Phpstan parser #12
Phpstan parser #12
Conversation
Current run on deptrac itself finds new dependencies not found with the old method when calling `parent::method()` by resolving the `parent` to the actual class. There is 10_000 things surely wrong with the implementation that needs fixing in the next passes, there are no tests, BUT, it works. Big thanks to Tomas Votruba for Rector and his blog as it allowed me by example to integrate PHPStan DIC to deptrac.
…ing and leave creating references to dedicated extractors. As a consequence the FileReferenceVisitor code is way more readable now.
Allows for much cleaner extractor implementations. Similar to how PHPStan custom rules work. Also allows us to call extractors only when they apply to the node in question.
Fix: composer-dependency-analyser: improve config, update * Fix filepath * 1.4.0
…olving issues with discovering symbols.
patrickkusebauch
force-pushed
the
feature/phpstan-parser
branch
from
f7d1339 to
ea83b1e
Compare
…covering symbols via PHPStan. This seems to be working?
|
I am ready to tentatively declare this PR as ready for review. There are still some small things missing (see main comment), but the business logic is implemented. |
| 'Foo\Bar:30 on Foo\SomeClass', | ||
| 'Foo\Bar:32 on Foo\SomeClass', | ||
| 'Foo\Bar:36 on Foo\string2', | ||
| 'Foo\Bar:38 on string', // todo: This is a bug in the parser for FQ "\string" |
There was a problem hiding this comment.
Here we have an undiscovered bug where \string is treated as class string instead of the scalar type string.
Resolve Psalm issues.
|
I am at a loss with regards to how to scope PhpStan. Any help would be appreciated. |
Right now I can not help with this, but I have seen that rector does not scoped phpstan. |
As far as I have discerned, Rector does things differently than we do. It does not scope itself, It only scopes its dependencies. And then from those it excludes stuff like PHPStan and PHPParser. Furthermore, Rector is not distributed as a PHAR, only as a composer dependency, so there is not one PHIVE installable file to download. We can look into going that route, but it will change how we distribute Deptrac. |
|
@patrickkusebauch Is the main issue with rescoping the PHPStan the composer autoloader configuration or have you encountered something else? Wild idea: what about releasing the PHPStan parser variant as a deptrac "extension"? This would mean that the default distribution is not affected, the design of deptrac is still correctly modular, yet we can experiment with the benefits of PHPStan parser. Such extension would include dependency on phpstan, and to use it, you would have to reconfigure a few services in the configuration file. |
|
@olsavmic it is strictly with rescoping PHPStan and getting it to work with Deptrac's own scoping. Thinking about an extension - I might work. I would require to still merge some of the refactoring done in this PR as well as Deptrac exposing a lot of classes that are currently internal into the |
The current run on deptrac itself finds new dependencies not found with the old method when calling parent::method() by resolving the parent to the actual class.
Big thanks to Tomas Votruba for Rector and his blog as it allowed me by example to integrate PHPStan DIC to deptrac.
While it does not add much on its own, I see how this might allow us to resolve many open issues in the future, like the oldest open issue qossmic/deptrac#449.
Missing: