feat: Ability to import FQCNs found during analysis #7597
Conversation
Wirone
added
kind/enhancement
kind/feature
topic/types
| // We handle `extends` and `implements` with similar logic, but we need to exit the loop under different conditions. | ||
| $isExtends = $tokens[$index]->equals([T_EXTENDS]); | ||
| $index = $tokens->getNextMeaningfulToken($index); | ||
| $extend = ['content' => '', 'tokens' => []]; | ||
|
|
||
| while (true) { | ||
| if ($tokens[$index]->equalsAny([',', '{', [T_IMPLEMENTS]])) { | ||
| $this->shortenClassIfPossible($tokens, $extend, $uses, $namespaceName); | ||
|
|
||
| if ($tokens[$index]->equals('{')) { | ||
| if ($tokens[$index]->equalsAny($isExtends ? [[T_IMPLEMENTS], '{'] : ['{'])) { |
There was a problem hiding this comment.
FYI: basically it is a fix for redundant analysis that is currently executed when extends and implements are both used in a class. This method is invoked 2 times in applyFix() because both T_EXTENDS and T_IMPLEMENTS trigger it, but currently both are exiting on {, which means implements part is processed 2 times.
In the context of this particular PR, it resulted in symbols being imported twice, because e.g. implements \Other\Foo was shortened to implements Foo in the first run (for extends), and when processing implements it was registering Foo for importing. I've fixed the issue of importing non-FQNs (symbol must start with \ to be registered for importing), but this scenario allowed me to find the redundant loop.
Seems like leftover from previous versions of array shape in `ImportProcessor::insertImports()`.
If symbol does not start with `\` it means that it's already imported (as a full import or partial like `use Foo\Bar;` + `Bar\Baz` usage), or it's a global symbol in a global namespace. Either way, it's tests' and static analysis' responsibility to catch invalid symbols, we just treat them as valid ones and handle only FQCNs.
There is `GlobalNamespaceImportFixer` which handle this. Maybe in the future it could be unified, but at this point let's have single fixer for that.
Wirone
force-pushed
the
codito/fqcn-import
branch
from
c943f82
to
3472b55
Compare
|
Thank you @Wirone |
…ixer#7597)" This reverts commit 27765a6.
Fixes #6582 🎉 .