Introduce lowercase-string

[VincentLanglet]

No description provided.

[staabm]

this cleanup regarding count-of-args and the indent change over the whole block could be a separate PR to make this diff here more readable

[VincentLanglet]

I opened #3439

[staabm]

edge-case which came to mind while looking at the PR:

offset-lookup in a lower-case string yields a lower-case string:
https://3v4l.org/1eWRH

[VincentLanglet]

There is at least

src/Type/Php/ReplaceFunctionsDynamicReturnTypeExtension.php
src/Type/Php/SprintfFunctionDynamicReturnTypeExtension.php
src/Type/Php/StrPadFunctionReturnTypeExtension.php
src/Type/Php/StrRepeatFunctionReturnTypeExtension.php
src/Type/Php/StrTokFunctionReturnTypeExtension.php

wich can be improved too, but maybe it's better in another PR.

[thg2k]

What's the use case? Is there an issue linked explaining why we need this? Why not also uppercase string?

[VincentLanglet]

Why not also uppercase string?

lowercase-string is the only one supported by Psalm so far and codebase more often use lowercase-string than uppercase string. https://psalm.dev/r/c4d35bbe26

You can see it's used a lot in their codebase https://github.com/search?q=repo%3Avimeo%2Fpsalm%20lowercase-string&type=code
But it's also used in some others codebase like https://github.com/search?q=repo%3ARoave%2FBetterReflection%20lowercase-string&type=code (and certainly more).

It's useful

• For case-insensitive mapping/behavior
• When working with case insensitive data, like URL.

I personally work a lot with URL recently, and it's useful to unsure/checks that I always work with lowercase string.

[staabm]

Use separate test files in the nsrt/ folder and use a different // lint comment condition in the first line and PhpVersion method for the version check in the impl

[phpstan-bot]

This pull request has been marked as ready for review.

[VincentLanglet]

I think it will be easier and better to have

src/Type/Php/ReplaceFunctionsDynamicReturnTypeExtension.php
src/Type/Php/SprintfFunctionDynamicReturnTypeExtension.php
src/Type/Php/StrPadFunctionReturnTypeExtension.php
src/Type/Php/StrRepeatFunctionReturnTypeExtension.php
src/Type/Php/StrTokFunctionReturnTypeExtension.php

in another PR.

[staabm]

I think this should include str-concatenation of 2 lowercase strings, because its a super low level thing

[VincentLanglet]

I think this should include str-concatenation of 2 lowercase strings, because its a super low level thing

Indeed, it's supported now.

[staabm]

I think we are missing a rule test for e.g.

/** @param lowercase-string $s */
function doFoo($s) {
  if ($s === 'AB') {}
  if ('AB' === $s) {}
  if ('AB' !== $s) {}
  if ($s === 'ab') {}
  if ($s !== 'ab') {}
  if ($s === 'aBc') {}
  if ($s !== 'aBc') {}
  if ($s === '01') {}
  if ($s === '1e2') {}
  if ($s === MyClass::myConst) {}
  if ($s === A_GLOBAL_CONST) {}
  if (doFoo('hi') === 'AB') {}
}

which asserts rule error(s) when comparing constant with lowercase-string.

---

as a followup it also can be useful to test comparisons for string-functions

/** @param lowercase-string $s */
function doFoo($s, int $x, int $y) {
  $part = substr($s, $x, $y);
  if ($part === 'AB) {} // always false
}

[staabm]

do we have a test somewhere for this type?
(I think I see it the very first time today)

[VincentLanglet]

Added in 9636e83

[VincentLanglet]

After this PR, we may want to introduce check for always false result
StrPosRule/StrStartsWithRule/StrContainRule/StrEndswithRule
cf https://phpstan.org/r/49d9f64f-9488-44ca-81e4-8d73767c4084

[ondrejmirtes]

1. There are no tests about accepting lowercase-string. Add a new test method to CallMethodsRuleTest and try passing different things like a constant string, another lowercase-string and general string to a parameter that accepts lowecase-string.
2. This change is quite disruptive as you can see from the tests that needed changing. We should do this only for bleeding edge, similarly to the list type. This means that all the places that currently do new IntersectionType and add the new accesory type will need to do AccessoryLowercaseStringType::intersectWith() instead. Also, TypeNodeResolver will interpret lowercase-string type the new way only with bleeding edge.

[ondrejmirtes]

You copied this from accessory literal string type, but it does not belong here.

[ondrejmirtes]

Also: We need excessive tests in TypeCombinatorTest::dataUnion and dataIntersect about the interaction with other string types.

[ondrejmirtes]

I'd say yes.

[thg2k]

"10E4" is a numeric string not lowercase

[VincentLanglet]

Yes that is the reason I kept maybe

[ondrejmirtes]

Maybe we don't need to do this only for bleeding edgw, just do a better job in VerbosityLevel::getRecommendedLevelByType

[VincentLanglet]

Maybe we don't need to do this only for bleeding edgw, just do a better job in VerbosityLevel::getRecommendedLevelByType

Do you have a preferred solution between bleeding edge and changing getRecommendedLevelByType ?

I unsure to understand what is needed to change in getRecommendedLevelByType.

[ondrejmirtes]

I'd like for lowercase-string to be in an error message about not-accepted type only if there's lowercase-string on the accepting side.

Please start by showing in a test how IntersectionType::describe() behaves with different VerbosityLevel when there's AccessoryLowercaseStringType involved.

I think we could bump it from value to precise.

And then in getRecommendedLevelByType we should look at $acceptedType and bump the returned level from value to precise only if $acceptingType->isLowecaseString()->yes().

After you do that, the impact on existing rules tests should be minimal. And I could merge it for everyone, not just on bleeding edge.

[VincentLanglet]

This is in a very good shape and close to merging.

Thanks for the kind works. I wasn't sure about it, it motivates a lot.

I feel like we don't need to update testParsingDesiredTypeDescription and testDesiredTypeDescription. Instead, add a test to TypeToPhpDocNodeTest.

I removed the test in testParsingDesiredTypeDescription and testDesiredTypeDescription.
I already added one in TypeToPhpDocNodeTest

yield [
			new IntersectionType([new StringType(), new AccessoryLowercaseStringType()]),
			'lowercase-string',
		];

Just revert the change in this rule and we'll see what fails.

I reverted #3438 (comment), you can now see the failure (cf https://github.com/phpstan/phpstan-src/actions/runs/11049021257/job/30693708332?pr=3438) ; since StrictComparisonOfDifferentTypesRule use VerbosityLevel::value(), the error message are like

Strict comparison using === between string and 'AB' will always evaluate to false.

which are not helping, instead of

Strict comparison using === between lowercase-string and 'AB' will always evaluate to false.

[ondrejmirtes]

See 7b1ade2

[ondrejmirtes]

Thank you.

[ondrejmirtes]

Now you can tackle phpstan/phpstan#7703 with Type::isNominalString(string $name): TrinaryLogic :) That would be awesome.

[staabm]

Excellent!