Don't stop PSR-4 service discovery if a parent class is missing #25932
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Member
derrabus
commented
Jan 25, 2018
•
derrabus
commented
| Q | A |
|---|---|
| Branch? | 3.4 |
| Bug fix? | yes |
| New feature? | no |
| BC breaks? | no |
| Deprecations? | no |
| Tests pass? | yes |
| Fixed tickets | #25929 |
| License | MIT |
| Doc PR | N/A |
derrabus
force-pushed
the
di-psr4-missing-parent
branch
from
0efe70e
to
d0f53e3
Compare
weaverryan
requested changes
Jan 26, 2018
| try { | ||
| $r = $this->container->getReflectionClass($class); | ||
| } catch (\ReflectionException $e) { | ||
| $classes[$class] = $e->getMessage(); |
There was a problem hiding this comment.
It's a little awkward to return an array with the error as the key, but since this method is private, it seems safe.
There was a problem hiding this comment.
Actually, the class is the key here. I agree that the structure I'm using now is a bit more awkward than the old one, but as you said, the method is private, so it's a purely internal data structure.
That being said, if you have a suggestion regarding a better data structure, I'm all ears. 😃
| foreach ($classes as $class) { | ||
| foreach ($classes as $class => $errorMessage) { | ||
| if (null !== $errorMessage) { | ||
| $definition = new Definition($class); |
There was a problem hiding this comment.
This is a nit-pick, but what happens of $class is actually an alias? I imagine it would make no difference: if that Definition is ultimately used (likely due to autowiring), it would throw the error.
There was a problem hiding this comment.
I don't expect this case to be any different than with an alias to a non-broken class. But I'll test it.
| foreach ($classes as $class => $errorMessage) { | ||
| if (null !== $errorMessage) { | ||
| $definition = new Definition($class); | ||
| $definition->addError($errorMessage); |
There was a problem hiding this comment.
We need to think about this error message holistically :). Currently, the user will see:
Class App\Foo\MissingParentClass not found
I think we need to add more... as much as possible really. Something like:
Error when loading registering services from resource "../src/*": an error was thrown when processing the class "App\Foo\ActualClass": "Class App\Foo\MissingParentClass not found".
I would LOVE to even say the line number in YAML or XML... but we definitely do not know that :p (pipe dream for someday?)
There was a problem hiding this comment.
👍 This code was more or less meant as a POC. I'll see what I can do about improving the DX.
|
Well-done @derrabus! The question is: feature or bug-fix? It's a safe change: I mean, anyone with bad classes is currently getting an error... and this just makes it not an error. I vote bug fix - we'll see what other people think. |
|
Agree on it being a bugfix. It doesn't introduce a new public-facing API, it just fixes incorrect behavior. |
| @@ -60,7 +60,16 @@ public function registerClasses(Definition $prototype, $namespace, $resource, $e | |||
| $interfaces = array(); | |||
| $singlyImplemented = array(); | |||
|
|
|||
| foreach ($classes as $class) { | |||
| foreach ($classes as $class => $errorMessage) { | |||
| if (null !== $errorMessage) { | |||
There was a problem hiding this comment.
what about putting this "if" just after the existing $this->setDefinition($class, unserialize($serializedPrototype)); line?
no need to explicitly create a new definition, the prototype-based one is better, isn't it?
There was a problem hiding this comment.
I agree about the prototype definition. About moving this block down: I'd like to avoid the interface_exists() call below, so the block should probably stay here.
There was a problem hiding this comment.
I'd like to avoid the interface_exists() call below
is there any reason for that? I'm not sure this is a valid concern :)
There was a problem hiding this comment.
I expected the interface_exists call to trigger the autoloader again, but appearently it doesn't. I'm moving the block.
There was a problem hiding this comment.
The call uses false as second argument, precisely to avoid the autoloading call.
There was a problem hiding this comment.
Exactly. Block moved. ✅
derrabus
force-pushed
the
di-psr4-missing-parent
branch
from
d0f53e3
to
08e17eb
Compare
nicolas-grekas
approved these changes
Jan 26, 2018
|
@nicolas-grekas, @stof: Do you agree with @weaverryan on turning this PR into a bugfix? I'd try to apply my patch against 3.4 then. |
|
I do, go for 3.4 on my side |
derrabus
force-pushed
the
di-psr4-missing-parent
branch
from
08e17eb
to
12d93fc
Compare
derrabus
force-pushed
the
di-psr4-missing-parent
branch
from
12d93fc
to
bc34213
Compare
|
I've added some context to the error message extracted from the exception. Please have a look @weaverryan. |
|
I've set up a
When I link the project to the code of this PR, the console is executed without failure. Now, if I add a public alias to the service discovered from my broken class, the console fails again.
|
derrabus
force-pushed
the
di-psr4-missing-parent
branch
from
bc34213
to
ba512d9
Compare
|
The AppVeyor failure looks unrelated. |
dunglas
approved these changes
Jan 26, 2018
weaverryan
reviewed
Jan 26, 2018
| $r = $this->container->getReflectionClass($class); | ||
| } catch (\ReflectionException $e) { | ||
| $classes[$class] = sprintf( | ||
| 'While discovering services from namespace "%s", an error was thrown when processing the class "%s": %s', |
There was a problem hiding this comment.
nit pick :). I think we usually try to end errors with a period. So the end of this could be:
... processing the class "%s": "%s".
derrabus
force-pushed
the
di-psr4-missing-parent
branch
from
ba512d9
to
3d6c3ba
Compare
chalasr
approved these changes
Jan 27, 2018
fabpot
approved these changes
Jan 29, 2018
|
Thank you @derrabus. |
fabpot
added a commit
that referenced
this pull request
Jan 29, 2018
…ssing (derrabus) This PR was merged into the 3.4 branch. Discussion ---------- Don't stop PSR-4 service discovery if a parent class is missing | Q | A | ------------- | --- | Branch? | 3.4 | Bug fix? | yes | New feature? | no | BC breaks? | no | Deprecations? | no | Tests pass? | yes | Fixed tickets | #25929 | License | MIT | Doc PR | N/A Commits ------- 3d6c3ba Don't stop PSR-4 service discovery if a parent class is missing.
Merged
Merged
Merged
|
not sure if it's related to this, however I'm not able to run "composer update --no-dev" the output: |
|
You're looking for #41673 |
|
thanks @nicolas-grekas . I'll try not to do this again |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.