console Kernel has run in another environment like "production". Which causes my system didn't work as expected.

Example, I have 2-3 enviroment which I run cronjob like

* * * * * php artisan schedule:run --env=hello
* * * * * php artisan schedule:run --env=world

But all sub scheduled commands has been run with "production" instead of "hello" or "world"

So this is my fix, I created a lambda function which pass current environment to the child command.

class Kernel extends ConsoleKernel
{
    protected function schedule(Schedule $schedule)
    {
        // Added 2 lines to get lamda function
        $env = app()->environment();
        $schedule_command = fn ($cmd) => $schedule->command($cmd, ['--env' => $env]);

        // Then schedule using new lambda function
        $schedule_command('check_product_stock')->everyTenMinutes();
        $schedule_command('run_job_queue')->everyMinute();
        $schedule_command('clean_old_data')->dailyAt('00:25');
    }

    protected function commands()
    {
        $this->load(__DIR__ . '/Commands');

        require base_path('routes/console.php');
    }
}

Enjoy ! :) Hope it works for you !

In case you use google/apiclient package to connect to Google API services, remember to cleaning unused services after deploying to production.

So let me show you why

First, lets counting number of classes of Google API services inside composer autoload classmap file (Composer use this file as a mapping to lazyload your needed class)

$ cat vendor/composer/autoload_classmap.php | grep 'Google' | wc -l
15436
$ du -h vendor/composer/autoload_classmap.php
3.5M    vendor/composer/autoload_classmap.php

As you can see, it has almost 15k classes having Google in class namespace. And your autoload_classmap.php has 3.5Mb in size (so large)

So, let cleaning unused services (by keeping only services you need)

Editing your composer.json file by adding "pre-autoload-dump" in scripts and "google/apiclient-services" in extra, like below (replace services you need like Drive, Youtube, Sheets, etc..)

{
    "scripts": {
        "pre-autoload-dump": "Google\\Task\\Composer::cleanup"
    },
    "extra": {
        "google/apiclient-services": [
            "Drive",
            "YouTube",
	    "Sheets"
        ]
    }
}

Finally, run composer dump-autoload again and see the result

$ cat vendor/composer/autoload_classmap.php | grep 'Google' | wc -l
383
$ du -m vendor/composer/autoload_classmap.php
1.3M    vendor/composer/autoload_classmap.php

Class map reduce from 15k to 400 lines, and size from 3.5M to 1.3M ! Such an optimize !

Beware : If you add services after cleaning, try to run rm -r vendor/google/apiclient-services && composer update to redownloading all services classes.

Laravel 9 uses "Spatie/Ignition" as default error handling pages , which is beatiful but bloated within add-on service like Flare.

So if you want to get back the good-old "Whoops" error pages, this tutorial is for you :D

First, we remove "spatie/ignition" from composer packages

$ composer remove "spatie/laravel-ignition"

Then, add 3 lines in the boot function of AppServiceProvider

use Illuminate\Contracts\Foundation\ExceptionRenderer;
use Illuminate\Foundation\Exceptions\Whoops\WhoopsExceptionRenderer;

class AppServiceProvider extends ServiceProvider {
    public function boot()
    {
        // Add 3 lines below between boot function
        if (config('app.debug')) {
            $this->app->bind(ExceptionRenderer::class, WhoopsExceptionRenderer::class);
        }
    }
}

And last but not least, open the config/app.php then add to this line to array config (try to replace the name of Editor you use in this list )

    ...
    'editor' => 'vscode'
    ...

Now try to add this line to second line in web.php route file to test

$x = 1 / 0;

By default, the document root of a domain in DirectAdmin (DA) has the format

/home/[your_username]/domains/[your_domain]/public_html

But in case you have to run modern PHP app like Laravel or Symfony framework (they use child public directory as their docroot). In some case, using .htaccess file to rewrite is a hacky way to solve problem (or this is the option when you don't have Admin account of DA)

So, if you have an Admin account, try this

Open menu Server Manager > Custom HTTPD Configurations, then click on your domain httpd config file, then click CUSTOMIZE

In the large textbox, add this line (remind to replace username and domain) to

|?DOCROOT=/home/[your_username]/domains/[your_domain]/public_html/public|

Click SAVE to save modification. Then go to Service Monitor and reload your webservers (Nginx, Apache or OpenLiteSpeed)

Yo! Back to my #TIL

Today, I find out a way to insert fastly current date or current time to a cell. It makes you faster 3-5 seconds and reduce confusing about what the time is?

• To insert current date : press Ctrl + :
• To insert current time : press Ctrl + Shift + :

Excel (Spreadsheets) never dies !

TLDR;

Use JSON.parse("[your data in json string]") if your data is big (>10KB for instance)

Short Explaination :

JSON.parse parses a string to object, so it has many strict requirements than Javascript parses the source code (more syntax, more context)

Long Explaination :

https://www.youtube.com/watch?v=ff4fgQxPaO0

NGINX is well known as a simple and good web server right now, but not everyone knows that NGINX can act like a TCP-UDP loadbalancer. So you won't need to install HAProxy when you need a LoadBalancer.

This feature is released on NGINX 1.9+. So you can setup it by this rule

stream {
	upstream backend1 {
		server s1.backend1.com:12345;
		server s2.backend1.com:12345;
	}
	server {
		listen 54321;
		proxy_pass backend1;
	}
	upstream backend2 {
		server s1.backend2.com:7777;
		server s2.backend2.com:7777;
		server s3.backend2.com:7777;
	}
	server {
		listen 8888 udp; # add udp keyword if you want UDP server
		proxy_pass backend2;
	}
}

To learn more, click here : https://docs.nginx.com/nginx/admin-guide/load-balancer/tcp-udp-load-balancer/