4 ms·

Something I'd really like is for PHP to somehow be stricter on the number of arguments passed to a function. As of now, PHP emits an error if arguments are mis

• MH15 2y ago
  I'd be curious to read about what percentage of active PHP devs use the recent features. The last time I worked in a PHP codebase (2020?) was half PHP 5 (bad) and half PHP 7 (much nicer). Curious if there's any real info out there on this
  • zuppy 2y ago
    php 7 has been released 9 years ago.
    • noduerme 2y ago
      Yeah, and I just finished porting an enormous amount of production code from PHP 5 to 7.x before fully moving it to 8. There are so many breaking changes in each major version, when you have a lot of live projects and clients don't have the budget to pay you to upgrade them, they can lay stagnant for years until way past EOL. It would have been nice to know, for instance, that future versions of PHP would throw warnings about undeclared variables or unaccessible named properties of "arrays" - which could previously be relied upon to be false-ish. That's a major pain point in code bases that treated arrays as simply dynamic objects that could be checked or defined at will. Lots of isset() and !empty() and other BS. Fine, but it takes time to sit down and check it all. I really preferred it when it let you just screw up or try to access a null property or define a variable inside a block and access it later without throwing any errors at all. Nothing about its actual functionality has changed in that regard; it's just errors you have to suppress or be more verbose to get around. In PHP 8 you can still do this: `if ($a) { $previouslyUndefined = 2; } if ($previouslyUndefined) { echo "yeah"; }` PHP still knows what $previouslyUndefined is or isn't at the second if statement, but it'll throw an error now in the first statement if you hadn't declared it outside the block. Why? Who cares? Scope in PHP is still understood to be inline, not in block; there is no equivalent to let vs var in JS. Stop telling me where I can check a variable if you're not enforcing various kinds of scope.
      • zerocrates 2y ago
        Your $previouslyUndefined thing as something that's changed, as far as I know, isn't true? Unless I've missed some very recent change. If $a is true, that snippet will just execute with no errors. If $a is false you'll get a warning trying to check $previouslyUndefined in the second if. That behavior's been the same for a very long time. The blocks don't matter for scope but the fact that you never executed the line that would have defined the variable does. Similarly, warnings on accessing array keys that don't exist, that's been a thing forever too. Pretty sure both go back with the same behavior to PHP 4, and probably earlier.
      • asddubs 2y ago
        >Why? Who cares? I do, if I typoed previouslyUndefined the first time. I get it adds boilerplate, but it also catches stupid bugs
  • reddalo 2y ago
    The Laravel ecosystem folks seem to be always up to date in recent PHP developments. At least, that's my impression.
    • naiv 2y ago
      Symfony also does a great job adding polyfills way ahead of a PHP release , eg https://github.com/symfony/polyfill-php84 https://github.com/symfony/polyfill-php84
      • withinboredom 2y ago
        Yes, and this is incredibly annoying. Many packages add them as a dependency, and then you get subtle bugs because of it. Or worse, they add a dependency for the polyfill that is related to an extension and suffer performance issues when the extensions are not installed; yet no warning is output.
    • chx 2y ago
      Drupal is very hot on attributes, has fiber support, uses readonly (a PHP 8.2 feature) extensively, enums are used -- overall it's fairly up to date.
  • post_below 2y ago
    PHP 5 is as close to phased out as it gets at this point. No doubt it's still in a lot of legacy enterprise codebases (lots of breaking changes going from 5 to 7 or 8), but outside of that no one is using it.
  • tored 2y ago
    I work on multiple projects from PHP 5.2 to PHP 8.3 and everything in between. Statistics based on packagist(composer) https://stitcher.io/blog/php-version-stats-january-2024 https://stitcher.io/blog/php-version-stats-january-2024 Statistics based on web servers in the wild https://w3techs.com/technologies/details/pl-php https://w3techs.com/technologies/details/pl-php https://w3techs.com/technologies/history_details/pl-php https://w3techs.com/technologies/history_details/pl-php
• withinboredom 2y ago
  I don’t really understand the issue. Already if you have a mismatch, the only way you’d ever know is through static analysis. It will run and maybe crash during run time. I always joke that changing a function signature is the single most risky thing you can do in php (especially if you have any dynamic dispatch). Making it even more risky isn’t the right answer, IMHO. Oh, and doing this would literally break class autoloading in symfony, and even the engine itself, which relies on this feature.
  • p4bl0 2y ago
    > the only way you’d ever know is through static analysis Not for builtin PHP functions which already throw errors on arity mismatch. > this would literally break class autoloading in symfony, and even the engine itself, which relies on this feature I don't understand. Could you point to where in the Symfony code it relies on being able to wrongly call a function with more arguments than it expects and will use? For variadic functions there is the ... operator already in the language since version 5.6, and my proposal wouldn't break that. Also note that builtin functions already emit deprecated warning in PHP 8 when called with too many arguments. Here is the full thread discussing it on the PHP internal mailing list: https://news-web.php.net/php.internals/122928 https://news-web.php.net/php.internals/122928
    • notresidenter 2y ago
      You should look at `func_get_args()` usage in the wild. This is sometimes used for (mostly outdated) good-enough reasons and doing this might break it?
