markus staab | @[email protected]

Do you need help with #PHPStan in some form? I have plenty of experience in contributing changes to PHPStan core, or implementing custom extensions. staabm.github.io/2024/01/01/php… plz RT

@jarredsumner Lets see how many people can afford the climbing prices for AI in a few months

Today, a human presses merge. Within a few months I expect: automated reviewer LGTMs -> another Claude with a fresh context window judges if it’s simple and low blast-radius -> auto-merge

🔑 New OSS from ShipMonk: Passkeys (WebAuthn) for PHP 8.4+ with zero dependencies. High-level flow API, conditional mediation, a fake authenticator for easy testing, 100% code coverage and phpstan max level with no ignores. github.com/shipmonk-rnd/p…

#phpstan 2.2.3+ is able to detect package updates based on composer.lock. utilizing the pre-existing dependency graph, we need to re-analyze only affected files. Previously we had todo a full project analysis. contributed by @sander_scode

@MarcJSchmidt Rent a cloud gaming gpu ;-)

I wanted to play Black Flag Resynced since it came out but my gaming machine's GPU is used nonstop for ML experiments/unit tests/compilers/benchmarks. Would need a second machine which my wife would not approve. help?

We have a few really cool perf experiments running atm to get #phpstan even faster.. some of them look really promising. 🫰

PHPStan Config : Set `reportIgnoresWithoutComments: true` to require targeted "@phpstan-ignore" identifiers with an explanation, while rejecting broad "@phpstan-ignore-line" and "@phpstan-ignore-next-line" shortcuts. -> discover even more phpstan configurations here: voku.github.io/PHPStanConfigu…

Its implemented in the upcoming release and will ship with a new phpdoc @pure-unless-callable-is-impure which allows you to achieve the same thing on your own code

contributed to #phpstan by @zonuexe: PHPStan will have a better idea about purity of callables taking a callback argument. for example will `array_map($pureCb, $arr)` be considered pure: - less false positives - better dead code detection

@VotrubaT Seeing php_ast extension beeing so fast, makes me wonder whether a new php extension could be implemented which is API compatibe with nikoc/php-parser but as fast as the ast extension I

Someone asked about memory consumption of Go parser... so I've added a new column to see how it runs compared to other parser ⬇️ (z7zmey-Go being the least hungry on multi cores)

@VotrubaT I tried running mago on some projects which use phpstan atm. My experience until now is, that mago is very fast in delivering a huge amount of false positives

This is my experience so far as well. PHP compilation architecture was never meant to handle processing whole codebases at once. Anyone using Mago in my circles?

@VotrubaT @wilsenhc so the idea is to built a tool which implements all rector built-in transformations and keep rector separat for the "users can build their own rules in php" use-case?

@markusstaab @wilsenhc It's not a bug, it's feature :). Goal is to make simpler, and faster tool that delivers more value, not 1:1 clone in different language. Same way FrankenPHP is not nginx, iPhone is not a homephone.

Basically, Rector-like tool written in Go is finished before Rector (in PHP) start to process a 1st file. How can it be so far? Maybe I use it wrong, I'll keep you posted 😉

@VotrubaT @wilsenhc I think mago has the same problem. You are leaving a lot behind as soon as custom extensions/rules cannot be built in php

@wilsenhc No idea yet. I'm in very early stage of testing. Considering paid products, less rules, but more practical, running in loop as standalone agent, instead of manually triggering... Open to explore.

@alexdaubois In PHPStan we already AST parse regex patterns. In case you can come up with a idea how it could work, please open a issue and describe your idea.

@markusstaab An excellent idea! I think yes. It will require some parsing but it’s pretty deterministic so my guess is that it’s totally doable 🙂

One regex can freeze your whole #PHP process. So I capped it in #Symfony latest releases. JSONPath's match() and search() run user-supplied patterns. Feed (a+)+ a long non-matching string and PCRE backtracks exponentially: classic ReDoS. I bounded pcre.backtrack_limit around those calls, so a pathological pattern fails fast instead of hanging.

PHPStan 2.2.4 addresses the RAMpocalypse! It now consumes a lot less memory. See for yourself by running "analyse -v". Running PHPStan itself used to consume 2,8 GB, now it consumes 2,1 GB. On a big real-world project, the consumption went down from 17,45 GB to 14,1 GB.

@nicksdot @Sir_Mishaa As long as mago produces less useful results the speed comparison does not make sense

@markusstaab @Sir_Mishaa Brendt posted this PHPStan vs. Mago video the other day. Maybe running it on Tempest can serve as a reproducer. x.com/brendt_gd/stat…

Mago parsing PHP 30-58 times faster than PHP Parser is actually insane. 😅

@Sir_Mishaa @nicksdot If you have a reproducer for phpstan beeing slow please report a bug

@nicksdot Could it be used to increase performance of PHPStan ? In big codebase, this quite take some time to run and would be nice if it could be improved

New release: PHPStan 2.2.3 github.com/phpstan/phpsta… #phpstan

Infection 0.34.0 is out! 🎉 It introduces a new feature for filtering source files to mutate and tests to run. Learn why it matters and how to use it: infection.github.io/guide/argument…

just reviewing a #phpstan PR which fixes a old feature request with a 3 digit issue number... nearly 10 years old. seems we can consider it implemented in the near future =)