localeCompare() doesn't return the expected value

[apptimise]

Minimal example:

var b = "aZ";
var a= "ab"
var ret = b.localeCompare(a); // returns -8 on Samsung Internet for Android 4.0

[emilio]

Hi, thanks for reporting this :)

[emilio]

We never call JS_SetLocaleCallbacks, so they should use the default. I'll try with a debug-mozjs build.

[Ms2ger]

Assigned to @emilio for investigation

[emilio]

This is not a bug, SpiderMonkey by default (in js::str_localecompare returns the difference between the first two different characters, which in this case is -8.

From the ECMA spec this is completely valid:

The actual return values are implementation-defined to permit implementers to encode additional information in the value, but the function is required to define a total ordering on all Strings. This function must treat Strings that are canonically equivalent according to the Unicode standard as identical and must return 0 when comparing Strings that are considered canonically equivalent.

So I'm closing this, feel free to reopen if anyone disagrees with this diagnostic :)

[apptimise]

I don't agree. The result of the comparison should be in the range {-1, 0, +1} http://www.w3schools.com/jsref/jsref_localecompare.asp
Even if the return value is "implementation-defined", it should return a positive number. "aZ" is definitely larger than "ab". All other browsers return "+1". It should be okay to return "+8" or a positive number, but not "-8"! It changes the whole functionality

[emilio]

Hmm... Ok, so this is spec compliant (because the spec only requires to define a total order), and the current behavior is the same of strcoll, which is sort of the expected thing.

That being said, since nobody else does this, I guess it makes sense to switch.

Seems like the easiest way to get this working is compiling SM with EXPOSE_INTL_API, so we get the self-hosted localeCompare, which ends-up calling http://searchfox.org/mozilla-central/source/js/src/builtin/Intl.cpp#1083.

@jdm, @Ms2ger, do you think that's reasonable? I can try to do that tomorrow, I guess.

[jdm]

Seems reasonable.