2.1.34 #13976
Replies: 20 comments 28 replies
-
|
i'm seeing a 25% decrase time from 33s . Great work |
Beta Was this translation helpful? Give feedback.
-
|
19m58s -> 15m34s (22% improvment) Thanks! |
Beta Was this translation helpful? Give feedback.
-
|
Before: After: The difference is 54 seconds. Thank you so much! |
Beta Was this translation helpful? Give feedback.
-
|
261.09s → 187.54s (-73.55s) |
Beta Was this translation helpful? Give feedback.
-
|
537s -> 472s (13%), not bad! @ondrejmirtes But if there are so many bigger changes, why not adjust the version number appropriately? When I just see a patch version number change, such a massive changelog is very surprising - and just makes planning updates with phpstan more difficult. |
Beta Was this translation helpful? Give feedback.
-
Please elaborate on that. I outlined the versioning strategy here: #12696 |
Beta Was this translation helpful? Give feedback.
-
|
Before v2.1.31
After v2.1.34
so unfortunately not really an improvement for us at https://github.com/shopware/shopware 🫣 |
Beta Was this translation helpful? Give feedback.
-
|
That usually points to garbage collector problems. Did you experience that even on PHPStan 2.x? |
Beta Was this translation helpful? Give feedback.
-
|
yes, currently we are on |
Beta Was this translation helpful? Give feedback.
-
|
results with Before v2.1.31 Elapsed time: 2 minutes 13 seconds After v2.1.34 Elapsed time: 2 minutes 9 seconds |
Beta Was this translation helpful? Give feedback.
-
|
@mitelg in case you find single files to be extra slow, please try to isolate a reproducer and report it as a new issue. would love looking into such cases (or in case its at least reproducible somehow, please leave instructions in a new issue) |
Beta Was this translation helpful? Give feedback.
-
|
@staabm thank you very much for your offer, much appreciated! and as you are here: thanks for your amazing work here and on other projects! ❤️ but as already mentioned above the behaviour seems a bit random to me 😅 I can ensure you, I would have created an issue, if it was reproducible consistently without checking whole shopware repo 😁 |
Beta Was this translation helpful? Give feedback.
-
|
I kinda wish that "Major new features" would not be added to a patch-level release, that's a bit of a violation of semantic versioning. I'm getting segmentation faults with this version and some of my custom rules no longer work. I'll try to investigate what's happening. |
Beta Was this translation helpful? Give feedback.
-
|
There's no practical difference in releasing this as 2.1.34 as opposed to 2.2.0. Everyone uses " I'm looking forward to your bug reports, I'll make sure to fix problems introduced in 2.1.34 fast. If some of your custom rules broke, it's probably because of brittle assumptions, like the order of in which they are invoked. |
Beta Was this translation helpful? Give feedback.
-
|
Well, there is a practical difference for me: It no longer works. For me the version number is also communicating the risk of upgrading. If it's "1.2.3 -> 1.2.4" I will think that it's only bug fixes and the risk of upgrading is minimal. But if it's "1.2.3 -> 1.3.0" then I know I better check the release notes and hope that all changes are compatible, but sometimes they just aren't, because nobody is perfect. I don't think major internal refactoring should go into a patch release, because the risk of it breaking things is too high. I'm still analyzing the issue, but found that it only happens on PHP 8, not PHP 7. Anyway, it's just my two cents. As always I appreciate your work and it's amazing to see how you have changed the entire PHP landscape with PHPStan! ❤️ I also just realize it's been 10 years now! 🥳 (btw, I avoid ^ as much as possible, because too many packages keep breaking things on a regular basis. I just run "composer outdated" very regularly. For some packages (notoriously Rector) I even stopped using ~ and use a fixed version instead.) |
Beta Was this translation helpful? Give feedback.
-
|
I know at least a few projects using ~ for Phpstan specifically not to get affect by a major new feature without planning it. |
Beta Was this translation helpful? Give feedback.
-
|
The recommended approach is outlined here: https://phpstan.org/user-guide/backward-compatibility-promise#minimizing-impact-of-type-inference-changes TL; DR: Caret range with a composer.lock file, or a specific PHPStan version (without range) without a lock file. If you want a super-stable build, both ~ and ^ pose similar risks of something breaking. |
Beta Was this translation helpful? Give feedback.
-
|
Before v2.1.33 After v2.1.34 |
Beta Was this translation helpful? Give feedback.
-
|
Our time went from 44s down to 39s (12% difference). |
Beta Was this translation helpful? Give feedback.
-
|
We see ~13% time improvement, at a cost of ~17% more memory usage.
|
Beta Was this translation helpful? Give feedback.
-
@billypoke IIRC depending on the used composer version, using a composer script might be less efficient. |
Beta Was this translation helpful? Give feedback.
-
|
@staabm This was using the latest composer version Here are the numbers for a cli run. I also re-ran it with the composer script since I can't remember exactly which programs were running when I ran the initial tests:
Looks like there is a small time penalty of ~2-3% (for a script lasting ~35 seconds) associated with the composer script. We use this approach so our developers can run phpstan and other tools (phpunit, rector, pint, etc.) with the same settings as the CI/CD runner will. |
Beta Was this translation helpful? Give feedback.
