Cleanup performance

[vencakrecl]

For the project where I use this package, the cleanup process was slow. So, I created this solution which reduced the time from 537s to 17s. It would be nice to integrate the solution into the package.

        $content = $generator?->toYaml() ?? '';

        $stop = false;
        $i = 1;
        while ($stop === false) {
            $matches = [];
            preg_match_all("@'#/components/schemas/(.*)'@", $content, $matches);

            $data = Yaml::parse($content);

            $unset = false;
            foreach ($data['components']['schemas'] as $key => $match) {
                if (!in_array($key, $matches[1])) {
                    unset($data['components']['schemas'][$key]);
                    $unset = true;
                }
            }

            $content = Yaml::dump($data, 10, 2, Yaml::DUMP_OBJECT_AS_MAP | Yaml::DUMP_NUMERIC_KEY_AS_STRING);

            if ($unset === false) {
                $stop = true;
            }

            echo "cleaning loop $i";
            $i++;
        }

        file_put_contents(
            filename: sprintf('%s/openapi.yaml', __DIR__),
            data: Yaml::dump($data, 10, 2, Yaml::DUMP_OBJECT_AS_MAP | Yaml::DUMP_NUMERIC_KEY_AS_STRING)
        );

[DerManoMann]

Wow! Impressive numbers. Is that already using the new option to turn off recursive cleanup on path filter?

[vencakrecl]

@DerManoMann No, it's not. The code uses yaml output, which is cleaned in a loop until all unused components are removed.

[DerManoMann]

@vencakrecl Ok, thanks.
It would be great if you could try to benchmark the new flag (assuming you are using PathFilter).
Alternatively, if there is a big enough public project out there that I could use for benchmarking then that would be cool too. Any suggestions welcome.

[DerManoMann]

I'll leave this here as a possible enhancement - this is outside the scope of processors (unless we allow a 2nd pipeline after serialization)