[12.x] Add json:unicode cast to support JSON_UNESCAPED_UNICODE encoding

[fuwasegu]

Description

This PR introduces a new json:unicode cast type for Eloquent attributes, allowing JSON encoding with JSON_UNESCAPED_UNICODE. This makes it easier to store and retrieve JSON data without Unicode escaping (\uXXXX), which is particularly useful for applications that handle non-ASCII characters like Japanese, Chinese, or emojis.

Why?

Currently, Eloquent's json cast escapes Unicode characters by default, making JSON-encoded attributes harder to read for humans and increasing the need for manual decoding. Developers often work around this by overriding accessors/mutators or manually using json_encode($value, JSON_UNESCAPED_UNICODE), which is cumbersome.

By introducing json:unicode, developers can simply specify it in the $casts array:

protected $casts = [
    'data' => 'json:unicode',
];

This automatically applies JSON_UNESCAPED_UNICODE when encoding the attribute.

Changes

Core Modifications

• Updated Illuminate\Database\Eloquent\Casts\Json:
  • Added an optional $flags parameter to Json::encode(), allowing different encoding options.
• Updated Illuminate\Database\Eloquent\Concerns\HasAttributes:
  • Added json:unicode to the list of castable types.
  • Modified asJson() to check if json:unicode is used and apply JSON_UNESCAPED_UNICODE.

Tests

• Added testJsonCastingRespectsUnicodeOption()
  • Ensures that json:unicode encodes JSON without escaping Unicode characters.
• Added testModelJsonUnicodeCastingFailsOnUnencodableData()
  • Verifies that JsonEncodingException is thrown when json:unicode encounters malformed UTF-8 characters.

Before & After Example

Before (using default json cast)

protected $casts = ['data' => 'json'];

$model = new SomeModel();
$model->data = ['こんにちは' => '世界'];
$model->save();

echo $model->toJson();
// {"data":{"\u3053\u3093\u306b\u3061\u306f":"\u4e16\u754c"}}

After (using json:unicode cast)

protected $casts = ['data' => 'json:unicode'];

$model = new SomeModel();
$model->data = ['こんにちは' => '世界'];
$model->save();

echo $model->toJson();
// {"data":{"こんにちは":"世界"}}

Backward Compatibility

• The default json cast remains unchanged, ensuring full backward compatibility.
• Developers can opt-in to json:unicode when needed.

Potential Edge Cases Considered

• Malformed UTF-8 data: Ensured json:unicode still throws JsonEncodingException when encoding fails.
• Null handling: Verified that null values are properly handled.
• Attribute retrieval: Ensured both json and json:unicode return arrays when accessed.

[AndrewMast]

I don't like idea of adding a single-use parameter to asJson. What if we wanted to add (for whatever reason), the ability to do JSON_PRETTY_PRINT? Or maybe JSON_PARTIAL_OUTPUT_ON_ERROR?

I think it should be asJson($value, int $flags = 0) and we can add an additional method to process the flags.

Maybe something like this?

protected function getJsonCastFlags($key): int
{
    $flags = 0;

    if ($this->hasCast($key, ['json:unicode'])) {
        $flags |= JSON_UNESCAPED_UNICODE;
    }

    return $flags;
}

[fuwasegu]

Thank you for your review!
I've addressed your feedback in commit 61fad07.
Please let me know if there's anything else that needs adjustment.

[shaedrich]

You could probably even add a type-check to ./types/Database/Eloquent/Casts to ensure that only a valid integer is returned 🤔

[shaedrich]

You could possibly make this type more precise:

Suggested change

	* @param int $flags
	* @return string
	*/
	protected function asJson($value)
	protected function asJson($value, $flags = 0)
	* @param int-mask-of<JSON_*> $flags
	* @return string
	*/
	protected function asJson($value, $flags = 0)

[shaedrich]

Therefore, you can narrow the return here as well:

Suggested change

	* @return int
	* @return int-mask-of<JSON_*>