feat: Introduce FullyMultilineFixer for ControlStructure, Array & Functions. #7813

VincentLanglet · 2024-02-04T11:56:08Z

Closes #6675
Closes #5393

coveralls · 2024-02-04T12:01:19Z

coverage: 96.061% (+0.03%) from 96.034%
when pulling 4b1a181 on VincentLanglet:fully-multiline
into 5cfa1f9 on PHP-CS-Fixer:master.

doc/rules/control_structure/fully_multiline.rst

VincentLanglet · 2024-02-04T12:04:13Z

doc/rules/control_structure/fully_multiline.rst

+
+Allowed values: a subset of ``['arguments', 'arrays', 'case', 'control_structures', 'match', 'parameters']``
+
+Default value: ``['arrays']``


Not sure about the default value:

Since this rule is kinda similar/related to the TrailingCommaInMultilineFixer, I copied the default value

But if we're driven by the PSR12, the default value might be something like control_structure + parameters.

As this is a new rule, what about enabling everything by default?

doc/rules/control_structure/fully_multiline.rst

VincentLanglet · 2024-02-04T12:08:26Z

I don't think I ever created a rule so far.

I'll be happy to get some feedback (for instance about the name or the config)
@Wirone @julienfalque @mvorisek @kubawerlos before I add more example/tests. :)

doc/rules/control_structure/fully_multiline.rst

piljac1 · 2024-02-08T16:44:49Z

Great job @VincentLanglet 😄 Looking forward to this PR being merged. Not having this enforced in the PSR12 rule currently causes PHP linters (configured to PSR-12) to fail even after running the fixer on the given file(s).

Keep up the good work ❤️

Edit: I know it's a WIP, but from what I tested (and what the tests expect as well), it transforms this:

        $arr = ['test',
        'test2'];

Into this:

        $arr = [
'test',
        'test2',
];

Shouldn't it base the fixes indentation based on the line containing the opening token (bracket, brace, etc.)? So the closing token would be aligned with the opening statement and everything inside would get indented once per level:

        $arr = [
            'test',
            'test2',
        ];

Yes the array_indentation rule will probably do the trick for arrays, but I'm wondering what will happen with other things such as a multiline if condition for instance.

VincentLanglet · 2024-02-18T15:56:00Z

Edit: I know it's a WIP, but from what I tested (and what the tests expect as well), it transforms this:
        $arr = ['test',
        'test2'];
Into this:
        $arr = [
'test',
        'test2',
];
Shouldn't it base the fixes indentation based on the line containing the opening token (bracket, brace, etc.)? So the closing token would be aligned with the opening statement and everything inside would get indented once per level:
        $arr = [
            'test',
            'test2',
        ];
Yes the array_indentation rule will probably do the trick for arrays, but I'm wondering what will happen with other things such as a multiline if condition for instance.

Yes, you'll get

       $arr = [
'test',
       'test2',
];

I think it's better to rely to the array_indentation rule for the array and statement_indentation for others and it's easier for writing the rule.

doc/rules/control_structure/single_expression_per_line.rst

doc/ruleSets/PER-CS2.0.rst

doc/ruleSets/Symfony.rst

doc/rules/control_structure/single_expression_per_line.rst

src/Fixer/ControlStructure/SingleExpressionPerLineFixer.php

tests/Fixer/ControlStructure/SingleExpressionPerLineFixerTest.php

julienfalque · 2024-03-17T19:10:23Z

tests/Fixer/ControlStructure/SingleExpressionPerLineFixerTest.php

+                    $a = foo(1, 2,
+                        3
+                    );',
+            ['elements' => [SingleExpressionPerLineFixer::ELEMENTS_ARGUMENTS]],


I think it's better not to use such constants here and to use hardcoded values instead. You want your tests to fail if the accepted values change in the implementation, which is not possible when the implementation's constants are also used in the test.

Then maybe the constants could be made private in the implementation.

julienfalque · 2024-03-17T19:13:58Z

src/Fixer/ControlStructure/SingleExpressionPerLineFixer.php

+/**
+ * @author Vincent Langlet
+ */
+final class SingleExpressionPerLineFixer extends AbstractFixer implements ConfigurableFixerInterface


Part of this new rule are already implemented in method_argument_space (with the on_multiline option). Should we rework that rule somehow?

Yes, I just discover that there is an overlap with method_argument_space.

