@beberlei@phpc.social

· 2mo ago

I dug deep into our huge dataset of production PHP applications and found: In one third of PHP projects processing takes up 50% or more of the time, and I/O less than that. https://tideways.com/profiler/blog/mythbusting-php-is-rarely-the-bottleneck-data-shows-40-of-time-spent-in-php #php #performance

tideways.com

7

1

11

0

Benjamin Eberlei @beberlei@phpc.social

· 8mo ago

Spontaneous one day trip to Brussels for #fosdem #fosdem2026 to meet #PHP community, and also releasing a small part of the herd. Who is there?

phpc.social

17

6

4

0

Benjamin Eberlei @beberlei@phpc.social

· 2mo ago

Replying to

@nCrazed@fd00.space i didn't expect it to be - we tested this using a synthetic benchmark across PHP 7.4 - 8.5 and there is practically no difference in all 8.x versions: https://tideways.com/profiler/blog/php-benchmarks-8-5-vs-8-4-8-3-and-7-4

Tideways

PHP Benchmarks: 8.5 vs 8.4, 8.3 and 7.4

Find Out Why Your PHP App Is Slow Before It Costs You. Tideways shows you exactly where PHP performance breaks down - in production, in real time.

1

0

Benjamin Eberlei @beberlei@phpc.social

· 2mo ago

Replying to

@afilina@phpc.social i really liked playing it! Good mechanics

1

0

Benjamin Eberlei @beberlei@phpc.social

· 8mo ago

What if we could run a list of middlewares without the stack trace becoming deeper than 1?

After talking about crazy PHP things with @sebastian@phpc.social @nielsdos@phpc.social and @Girgias@phpc.social this mad idea came to me on the train back from #fosdem

Took 3 attempts to make it work, and its clumsy becuse you need to call yield and then $response = yield;

Or is there a better way?

Anyways, stacktrace only has depth 1 here for each Middleware::process call.

Wdyt?

https://gist.github.com/beberlei/c958c24bcea4b08013eb1126e65fefce

phpc.social

3

0

Benjamin Eberlei @beberlei@phpc.social

· 4mo ago

Replying to

@SenseException@phpc.social you need to give it this as architectural constraints. Via agents.md for example. After you have some code already it will learn from existing architecture

1

0

Benjamin Eberlei @beberlei@phpc.social

· 6mo ago

Replying to

@LeedsPHP Maybe https://eventy.io/ - its from several people in the Laravel community /cc @driesvints

eventy.io

2

0

Benjamin Eberlei @beberlei@phpc.social

· 8mo ago

Replying to

@julian@phpc.social good question, opened an issue to evaluate.

2

0

Benjamin Eberlei @beberlei@phpc.social

· 5mo ago

Replying to

@OndrejMirtes@phpc.social @bobmagicii@phpc.social @asgrim@phpc.social I eed it for typehint based parameter injection, and my "80%" rule is that it just tests all classes for a converter from left to right until it finds one.

1

2

0

Benjamin Eberlei @beberlei@phpc.social

· 5mo ago

Replying to

@OndrejMirtes@phpc.social @bobmagicii@phpc.social @asgrim@phpc.social you are right that use case is important, for PHPStan the exact type relationship and compisition is important, for me: I only need a simple list of all classes that could go in there. :)

1

0

Benjamin Eberlei @beberlei@phpc.social

· 5mo ago

Replying to

@asgrim@phpc.social no, i mean a library that simplifies ReflectionParameter::getType() because it can be either named type, union type and its super clunky to work with compared to previous ::getClass()

1

9

0

Benjamin Eberlei @beberlei@phpc.social

· 5mo ago

Do i misremember that there was a package that simplified working with ReflectionType instances?

1

13

0

Benjamin Eberlei @beberlei@phpc.social

· 8mo ago

Replying to

@heiglandreas@phpc.social Its honorable that you think this way, but the train has left the station and there is no going back to the old ways. Even if AI bubble bursts, the tech exists now and local models can still be powerful enough to completly change coding forever. Its just a different way of programming that solves a lot of problems, mainly the scarcity of available human programmer brains. Think of woodworkers witching from hand tools to electrical tools. The profession stays the same.

1

0

Benjamin Eberlei @beberlei@phpc.social

· 11mo ago

Replying to

@shochdoerfer@phpc.social Ugh just realized the sentence is misleading, the extension was written primarily by Mate, the blog post was written by Tim :)

2

1

0

Benjamin Eberlei @beberlei@phpc.social

· 8mo ago

@ocramius@mastodon.social this one? https://en.wikipedia.org/wiki/Zoropsis_spinimana

en.wikipedia.org

1

0

Benjamin Eberlei @beberlei@phpc.social

· 16mo ago

Replying to

@flowcontrol@phpc.social @Tideways@phpc.social 🤫

3

1

0

Benjamin Eberlei @beberlei@phpc.social

· 11mo ago

Replying to on phpc.social

And there is a discussion on Reddit on the topic: https://www.reddit.com/r/PHP/comments/1o4pncv/comment/nj8ihvb/

reddit.com

1

0

Benjamin Eberlei @beberlei@phpc.social

· 1w ago

Confused, did GitHub forget during their last Issue Labels screen redesign to add a button for creating a new label? I can't find any way to do that 😆

0

1

0

Benjamin Eberlei @beberlei@phpc.social

· 2mo ago

Replying to

@mattbrowndev@mastodon.social good point. What we are measuring is „main thread“ blocking time, where itis spent. So right, if you parallelize I/O then share of I/O drops. Processing is much harder (almost impossible) in PHP right now from the response time pov

0

Benjamin Eberlei @beberlei@phpc.social

· 3mo ago

Replying to

@ocramius@mastodon.social @wanwizard@mastodon.scot depends heavily on the vendor though, i charge for 46 cent at ENBW fastchargers theough a 5 euro subscription. Since i have no wallbox I charge this way at the supermarket

0

Benjamin Eberlei @beberlei@phpc.social

· 2mo ago

Replying to

@ilia@phpc.social maybe url to article got lost? curious :)

0

1

0

Benjamin Eberlei @beberlei@phpc.social

· 2mo ago

Replying to

@brendt@phpc.social Its used on the Tideways Wordpress relaunch as a custom Gutenberg block, example https://tideways.com/profiler/blog/new-in-php-8.6-narrowing-the-array_map-vs-foreach-performance-gap

tideways.com

0

1

0

Benjamin Eberlei @beberlei@phpc.social

· 2mo ago

Replying to

@pierre@phpc.social really looking forward to GD improving!!! However isnt 8.6 unlikely given RFC deadline looming or not? I think discussion must already be underway to make it.

0

1

0

Benjamin Eberlei @beberlei@phpc.social

· 2mo ago

Replying to

@pierre@phpc.social I would imagine the discussion on the API will take much longer than two weeks, as the APIs shipped in recent versions are actually decent because of the feedback loop from the community and less yolo mentality :)

0

1

0

Benjamin Eberlei @beberlei@phpc.social

· 1w ago

Replying to

@tomasnorre@phpc.social haha yes, not what I expected with prior knowledge of how that worked before. I helped myself with GitHub CLI 🤣

0

1

0