Implement named arguments for NEW DTO syntax

[Zuruuh]

I love using DTOs but it can get pretty annoying to understand what's going on when using a ton of arguments, having to keep track of what is what, in what order, etc... Using named arguments should really help a lot with that I think 👍
Small note but I also made it that trailing commas don't crash the lexer which is a nice qol feature to have imo. I can remove it / move it to another PR if needed 👌

Now the following DQL works as expected

SELECT NEW CustomerDTO(
  email: e.email,
  name: u.name,
  city: a.address,
) FROM ...

Closes #9182

[derrabus]

Thank you. Your new tests are failing, however.

[Zuruuh]

yeah you're right, I only tested locally with a php version above 8 (and it worked fine), but it seems like ReflectionClass::newInstanceArgs didn't work with an associative array before the introduction of named params in php, so it will require a bit more work to handle php 7.x versions

[Zuruuh]

Tested locally with php 7.1, tests should pass fine now 👍

[greg0ire]

This probably needs to be updated as well:

orm/docs/en/reference/dql-doctrine-query-language.rst

Lines 1690 to 1691 in 2b91edc

	NewObjectExpression ::= "NEW" AbstractSchemaName "(" NewObjectArg {"," NewObjectArg}* ")"
	NewObjectArg ::= ScalarExpression | "(" Subselect ")"

[greg0ire]

The Parser is neither final nor internal. This means this is a breaking change for extending classes. You can obtain the new argument with func_get_args() without changing the signature.

Suggested change

	public function NewObjectArg(bool $namedArgAlreadyParsed = false)
	public function NewObjectArg(/* bool $namedArgAlreadyParsed = false */)

[derrabus]

Suggested change

	The ``NEW`` operator also supports named arguments, similarly to php 8.0 :
	The ``NEW`` operator also supports named arguments:

[derrabus]

Suggested change

	Note that you cannot pass unnamed arguments after named ones, just like in php.
	Note that you must not pass ordered arguments after named ones.

[derrabus]

Suggested change

	$unnamedArgs = [];
	$orderedArgs = [];

[derrabus]

Suggested change

	/** @var ?string */
	/** @var string|null */

[derrabus]

Do we actually need this parameter? Couldn't NewObjectExpression() just check if the kind of argument it receives is supported at its position?

[Zuruuh]

I think this should work (but idk how to feel abt it): when parsing a NewObjectExpression, we parse the current one, consume the tokens, then parse the next one (if there is any) without consuming it, and throw if it's not named while the current one is. It would work without the need of the parameter, but i feel like it's kinda doing too much ? also we're parsing twice all the arguments while it's not really needed (idk if it's really impactant in terms of perf). wdyt ?

[Zuruuh]

Hello @derrabus, I didn't want to ping earlier since it was holiday season; what do you think about the approach describe above ? did you have something else in mind ?

[Zuruuh]

just saw #11208, I will update the mr to target 3.1 👍

[Zuruuh]

Update done

• dropped the custom implementation for php versions prior to 8.0 since the lib requires >=8.1
• Implemented the changes that would have been breaking in src/Query/Parser.php since the class is now finalized

[beberlei]

This was taken over by #11575