PHP 8.2.4
php.net
php.net
What kind of things are you having problems with in regards to backwards compatibility?
You know they'll pull the same crap just as fast next time.
Some applications might be easy to upgrade. Others will not. You are forced to do work, because for some minority it feels nicer when they get to use that new feature. And it has to look a certain way, no matter the cost! Some of the stuff that’s comming will be even worse.
The even bigger issue is that your underlying tool breaks, which has an impact on dependecies as well. Foo needs version N now, but you also depend on Bar which only works with N-1.
It’s not fun, and it’s expensive.
You are under no obligation to upgrade your php version as soon as released without being able to plan for it. PHP 8.1.17 was released alongside 8.2.4 for exemple.
However, I had to migrate some 30k LOC app from Dotnet 3 to 6 this year and it was quite straight forward. Not that many thinks broke, and when then they were in the bootstrapping of the app.
I'm not that much into PHP anymore, so my take might be wrong. I wouldn't like to upgrade each year to a new version which breaks syntax. I would be ok to migrate every 3 or 4 years from LTS to LTS.
To use the latest PHP 8.2 version Debian or Ubuntu users need to add an external package source, meant specifically to be "on the edge" for PHP, deb.sury.org
But GP comment is not that, it reeks of "old man yells at cloud" with zero mention of what is an issue and what isn't.
5.x -> 7+ was a pain, mainly down to those projects being without composer or other really suspect practices that need fixing at the same time. Much of the actual issue was due to particular string concatenation methods or the mysql_* lib which is easily shimmed to something like PDO.
> 5.x -> 7+ was a pain, mainly down to those projects being without composer or other really suspect practices that need fixing at the same time. Much of the actual issue was due to particular string concatenation methods or the mysql_* lib which is easily shimmed to something like PDO.
That's what I meant in my other comment, while it's true, mysql_ was already being removed in favor or PDO in PHP 5's time, and PHP 5 to PHP 7 was almost a decade of developpement, so I don't think it's fair to put these issues on "too fast a developpement cycle".
The one time you actually may have had to work on things was with PHP 7, but that's because they finally removed things that had been deprecated for years, some for over a decade, and all of them were security risks or massive code smell anyway.
The one thing with PHP 8 was the removal of dynamic propery, and again that's a massive code smell that was holding the language back. Any code still using it when PHP 8 released was either written over a decade ago, or something that has way more issues than that with upgrading.
The current solution for code reuse (use longer class names to avoid clashes, we help you with syntactic sugar called "namespaces") is the wrong approach imo.
I'm curious as to why you consider autoloading over PSR-4 namespaces as insufficient.
https://www.php.net/manual/en/language.namespaces.importing....
At least in the PHP I'm familiar with, nothing "happens" at `use` time; the first time you call a method on a class it autoloads. The autoloader is then what does the work of finding and requiring the file. No class, no autoloader.
As a maker, I prefer a clean stack. So that its complexity does not slow me down. Workarounds make development slower and slower, the more it becomes a giant hairball of workarounds and workarounds around workarounds.
In Python and JavaScript, I don't need those workarounds. Would be awesome if PHP catches up.
If two packages I require depend on incompatible versions of some third package, there aren't any easy ways for me to resolve it.
This problem exists because of the global nature of PHP namespaces.
A module system and less reliance on PSR the better. Why exactly does PSR still exist?
That's my problem with it. It should be focused on what's good for developers. Not what makes it easier to overengineer the next flavour of the month.
I will readily admit that 99% of my assertions are some variation of `$this->assertEquals()`
You can in theory create modules with require.
module.php
<?php
return (function(){
$foo = function(){echo "hello from PHP";};
return [
"greet"=>$foo
];
})();
<?php
$module = require('./module.php');
$module["greet"]();
Your solution works, but unfortunately not something that has any community support. I proposed something similar a year ago but use classes instead because classes seems to be the internal building block in PHP.
For both our solutions is would relatively easy to add a shim to older PHP versions.
module.php
<?php
return new class {
public const MY_CONST = 47;
private const MY_PRIVATE_CONST = 127;
// with PHP 8.1 you also have readonly properties
public string $myPublic = 'foo';
private int $myPrivate = 17;
public function helloWorld(string $name): void
{
echo "Hello, World {$name}!";
}
};
<?php
final class ModuleException extends Exception {
}
function module(string $name): object
{
static $modules = [];
if (isset($modules[$name])) {
return $modules[$name];
}
try {
$module = @require $name;
} catch (Error $error) {
throw new ModuleException("No such module {$name}", 0, $error);
}
if (!is_object($module)) {
throw new ModuleException("Module {$name} is not an object");
}
$modules[$name] = $module;
return $module;
}
$my = module('module.php');
$my->helloWorld('Module system');
$other = module('module.php');
assert($my === $other);
PHP dependencies are so intertwined that for projects that abuse and use more than a reasonable amount (aka their ability to invest time keeping them and the code using it up to date) the update process can become a nightmare of complexity.
Node's module system let you move each dependency at it's own pace. This is such a breath!
The problem is that when you only know of the solution of your ecosystem, it's hard to see the added value you miss out.
If you use a framework then try to write all your own code outside of the framework and pass in all needed framework data as arguments to your own classes/functions.
However, what I was speaking about is the version handling of (sometimes direct but mostly) indirect dependencies.
The domain of `composer why` and `composer why-not`.
One if the most speaking example I can think of is: if two libs depends on guzzle 5, then you can't update only one on a version that depends on guzzle 6: you must upgrade both at the same time. Each to an intercompatible lock on guzzle version.
All in all, super positive.