Parsing of dates in "ar" locale fails when typed

[jokraft]

Describe the bug
There is a problem with parsing type dates in "ar" locale (see: flutter/flutter#138635)

To Reproduce
see examples in flutter/flutter#138635
Alternatively:

import 'package:intl/date_symbol_data_local.dart';
import 'package:intl/intl.dart';

void main() {
  initializeDateFormatting("ar");
  // works
  DateFormat('yMd', "ar").parseStrict(String.fromCharCodes(
      [1633, 1633, 8207, 47, 1633, 1633, 8207, 47, 1634, 1632, 1632, 1633]));
  // fails (codeUnits of typed date)
  try {
    DateFormat('yMd', "ar").parseStrict(String.fromCharCodes(
        [1633, 1633, 47, 1633, 1633, 47, 1634, 1632, 1632, 1633]));
  } catch (e) {
    print(e);
  }
}

Did some digging and my current theory is, that date_time_patterns.dart contains some unneeded "RLM" characters. Are they supposed to be generated there?
If I change keyboard to arabic and type the date there will be no such characters and as such parsing fails.

System Info

Dart SDK 3.5.0-277.0.dev Flutter SDK 3.23.0-13.0.pre.277 flutter_personal_test 0.0.0

dependencies:

• flutter 0.0.0 [characters collection material_color_utilities meta vector_math sky_engine]
• flutter_localizations 0.0.0 [flutter intl characters clock collection material_color_utilities meta path vector_math]

dev dependencies:

• file 7.0.0 [meta path]

transitive dependencies:

• characters 1.3.0
• clock 1.1.1
• collection 1.18.0
• intl 0.19.0 [clock meta path]
• material_color_utilities 0.11.1 [collection]
• meta 1.15.0
• path 1.9.0
• sky_engine 0.0.99
• vector_math 2.1.4

[mosuem]

The date time patterns are autogenerated from the Unicode CLDR database. Date parsing is unfortunately very fragile.

[jokraft]

RLMs are also in ar.xml in CLDR
https://unicode-org.atlassian.net/jira/software/c/projects/CLDR/issues/CLDR-14076 suggests to me that is supposed to be like that. (I am personally not sure since the RLM is for strings with mixed text directions as far as I can see, which should not normally be the case if someone is simply using arabic locale, right?)
I did not manage to get RLM into the string by typing. (Also had a look at some arabic dates I found online and they also did not include RLMs)
Question remains where this should be mitigated. Does every caller of parseStrict need to handle adding the RLMs when in arabic locale? Or should DateFormat handle this?

[mosuem]

I don't know what the best strategy is - parsing is inherently fragile exactly because of cases like this. Maybe the logic of parsing could be special-cased for arabic parsing. I am open for reviewing a PR!