[browser] HybridGlobalization allows different hashes for strings that return true for Equals

[ilonatommy]

We do not have a native, locale sensitive algorithm for HybridGlobalization, that would replace the ICU data. Implementing it manually in JS would slow down hashing mechanism. From this reason, we decided to use Invariant hasing and sort key functions in HybridGlobalization mode (see: #96354).
This approach has a few downsides, including a situation where two strings, e.g. s1 = "igloo", s2 = "İGLOO"

that are equal under "tr-TR" linguistic comparer (ICU or HybridGlobalization), produce:

bool linguisticComparer = new CultureInfo("tr-TR").CompareInfo.Compare(s1, s2, CompareOptions.IgnoreCase); // true

The HasCodes in ICU mode and HybridGlobalization mode are produced differently:

int hashCode1 = new CultureInfo("tr-TR").CompareInfo.GetHashCode(s1, CompareOptions.IgnoreCase);
int hashCode2 = new CultureInfo("tr-TR").CompareInfo.GetHashCode(s2, CompareOptions.IgnoreCase);
// true when ICU loaded (Hybrid mode off), false when HybridGlobalization mode on (reduced ICU loaded)

Equality function should have the same effect as GetHashCode comparison, however for HybridGlobalization this does not stand.

There are HashCodeLocalized test cases excluded for HG marked with this issue.

Possible fix:

• switching HybridGlobalization off (this will increase the application size by loading additional ICU data)
• implementing JS-based, locale-sensitive hashing (this will be much slower than Invariant or ICU hashing).

[ghost]

Tagging subscribers to 'arch-wasm': @lewing
See info in area-owners.md if you want to be subscribed.

Issue Details

---

We do not have a native, locale sensitive algorithm for HybridGlobalization, that would replace the ICU data. Implementing it manually in JS would slow down hashing mechanism. From this reason, we decided to use Invariant hasing and sort key functions in HybridGlobalization mode (see: #96354).
This approach has a few downsides, including a situation where two strings, e.g. s1 = "Hello", s2 = "He\u200dllo"

that are equal under linguistic comparer (ICU or HybridGlobalization), produce:

bool linguisticComparer = StringComparer.InvariantCulture.Equals(s1, s2); // true

The HasCodes in ICU mode and HybridGlobalization mode are produced differently:

bool areHasCodesSame = StringComparer.InvariantCulture.GetHashCode(s1) == StringComparer.InvariantCulture.GetHashCode(s2);
// true when ICU loaded (Hybrid mode off), false when HybridGlobalization mode on (reduced ICU loaded)

Equality function should have the same effect as GetHashCode comparison, however for HybridGlobalization this does not stand.

Possible fix:

• switching HybridGlobalization off (this will increase the application size by loading additional ICU data)
• implementing JS-based, locale-sensitive hashing (this will be much slower than Invariant or ICU hashing).

Author:	ilonatommy
Assignees:	-
Labels:	arch-wasm, area-System.Globalization
Milestone:	-

[pavelsavara]

I think this breaks contract between Equals and GetHashCode in bad way.

[ilonatommy]

Closing - the PR that was going to introduce this behavior got changed.