I was wondering if I should remove the argument option from my Fixer.

src/Fixer/ControlStructure/SingleExpressionPerLineFixer.php

julienfalque · 2024-03-17T19:33:48Z

src/Fixer/ControlStructure/SingleExpressionPerLineFixer.php

+            if ($tokens[$prevIndex]->equalsAny([']', [T_CLASS], [T_STRING], [T_VARIABLE], [T_STATIC]])
+                && !$tokens[$prevPrevIndex]->isGivenKind(T_FUNCTION)
+            ) {


This should have been fixed automatically 😅

Suggested change

if ($tokens[$prevIndex]->equalsAny([']', [T_CLASS], [T_STRING], [T_VARIABLE], [T_STATIC]])

&& !$tokens[$prevPrevIndex]->isGivenKind(T_FUNCTION)

) {

if (

$tokens[$prevIndex]->equalsAny([']', [T_CLASS], [T_STRING], [T_VARIABLE], [T_STATIC]])

&& !$tokens[$prevPrevIndex]->isGivenKind(T_FUNCTION)

) {

PHPCSFixer extends Symfony ruleset which disable the FullyMultilineFixer...because it will do too much change to the symfony codebase.

If the rule is wanted by Symfony but temporarily disabled for practical reasons, I'd say it should be enabled in @Symfony ruleset (so it can be applied to e.g. third party bundles) and disabled in the Symfony repository.

julienfalque · 2024-03-17T19:39:17Z

src/Fixer/ControlStructure/SingleExpressionPerLineFixer.php

+
+    private function addNewLineAfter(Tokens $tokens, int $index, int $extraIndentation = 0): int
+    {
+        $indent = $this->getIndentation($tokens, $index, $extraIndentation);


I would remove indentation handling and leave that to statement_indentation and array_indentation rules.

I took the indentation handling from MethodArgumentSpaceFixer.

If the method argument is doing it, I was thinking it could be nice do it too.
But I'd like to avoid indeed.

If I want to rely on statement_indentation and array_indentation, I need them to be perfect.
And #7884 is one blocker for example.

IMO we should rather rework method_argument_space to not handle indentation too (maybe with a deprecation path though).

For the record, you should be able to work around #7884 by fixing the wrong indent manually. The rule should not "unfix" it on subsequent run.

For the record, you should be able to work around #7884 by fixing the wrong indent manually. The rule should not "unfix" it on subsequent run.

Yes, but the rule won't be popular (especially for big code base) if you have to fix every single multi-line controle structure indentation after a run... It's definitely a no-go IMHO.

This reverts commit 02282ae.

julienfalque · 2024-03-20T08:07:42Z

tests/Fixtures/Integration/priority/single_expression_per_line,statement_indentation.test

+if ($a
+    || $b) {
+    // foo
+}


I think the test will pass if you write:

Suggested change

if ($a

|| $b) {

// foo

}

if ($a

|| $b) {

// foo

}

This is because currently, statement_indentation handles $a || $b as a single statement, so line || $b is not fixed but simply adjusted with the same indentation delta as line $a, i.e. 4 spaces.

Yes, but since most of the existing code base are using

if ($a || $b) { // foo }

I think, we must fix this correctly.

That's why I consider #7884 as a blocker.
Imho this PR is ready only after this test (untouched) is finally green.

julienfalque · 2024-03-20T16:48:52Z

src/Fixer/ControlStructure/SingleExpressionPerLineFixer.php

+        ?array $lineDelimitersIndexes = null
+    ): int {
+        $tokenAddedCount = 0;
+        if (!$tokens->isPartialCodeMultiline($begin, $end)) {


Does this mean that a structure that does not directly contain newlines but contains another structure that does will be fixed anyway? e.g.:

-$a = [foo( +$a = [ +foo( 1, 2, -)]; +) +];

Yes.

In the example

$a = [foo( 1, 2, )];

the array is multiline by definition because the open [ and the close ] are not on the same line.

Maybe I'm wrong but my interpretation of those "fully multiline" rules has always been that we should only consider newlines that are directly part of the structure we are fixing. So in this example, the foo() function must be fully multiline, but not the array itself.

@Wirone @keradus What's your opinion on this?

You're right, cf https://www.php-fig.org/psr/psr-12/#47-method-and-function-calls

A single argument being split across multiple lines (as might be the case with an anonymous function or array) does not constitute splitting the argument list itself.

so I need to fix this...