      • p4bl0 2y ago
        I know of func_get_args, but proper variadic functions have been a thing since PHP 5.6 (released more than 10 years ago) using the ... operator. Also, my initial proposal doesn't break existing code :).
        • withinboredom 2y ago
• Cthulhu_ 2y ago
  It might be too big a change on a language level given this has been in since forever, but it might be picked up by static analysis / a linter. I'd argue it's always better to have additional protections like this in a linter as the process of adding linter rules is easier and less impactful than making a language change. It's also always preferred to not add anything to the language imo; in this case, I'd opt to have the interpreter emit a warning or info message. It's not broken, it's a developer error.
  • p4bl0 2y ago
    Indeed on the PHP internals mailing list some people were saying that it would be better to entirely deprecate passing extra arguments to a non-variadic function, without adding syntax/keyword to the language.
• Ayesh 2y ago
  FWIW, core PHP functions do throw an ArgumentCountError when passing fewer OR more parameters than the signature allows.
  • p4bl0 2y ago
    Yep! And currently the behavior is not consistent with user defined functions.
• rkharsan64 2y ago
  Somehow people seem to be missing the fact that this would've been an opt-in feature. > A way to bake that in without breaking old code would be to allow function definition to put an explicit stop to the argument list, ...
• mklepaczewski 2y ago
  I'm not a fan of the void argument syntax. Wouldn't something like the code below work? We already do it with `strict_types=1`. <?php declare(strict_args=1);
  • tored 2y ago
    Mark the function with an attribute to allow the old behavior. See link for similar use case. https://www.php.net/manual/en/class.allowdynamicproperties.php https://www.php.net/manual/en/class.allowdynamicproperties.p...
    • mklepaczewski 2y ago
      Yup, that's even better.
      • p4bl0 2y ago
        That's equivalent to my initial proposal, except that mine adds the information in the type signature of the function rather than in a decorator.
        • tored 2y ago
          If I understood your proposal correctly, to get the new behavior add an explicit stop to the function, my proposal add attribute to keep old behavior. Thus if understanding this right that would require to update every function signature to new behavior rather than marking a few functions to get the old behavior and automatically get the new better behavior of every other function for free.
          • p4bl0 2y ago
            Oh okay, you would do it in reverse. I strongly agree with that. But that means the new feature is opt-out rather than opt-in, and it may break some old code. Maybe it should be done in two-steps (opt-in + deprecation, then opt-out).
• noduerme 2y ago
  I think this would be nice ergonomically, from a coding perspective, but I'm curious as to how it would be a security threat to pass too many arguments. What's the potential exploit here?
  • p4bl0 2y ago
    Exploit I don't know, but as any stricter type verification, it would catch some bugs for sure. Note that builtin functions already throw an ArgumentCountError when passing fewer OR more parameters than the signature allows. My proposal consists in (optionally in a first place) make this behavior consistent for user-defined functions.
    • noduerme 2y ago
      The trouble for me, where the rubber meets the road, is external API calls that spread their arguments into a PHP function that takes a bunch of args. So I would love a way to detect if they're sending too many, which I don't think currently exists (?) but not at the expense of breaking the API if they actually do send too many.
• bwoebi 2y ago
  One of the main advantages of actually allowing more arguments is forward compatibility: You can, within a library, provide an additional argument to a callback without actually introducing a BC break for all users. My favorite approach would be allowing too many args on dynamic calls (closures, and function calls with dynamic name, not method calls in general) and otherwise rejecting it.
  • berkes 2y ago
    There's no need to have this in the language just to solve the case you describe. "function current_thing(id: Int, callback: Fn(Int)) {}" and, when you decide you need more you have a myriad of options to add these. From the despised "function current_real_thing(id: Ind, success_callback: Fn(Int), error_callback: Fn(Error)) {}" to "some_namespace:current_thing(...)" via "OtherClass::current_thing(...)" to "load current_thing from NewImplementationOfThing" and so on. Being strict and explicit isn't opposed to being flexible. And strictness and explicitness is most often a predicament to allow for future change, rather than hampering that future change. It's far easier to refactor, maintain, test and reason about strict and limited implementations than to do so with dynamic, runtime-magically-changing implementations.
    • bwoebi 2y ago
      I found this approach works best with languages having method overloading. For PHP it felt quite limiting, and it also requires you to have more complexity and overhead with wrapping. But I have no hard evidence at hand, only how I experienced that in PHP.