-
|
As long as all your devs are using a recent composer version it should be fine. IIRC it was fixed recently |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for the heads up. Appreciate all the hard work on this release 🚀 |
Beta Was this translation helpful? Give feedback.
-
|
For our project, we see the following results:
Logs:
|
Beta Was this translation helpful? Give feedback.
-
|
-33% on our project too! (measured on Github CI)
|
Beta Was this translation helpful? Give feedback.
-
|
2.1.33 2.1.34 Walltime: (Project also uses shipmonk/dead-code-detector extension) |
Beta Was this translation helpful? Give feedback.
-
|
In the FreshRSS project, before/after (on a slow computer), ~22% improvement 🏎 $ rm -fr /tmp/phpstan/
$ time composer run-script phpstan
> phpstan analyse --memory-limit 512M .
Note: Using configuration file /home/alex/GitHub/FreshRSS/phpstan.dist.neon.
real 1m2.924s
user 4m54.812s
sys 0m16.494s
$ rm -fr /tmp/phpstan/
$ time composer run-script phpstan
> phpstan analyse --memory-limit 512M .
Note: Using configuration file /home/alex/GitHub/FreshRSS/phpstan.dist.neon.
real 0m49.215s
user 3m40.906s
sys 0m18.413s |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
I'm using KnpLabs/DoctrineBehaviors#710 for the template/generics annotations. |
Beta Was this translation helpful? Give feedback.
-
|
I can't reproduce it on the playground 😢 |
Beta Was this translation helpful? Give feedback.
-
|
Small reproducing repository then please. |
Beta Was this translation helpful? Give feedback.
-
|
I am sure it is related to my poor knowledge of generics and templates, but here it is: |
Beta Was this translation helpful? Give feedback.
-
|
I had a look at #14003 which suggested https://phpstan.org/r/0d11d8fb-b844-45c7-b050-b5a685e0b1ac but unfortunately it is a bit simpler than my setup. It uses the same approach though, so I do not understand where I went wrong. |
Beta Was this translation helpful? Give feedback.
-
|
A lot improvements since 2.1.34, a big thanks!! before: after: ~ 31% faster! |
Beta Was this translation helpful? Give feedback.
-
|
I got ~24% improvement (2.1.33 vs 2.1.36) with 30 workers. Thank you! 2.1.33 2.1.36 I also compared baselines (~14k errors) and there's not much difference (good). |
Beta Was this translation helpful? Give feedback.
-
|
Impressive ~30% performance improvement, reducing time from |
Beta Was this translation helpful? Give feedback.
-
|
Thanks, this is awesome!
With 2.1.33 output2.1.37 output |
Beta Was this translation helpful? Give feedback.
-
|
Here's our results for approx. 4,410 files (forceably clearing all caches with TLDR
v2.1.33$ time sail bin phpstan -v
Note: Using configuration file /var/www/html/phpstan.neon.
4409/4409 [▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓] 100% 1 min
// ...
Elapsed time: 1 minute 24 seconds
Used memory: 6.14 GB
________________________________________________________
Executed in 85.12 secs fish external
usr time 228.59 millis 0.00 micros 228.59 millis
sys time 151.13 millis 770.00 micros 150.36 millisv2.1.37$ time sail bin phpstan -v
Note: Using configuration file /var/www/html/phpstan.neon.
4420/4420 [▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓] 100% 50 secs
// ...
Elapsed time: 52.48 seconds
Used memory: 6.99 GB
________________________________________________________
Executed in 53.14 secs fish external
usr time 254.43 millis 891.00 micros 253.54 millis
sys time 127.74 millis 278.00 micros 127.47 millisKeep up the good work! |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Major new features 🚀
doFoo($a = 1, $a)will see better type inference for 2nd call argument.$scope->getType($node->args[1]), without Fibers it getsmixed, because$scopesees the state before the method call.$node->args[1]is available in the analyser.$scope->getType($node->args[1])sees1as the type, notmixed.Improvements 🔧
array_count_valuesreturned value is not taking care of the key-type #13816, thanks @mad-briller!Bugfixes 🐛
UnionType::getObjectClassReflections(#4541), When one class mentioned by@phpstan-sealedis unknown all classes are incorrectly reputed unknown #13777, thanks @VincentLanglet!count(non-empty-array, COUNT_RECURSIVE)isint<1, max>(#4515), thanks @staabm!range()function to always yield anon-empty-list(#4688), thanks @FeBe95!@phpstan-sealedleads to wrong errors being reported when used on a non-abstract class #13734, thanks @stof!accepts()methods (#4766), non-empty-array&hasOffset has unexpected type compatiiblity #13964, thanks @staabm!Performance 🏎️
@template TKey as array-key#4670) (#4670), thanks @staabm!TypeCombinator::union()calls that are not necessary (phpstan/phpstan-src@12f2b23)TypeCombinator::intersect()calls that are not necessary (phpstan/phpstan-src@f1f97c3)NodeScopeResolver::processAssignVar()(#4716), thanks @staabm!StaticType::transformStaticTypeless often (phpstan/phpstan-src@d27df94)CallableType::equals()(phpstan/phpstan-src@1f1d382)Function signature fixes 🤖
Internals 🔍
self::class#4713) (#4713), thanks @staabm!This discussion was created from the release 2.1.34.
Beta Was this translation helpful? Give feedback.
All reactions