VincentLanglet changed the title ~~Introduce FullyMultilineFixer for ControlStructure, Array & Functions.~~ feat: Introduce FullyMultilineFixer for ControlStructure, Array & Functions. Feb 4, 2024

VincentLanglet commented Feb 4, 2024

View reviewed changes

doc/rules/control_structure/fully_multiline.rst Outdated Show resolved Hide resolved

VincentLanglet force-pushed the fully-multiline branch 14 times, most recently from a81d71d to 65ea695 Compare February 4, 2024 16:21

VincentLanglet commented Feb 18, 2024

View reviewed changes

doc/rules/control_structure/single_expression_per_line.rst Outdated Show resolved Hide resolved

VincentLanglet force-pushed the fully-multiline branch 8 times, most recently from f63d1b5 to 6375309 Compare February 18, 2024 19:08

VincentLanglet mentioned this pull request Feb 18, 2024

feat: Add strict option to array indentation #7846

Closed

VincentLanglet force-pushed the fully-multiline branch from fc31b96 to c658ba4 Compare March 10, 2024 16:09

VincentLanglet commented Mar 10, 2024

View reviewed changes

doc/ruleSets/PER-CS2.0.rst Outdated Show resolved Hide resolved

doc/ruleSets/Symfony.rst Show resolved Hide resolved

doc/rules/control_structure/single_expression_per_line.rst Outdated Show resolved Hide resolved

VincentLanglet added 6 commits March 14, 2024 08:31

First iteration

042c0bc

Rename

739ac5e

Fix some bugs

6c23228

Update rulesets

0a257a8

Refacto

8665d8a

Fix

44ecdc4

VincentLanglet force-pushed the fully-multiline branch from 80e571e to 44ecdc4 Compare March 14, 2024 07:31

Simplify

d9f7a09

VincentLanglet mentioned this pull request Mar 14, 2024

statement_indentation is indenting already correct line #7884

Open

VincentLanglet added 4 commits March 15, 2024 19:39

Add priority

8b36f53

Change default

0e315f7

Indent correctly

02282ae

Update doc

a773dee

julienfalque reviewed Mar 17, 2024

View reviewed changes

VincentLanglet added 4 commits March 17, 2024 20:44

Fix

c1055a5

Revert "Indent correctly"

514a9ef

This reverts commit 02282ae.

Wip

e8f5b24

Fix

ad7ef1d

julienfalque reviewed Mar 20, 2024

View reviewed changes

VincentLanglet added 4 commits March 21, 2024 09:59

Rework about what is multiline

4f4f5ab

docs

8e35b84

Fix

31c8f19

Fix

4b1a181

julienfalque mentioned this pull request Mar 23, 2024

fix: Add initial in-statements indent support #7901

Open

keradus force-pushed the master branch from a968b6e to 28209a2 Compare April 17, 2024 08:10

VincentLanglet mentioned this pull request Apr 26, 2024

Enforce new line for each array item if there's already at least one new line #6675

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: Introduce FullyMultilineFixer for ControlStructure, Array & Functions. #7813

feat: Introduce FullyMultilineFixer for ControlStructure, Array & Functions. #7813

VincentLanglet commented Feb 4, 2024 •

edited

coveralls commented Feb 4, 2024 •

edited

VincentLanglet Feb 4, 2024 •

edited

julienfalque Mar 17, 2024

VincentLanglet commented Feb 4, 2024

piljac1 commented Feb 8, 2024 •

edited

VincentLanglet commented Feb 18, 2024

julienfalque Mar 17, 2024

julienfalque Mar 17, 2024

VincentLanglet Mar 17, 2024

julienfalque Mar 17, 2024

VincentLanglet Mar 17, 2024

julienfalque Mar 20, 2024

julienfalque Mar 17, 2024

VincentLanglet Mar 17, 2024

julienfalque Mar 20, 2024

VincentLanglet Mar 20, 2024

julienfalque Mar 20, 2024

VincentLanglet Mar 20, 2024

julienfalque Mar 20, 2024

VincentLanglet Mar 20, 2024

julienfalque Mar 20, 2024

VincentLanglet Mar 20, 2024


		Allowed values: a subset of ``['arguments', 'arrays', 'case', 'control_structures', 'match', 'parameters']``

		Default value: ``['arrays']``

feat: Introduce FullyMultilineFixer for ControlStructure, Array & Functions. #7813

Are you sure you want to change the base?

feat: Introduce FullyMultilineFixer for ControlStructure, Array & Functions. #7813

Conversation

VincentLanglet commented Feb 4, 2024 • edited

coveralls commented Feb 4, 2024 • edited

VincentLanglet Feb 4, 2024 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

VincentLanglet commented Feb 4, 2024

piljac1 commented Feb 8, 2024 • edited

VincentLanglet commented Feb 18, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

VincentLanglet commented Feb 4, 2024 •

edited

coveralls commented Feb 4, 2024 •

edited

VincentLanglet Feb 4, 2024 •

edited

piljac1 commented Feb 8, 2024 •

edited