str_(starts|ends)_with variadic needle

[divinity76]

make str_starts_with and str_ends_with variadic:

str_starts_with(string $haystack, string ...$needle): bool
str_ends_with(string $haystack, string ...$needle): bool

a simple example could be

if (str_starts_with($url, "http://", "https://")) {
    // url
}
if (str_ends_with($filename, ".pdf", ".doc", ".docx")) {
    // document
}
if(str_ends_with($str, ...$validExtensions){
    // valid extension
}

which seems easier than

if (str_starts_with($url, "http://") || str_starts_with($url, "https://")) {
    // url
}
if (str_ends_with($filename, ".pdf") || str_ends_with($filename, ".doc") || str_ends_with($filename, ".docx")) {
    // document
}
$isValidExtension = false;
foreach ($validExtensions as $needle) {
    if (str_ends_with($str, $needle)) {
        $isValidExtension = true;
        break;
    }
}
if ($isValidExtension) {
    // valid extension
}

inspired by Python's str.startswith() which supports python-tuples like if str.startswith(("foo","bar")): ...

[iluuu1994]

This requires a discussion on the mailing list and probably an RFC.

[divinity76]

asked internals. https://news-web.php.net/php.internals/127637

[divinity76]

I see a theoretical issue with using variadic specifically: it will be difficult to add new arguments, like bool $case_sensitive=true in the future.

That can be avoided by accepting string|string[] $needle instead of string ...$needle

That also more closely resemble how Python handles it

[bcremer]

A signature of str_starts_with(string $haystack, string ...$needle): bool allows the function to be called without a second argument.

Current behavior:

str_starts_with('foo'); // Uncaught ArgumentCountError: str_starts_with() expects exactly 2 arguments, 1 given

If this should be maintained the signature should be

str_starts_with(string $haystack, string $needle, string ...$additionalNeedles): bool

I'm not sure how this is handled with other core functions that accept variadic inputs or if this minor behavior change is considered a bc break.

[divinity76]

@bcremer this branch, as currently written, still requires a minimum of 2 arguments.

hans@DESKTOP-EE15SLU:~/projects/php-src$ ./sapi/cli/php -r 'str_ends_with("foo");'

Fatal error: Uncaught ArgumentCountError: str_ends_with() expects at least 2 arguments, 1 given in Command line code:1
Stack trace:
#0 Command line code(1): str_ends_with('foo')
#1 {main}
  thrown in Command line code on line 1
hans@DESKTOP-EE15SLU:~/projects/php-src$ ./sapi/cli/php -r 'str_ends_with("foo", ...[]);'

Fatal error: Uncaught ArgumentCountError: str_ends_with() expects at least 2 arguments, 1 given in Command line code:1
Stack trace:
#0 Command line code(1): str_ends_with('foo')
#1 {main}
  thrown in Command line code on line 1

it's subtle, but i believe

        Z_PARAM_VARIADIC('+', needles, num_needles)

means require at least 1 argument while

        Z_PARAM_VARIADIC('*', needles, num_needles)

would mean no arguments required

[bcremer]

@divinity76
OK, but that in turn would mean the signature in basic_functions.stub.php is not correct and does not match the implementation.

Also tools like phpstan loose the information about the required arguments when using the provided stub. See: https://phpstan.org/r/8c158f00-493b-44c4-8cb6-e5b87b53511e

Is the stub file auto-generated or can it be modified manually?

[divinity76]

it can be modified manually. do you mean it should be

str_starts_with(string $haystack, string $needle, string ...$needles): bool

then?

[bcremer]

Disclaimer, I'm not a core contributor but user of static code analysis in userland code.

But yes, the following signature should match the implementation and produces the same error message in phpstan as before:

function str_starts_with(string $haystack, string $needle, string ...$needles): bool 

https://phpstan.org/r/00ec57e0-6c21-49d7-a9e5-c102ad02e8cc

[TimWolla]

it's subtle, but i believe

        Z_PARAM_VARIADIC('+', needles, num_needles)

The first parameter is ignored there (probably for backwards compatibility reasons).

[TimWolla]

Allowing zero needles would not be a BC break (it's widening the signature) and would be consistent with:

array_any($needles, static fn ($needle) => str_starts_with($haystack, $needle));

returning false for an empty array.

[Girgias]

Is this PR still being pursued?

[divinity76]

@Girgias nope. I still believe it's a good idea, but not enough to pursue it myself, I have too much other stuff to deal with these days. Closing.

Hope someone else may pursue it in the future though!