Multi Line Generic Types are invalid

[ServerExe]

Bug report

PhpStan-Version: 0.12.19

The following error occurs when trying to declare a multi lined Generics annotation:

PHPDoc tag @implements has invalid value (HelloInterface< Foo, Bar): Unexpected token "\n *", expected type at offset 34

Code snippet that reproduces the problem

https://phpstan.org/r/80a149d9-5820-4041-bf3b-4b6e19de6b09

Expected output

No errors! as it is in:
https://phpstan.org/r/08857dad-5e7b-4615-9c99-f3397aa204ce

[ondrejmirtes]

Yes, this isn't currently supported. Moving to phpdoc-parser repo if someone wanted to implement that.

[icanhazstring]

@ondrejmirtes thanks for moving. Can you point ppl in the right direction, like the file where the parsing for this takes place. So maybe someone can get startet right away :)

[ondrejmirtes]

See how multiline array shapes are implemented:

phpdoc-parser/src/Parser/TypeParser.php

Lines 282 to 304 in d4c4cd3

	private function parseArrayShape(TokenIterator $tokens, Ast\Type\TypeNode $type): Ast\Type\ArrayShapeNode
	{
	$tokens->consumeTokenType(Lexer::TOKEN_OPEN_CURLY_BRACKET);
	$tokens->tryConsumeTokenType(Lexer::TOKEN_PHPDOC_EOL);
	$items = [$this->parseArrayShapeItem($tokens)];
	$tokens->tryConsumeTokenType(Lexer::TOKEN_PHPDOC_EOL);
	while ($tokens->tryConsumeTokenType(Lexer::TOKEN_COMMA)) {
	$tokens->tryConsumeTokenType(Lexer::TOKEN_PHPDOC_EOL);
	if ($tokens->tryConsumeTokenType(Lexer::TOKEN_CLOSE_CURLY_BRACKET)) {
	// trailing comma case
	return new Ast\Type\ArrayShapeNode($items);
	}
	$items[] = $this->parseArrayShapeItem($tokens);
	$tokens->tryConsumeTokenType(Lexer::TOKEN_PHPDOC_EOL);
	}
	$tokens->tryConsumeTokenType(Lexer::TOKEN_PHPDOC_EOL);
	$tokens->consumeTokenType(Lexer::TOKEN_CLOSE_CURLY_BRACKET);
	return new Ast\Type\ArrayShapeNode($items);
	}

This is where generic types are parsed:

phpdoc-parser/src/Parser/TypeParser.php

Lines 165 to 176 in d4c4cd3

	public function parseGeneric(TokenIterator $tokens, Ast\Type\IdentifierTypeNode $baseType): Ast\Type\GenericTypeNode
	{
	$tokens->consumeTokenType(Lexer::TOKEN_OPEN_ANGLE_BRACKET);
	$genericTypes = [$this->parse($tokens)];
	while ($tokens->tryConsumeTokenType(Lexer::TOKEN_COMMA)) {
	$genericTypes[] = $this->parse($tokens);
	}
	$tokens->consumeTokenType(Lexer::TOKEN_CLOSE_ANGLE_BRACKET);
	return new Ast\Type\GenericTypeNode($baseType, $genericTypes);
	}

[icanhazstring]

Will try to tackle this one :)

[icanhazstring]

@ondrejmirtes would it be wise to update the Tokenzier to avoid tokenizing the \n as a token itself.
The only thing "breaking" the code right now is the "new_line" token. With this removed, everything should work fine.

[icanhazstring]

Nevermind, just found out how the parsing works exactly :)

[icanhazstring]

This can be closed then @ondrejmirtes :)

[ondrejmirtes]

Yes, implemented in #44 :)

[github-actions]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.