Add types tag

[drjayvee]

The is a draft implementation for #4165.

[drjayvee]

I just force-pushed a fix to the docs

[drjayvee]

This PR isn't complete yet. I forgot to add a TypesTokenParser (and tests).

[drjayvee]

To quote my last commit message:

Move parsing of {name: 'string'} mappings to ExpressionParser

This also adds support for optional keys as discussed in #4165.

I really don't like using a static method, but I couldn't figure out how
else to call the method in tests. See the notes on the method itself.

Any help is appreciated!

[drjayvee]

I don't understand the CI error:

Remaining self deprecation notices (2)

  2x: Since twig/twig 3.12: The "tag" constructor argument of the "Twig\Node\TypesNode" class is deprecated and ignored (check which TokenParser class set it to "null"), the tag is now automatically set by the Parser when needed.
    1x in TypesTest::getTests from Twig\Tests\Node
    1x in TypesTest::testConstructor from Twig\Tests\Node

That test does not pass the tag argument.

Update: I also can't reproduce this on my local dev stack.

I think the problem is that the TypesNode constructor passes tag to parent::__construct(). I'll remove that and push again.

[ruudk]

Looks great 👏

[fabpot]

LGTM. I've added some minor comments.

[fabpot]

Thank you @drjayvee.

[fabpot]

Not sure why Github didn't work as expected here. I'm closing but it's merged.

[VincentLanglet]

Hi @drjayvee,

Is there a reason to not support

{% types {
        'foo': 'bool',
        'bar'?: 'int',
    } %}

too ? Like it's done for classic objects {% set bar = {'foo': 42} %}.

[ruudk]

I guess we did not consider this as it looks a bit ugly for optionally. But sure, we could still accept this. Especially as it's better in line with mappings.

I do hope we can turn of the quoting for the types tag in Twig-CS-Fixer though 😉

[stof]

@VincentLanglet variable names need to be valid identifiers anyway to be usable, so I don't see the benefit of supporting string literals as keys there instead of only identifiers.
This syntax is not an actual hash anyway (the parser does not allow arbitrary expressions)