Missing splat collections

[withinboredom]

A way I've seen to enforce types with arrays is something like the psalm method + class method:

/**
 * @template T of Item
 */
class SplatCollection {
  public function append(Item ...$items) {
    array_push($this->iterator, ...$items);
  }
}

$collection->append(...$items);

I have no idea what the performance penalty is for using splats like this, but this is what the output is when added to the phpbench:

PHPBench (1.2.5) running benchmarks... #standwithukraine
with configuration file: /app/phpbench.json
with PHP version 8.1.4, xdebug ❌, opcache ✔

\App\Tests\TypHintBench

    benchWithoutType........................R2 I88 - Mo107.894μs (±0.49%)
    benchWithArrayType......................R2 I53 - Mo108.076μs (±0.47%)
    benchWithMixedType......................R1 I62 - Mo103.304μs (±0.47%)
    benchWithClassType......................R1 I60 - Mo153.612μs (±0.42%)
    benchWithSplatType......................R1 I86 - Mo167.429μs (±0.31%)

Subjects: 5, Assertions: 0, Failures: 0, Errors: 0
+--------------+--------------------+-----+------+-----+-----------+-----------+--------+
| benchmark    | subject            | set | revs | its | mem_peak  | mode      | rstdev |
+--------------+--------------------+-----+------+-----+-----------+-----------+--------+
| TypHintBench | benchWithoutType   |     | 1000 | 100 | 612.464kb | 107.894μs | ±0.49% |
| TypHintBench | benchWithArrayType |     | 1000 | 100 | 612.464kb | 108.076μs | ±0.47% |
| TypHintBench | benchWithMixedType |     | 1000 | 100 | 612.464kb | 103.304μs | ±0.47% |
| TypHintBench | benchWithClassType |     | 1000 | 100 | 612.464kb | 153.612μs | ±0.42% |
| TypHintBench | benchWithSplatType |     | 1000 | 100 | 612.464kb | 167.429μs | ±0.31% |
+--------------+--------------------+-----+------+-----+-----------+-----------+--------+

so it seems like it might be pretty slow.

Pros:

• type checking by php
• static analysis by psalm

Cons:

• slow?
• weird to write and read

[mrsuh]

Hi
It's great you did splat type test!
My goal was to compare the array/psalm/generics collections. Array collections use an array type and psalm/generics use a class type. After the first tests I decided to add without type and mixed tests just because it's very interesting)
Your test is also interesting.
I think the splat test with class type has the right place in your benchmark table: array is faster than class and class with splat takes extra time to handle the argument.

Can you make PR with your code and result?