ICU4X Collation failures about conformance HOT 7 CLOSED

Hey @hsivonen, can you take a look at these? Here is how we configure the ICU4X collator:

    let mut options = CollatorOptions::new();
    options.strength = Some(Strength::Tertiary);

    // Ignore punctuation only if using shifted test.
    if let Some(ip) = ignore_punctuation {
        if *ip {
            options.alternate_handling = Some(AlternateHandling::Shifted);
        }
    }

    let collator: Collator = Collator::try_new(&locale!("en").into(), options).unwrap();

    let comparison = collator.compare(str1, str2);

Here is the corresponding JS code which passes the tests:

    // Locale if provided in the test data.
    let testLocale = undefined;
    if ('locale' in json) {
      testLocale = json['locale'];
    }
    let testCollOptions = {};
    if ('ignorePunctuation' in json) {
      testCollOptions = {
        ignorePunctuation:json['ignorePunctuation']}
    }

    // Get other fields if provided
    let rules = undefined;
    if ('rules' in json) {
      rules = json['rules'];
    }

    // Set up collator object with optional locale and testOptions.
    let coll;
    try {
      coll = new Intl.Collator(testLocale, testCollOptions);

      let d1 = json['s1'];
      let d2 = json['s2'];

      const compared = coll.compare(d1, d2);

These cases have no locale option so I don't have reason to believe that the locale is at fault.

Can you advise on how to get the Intl.Collator behavior via ICU4X APIs?

from conformance.

Acknowledging that I've seen this. This needs stepping through in a debugger.

from conformance.

Do I understand correctly that your harness wants the outcome for all these three to be Less, but the outcome is Equal? If so, why is Less expected?

• {"label":"0010001","s1":"𑜿!","s2":"𑜿?","line":8661,"ignorePunctuation":true} CollatorOptions { strength: Some(Tertiary), alternate_handling: Some(Shifted), case_first: None, max_variable: None, case_level: None, numeric: None, backward_second_level: None }

These are Equal, because the inputs only differ by punctuation, the mechanism for ignoring punctuation is shifting punctuation to the quaternary level and you only compare on the tertiary level.

• {"label":"0243300","s1":"𑛁b","s2":"𑜱b","line":47434} CollatorOptions { strength: Some(Tertiary), alternate_handling: None, case_first: None, max_variable: None, case_level: None, numeric: None, backward_second_level: None }

These are Equal, because both strings are the same.

• {"label":"0373766","s1":"龜a","s2":"龜a","line":177900} CollatorOptions { strength: Some(Tertiary), alternate_handling: None, case_first: None, max_variable: None, case_level: None, numeric: None, backward_second_level: None }

These are Equal, because both compatibility ideographs normalize to U+9F9C, and the collator always normalizes.

from conformance.

I timed out on attempting to run the tests locally. Anyway, my guess is that this is a test harness bug, and the most suspicious part of the test harness is this bit:

    let result = comparison == Ordering::Less;


    // TODO: How to do this easier?
    let mut result_string = true;
    if !result {
        result_string = false;
    }

from conformance.

Thank you @hsivonen!

Yeah, @sven-oly, here is a difference in code:

In Rust you say:

let result = comparison == Ordering::Less;

In JS you say:

      let result_bool = true;
      if (compared > 0) {
        result_bool = false;
      }

So it appears perhaps Rust is looking for < and JS is looking for <=

from conformance.

from conformance.