Cli command classes

[weirdan]

Over the years Psalm's entrypoints have grown quite large. This PR aims to solve it by moving the entrypoint code to their own classes.

Overview of changes:

• src/*.php entrypoints moved to corresponding classes in src/Psalm/Internal/Cli folder
• Global functions in \Psalm namespace deprecated and moved to Psalm\Internal\CliUtils class
• Psalm\Internal\Cli\Psalm::run() (originally src/psalm.php) split into multiple methods
• Psalm\Internal\Cli\Psalter::run() (originally src/psalter.php) split into multiple methods

[weirdan]

The remaining CI issue is somewhat puzzling. It's only reproducible with --threads=1 (which is implicit on AppVeyor).

There seems to be an existing bug where multiple relative paths may reference the same file, ending up in the same file being scanned several times.

[orklah]

There's a lot of code being repeated (sometimes with mistakes) between entry points, for example, the php-version fixed here #5855.

I see you introduced a initPhpVersion on Psalm entry point. Maybe this method should be extracted into CliUtils class and used by everyone?

[weirdan]

@orklah I'd like to defer that to subsequent, more focused, PRs. While I certainly agree with the sentiment, this one is quite big already. Here I aimed for bug-for-bug compatibility, to ease the transition.

[orklah]

Well, seems weird you refactored things in methods only in Psalm entrypoint and not on Psalter then.

But apart from that, the PR seems great :)

[weirdan]

Well, seems weird you refactored things in methods only in Psalm entrypoint and not on Psalter then.

Initially I did not intend to do that in this PR, but Psalm was complaining about ComplexMethod 😄. So once it stopped doing that I deemed the PR good enough.

[orklah]

Can relate to that :D

[muglug]

Thanks!

Initially I did not intend to do that in this PR, but Psalm was complaining about ComplexMethod 😄. So once it stopped doing that I deemed the PR good enough.

Is this annoying or is it good?

[weirdan]

Is this annoying or is it good?

It's ok, but I did not expect that - the method was no more complex than the closure originally found in src/psalm.php, but Psalm did not flag that closure. Also the measure of complexity used is not very intuitive. IIRC, the most complexity reduction came from extracting code like this:

psalm/src/psalm.php

Lines 218 to 236 in 38c452a

	if (array_key_exists('help', $options)) {
	$options['h'] = false;
	}
	if (array_key_exists('version', $options)) {
	$options['v'] = false;
	}
	if (array_key_exists('init', $options)) {
	$options['i'] = false;
	}
	if (array_key_exists('monochrome', $options)) {
	$options['m'] = false;
	}
	if (isset($options['config'])) {
	$options['c'] = $options['config'];
	}

Arguably this does not look the most complex part of that method.