- PHPStorm supports
Generator|Foo[]syntax, which phan doesn't because it usesGenerator<Foo>- so you can use@return Generator|Foo[]and@phan-return Generator<Foo>, but that's another type of repetition - You could suppress PHPStorm's warnings, but that doesn't work well on projects/files that don't have other static analyzers set up and phpstorm may have its own checks you'd wish to keep using. I forget how that's done
- I'm not sure what PHPStorm's plans are for templates in general, and haven't seen anything in their announcements. I don't work on it and primarily use Vim.
Generator|Foo[] because that syntax implies that a native array of Foo could possibly be returned. At least I see all alternatives and didn't think about combining @return and @phan-return at the same time! Thanks again!
also saw an error i don't understand:
PhanTypeInvalidRightOperandOfBitwiseOp Invalid operator: right operand of & is bool (expected int|string)
the line in question is like $x & $key, where $key is an array key (so int|string) and is also asserted to be an int. any idea why phan thinks it's bool?
php-ast version 1.0.6
PHP version used to run Phan: 7.4.3
type[][] if all keys are unknown, there's also more elaborate types such as list<array<string, type>>
assert(is_int($b)) before the op and it still errors.
phan ±0198a2479⚡ » php internal/dump_fallback_ast.php --php-ast-native '$x = ($a & $b === $a);'
AST_STMT_LIST [] #1
0 => AST_ASSIGN [] #1
var => AST_VAR [] #1
name => x
expr => AST_BINARY_OP [BINARY_BITWISE_AND] #1
left => AST_VAR [] #1
name => a
right => AST_BINARY_OP [BINARY_IS_IDENTICAL] #1
left => AST_VAR [] #1
name => b
right => AST_VAR [] #1
name => a$x = ($a & $b) === $a;
phan 2.7.3 is emittingPhanEmptyFQSENInClasslike with code like
class C {
/** @var string Fully Qualified Classname. */
const FQCN = '';
public function f() {
if (! empty(static::FQCN)) {
$fqcn = static::FQCN;
new $fqcn();
}
}
}shouldn't it be obvious to phan that $fqcn is a non-empty string?
Currently, no - in abstract classes declaring a class constant with /** @phan-abstract */ public const X = ''; would help detect classes that didn't override the constant, but that doesn't help in your use case because traits can't have class constants.
- I think you might be able to do something with
@mixinfor a stub interface (e.g. .phan/stubs) with a abstract class constant but I don't know if that even works now and wouldn't rely on it even if it did work. - Phan only analyzes trait methods on the declaring trait right now, and there's no way to suppress something only on the declaring context
There were discussions of allowing traits to have class constants or require implementing interfaces in php itself in some future php version
IteratorAggregate as being a kind of Iterator<T> for a specific T? For example, in the following code, I’d like Phan to emit a deprecation error on lines 19 and 26, but it only does so on line 26.<?php
class X {
/** @deprecated */
public $x;
}
class Y implements IteratorAggregate {
/** @var list<X> */
public $xs;
/** @return Iterator<X> */
function getIterator() {
return new ArrayIterator($this->xs);
}
}
/** @param Y $y */
function f1($y) {
foreach ($y as $x) {
echo $x->x;
}
}
/** @param Y|Iterator<X> $y */
function f2($y) {
foreach ($y as $x) {
echo $x->x;
}
}Hello! I have a question about
input:22: PhanTypeNoPropertiesForeach Class \ActiveRecord was passed to foreach, but it does not extend Traversable and doesn't have any declared properties. (This check excludes dynamic properties)In this snippet
If I remove ActiveRecord from union type on line 5, why Phan doesn't warn about passing null|int|string|bool to foreach? Why only non-traversable object instance bothers it?
As you already guessed, in real code return type of that find() method depends on its arguments. Can I help Phan to define type in each case more precisely except of add exclusive @phan-var for each call of find()?
PhanTypeMismatchForeach detects the case where there are no valid types, but if some of the types are valid, it doesn't warn. I don't believe that --strict-type-checking or similar options currently cover array|null - there's probably hundreds of issues for which a warning about being partially invalid could start being emitted
https://github.com/phan/phan#readme
Phan is a static analyzer for PHP that prefers to minimize false-positives.
target_php_version? Faced with overlooked one like this for PHP 5.6. It worth to mention that PHPStorm also doesn't show it while Netbeans IDE does. And this code become valid in PHP 7+
https://github.com/phan/phan/wiki/Incrementally-Strengthening-Analysis#just-backward-compatibility
If you are still using versions of php older than 5.6, PHP53CompatibilityPlugin may be worth looking into if you are not running syntax checks for php 5.3 through another method such as InvokePHPNativeSyntaxCheckPlugin or phan --native-syntax-check php53 (see .phan/plugins/README.md).
For issues that php5.6 --syntax-check would check, I'd recommend using php5.6 --syntax-check in addition to Phan,, whether internally or externally - it's redundant to detect the hundreds of issue types that the syntax checks would already check for
phan --native-syntax-check /path/to/php53 assuming php5.6 is a binary in your PATH
Is Phan intended to detect PHP Parse errors corresponding its target_php_version?
It doesn't - it catches some of the parse/compile errors, in addition to incompatibilities (e.g. usingstringwould mean a class named string in php 5.6 and not be detected by php56 --syntax-check)
Also,
If you are migrating from PHP 5 to PHP 7, you should also look into using php7cc (no longer maintained) and php7mar, which have different backwards compatibility checks.
Hey there, I had a quick question if anyone had any output. Sorry if this is the wrong place.
I'm using Mockery with phan and getting a PhanTypeMismatchArgument similar to this issue. The issue was asking specifically about creating a plugin, but I wanted to ask if there was any specific way people are handling this right now aside from suppressing it.
I don't want to open an issue because it's already known but I also figured more people would have ran into this and talked about potential workarounds that work with Phan 4
