ext/intl: Add SpoofChecker::getBidiSkeleton()

[LamentXU123]

Follow up #13469. Since we have SpoofChecker::areBidiConfusable() with all these constant (UBIDI_LTR, UBIDI_RTL, UBIDI_MIXED and UBIDI_NEUTRAL) it can only determine whether two strings are visually confusable in a given text direction. Applications that need to detect confusable identifiers among a large existing set would otherwise need to repeatedly perform pairwise comparisons using areBidiConfusable().

Now since ICU 74+ and we have SpoofChecker::areBidiConfusable() already, it is convenient to add the corresponding SpoofChecker::areBidiConfusable() . ICU's bidi skeleton API transforms an identifier into a skeleton for a given text direction. So it is useful for larger compares.

e.g.

  $checker = new Spoofchecker();

  $skeleton = $checker->getBidiSkeleton(
      Spoofchecker::LTR,
      $identifier
  );

  if ($skeleton !== false && isset($knownSkeletons[$skeleton])) {
      // The identifier may be visually confusable with an existing one.
  }

The internal API is documented at: https://unicode-org.github.io/icu-docs/apidoc/dev/icu4c/uspoof_8h.html#a5aeeef353b4cca531b61480597a03e85

[devnexen]

why not uspoof_getBidiSkeletonUTF8 ?

[LamentXU123]

Correct, since PHP strings are already UTF-8 byte strings :)

[devnexen]

if you remove it, do you get a build error ?

[LamentXU123]

Yes. php_intl.h set U_SHOW_CPLUSPLUS_API to 0 so the declaration of uspoof_getBidiSkeletonUnicodeString() was hidden, even though ICU’s symbol-renaming macro still expanded the call to uspoof_getBidiSkeletonUnicodeString_74.

So I need to manually add this.

[LamentXU123]

However, if we use the UTF8 API here, we can remove this.

[devnexen]

nice to have

var_dump(
    $checker->getBidiSkeleton(Spoofchecker::LTR, "\u{05D0}A_1") ===
    $checker->getBidiSkeleton(Spoofchecker::LTR, "\u{05D0}1_A")
);
var_dump(
    $checker->getBidiSkeleton(Spoofchecker::RTL, "\u{05D0}A_1") ===
    $checker->getBidiSkeleton(Spoofchecker::RTL, "\u{05D0}1_A")
);

[LamentXU123]

RTL tests. Added.