Separator and digit characters used by locale-aware formatters
===============================================================
Each character below appears LITERALLY between the << >> guards. Guards are ASCII.

GROUP (THOUSANDS) SEPARATORS
  U+002C COMMA                        <<,>>    en-US, en-GB, en-IN, ja-JP, zh-CN
  U+002E FULL STOP                    <<.>>    de-DE, es-ES
  U+00A0 NO-BREAK SPACE               << >>    pl-PL, ru-RU
  U+202F NARROW NO-BREAK SPACE        << >>    fr-FR (changed from U+00A0 in CLDR 34)
  U+2019 RIGHT SINGLE QUOTATION MARK  <<’>>    de-CH (an apostrophe-LOOKING character, not U+0027)
  U+066C ARABIC THOUSANDS SEPARATOR   <<٬>>    ar-EG and other 'arab' numbering locales

DECIMAL SEPARATORS
  U+002E FULL STOP                    <<.>>    en-*, ja-JP, zh-CN, de-CH
  U+002C COMMA                        <<,>>    de-DE, fr-FR, es-ES, pl-PL, ru-RU
  U+066B ARABIC DECIMAL SEPARATOR     <<٫>>    ar-EG

THE GAP BETWEEN AN AMOUNT AND ITS CURRENCY SYMBOL
  U+00A0 NO-BREAK SPACE               << >>    de-DE, fr-FR, es-ES, pl-PL, ru-RU, de-CH
  (none)                                       en-US, en-GB, en-IN, ja-JP

DIGIT SETS (CLDR numbering systems)
  latn  0 1 2 3 4 5 6 7 8 9           <<0123456789>>
  arab  U+0660..U+0669                <<٠١٢٣٤٥٦٧٨٩>>    Arabic-Indic, default for ar-EG
  arabext U+06F0..U+06F9              <<۰۱۲۳۴۵۶۷۸۹>>    Extended Arabic-Indic, default for fa-IR
  deva  U+0966..U+096F                <<०१२३४५६७८९>>    Devanagari, selectable for hi-IN

PERCENT AND SIGN CHARACTERS
  U+0025 PERCENT SIGN                 <<%>>
  U+066A ARABIC PERCENT SIGN          <<٪>>
  U+2212 MINUS SIGN                   <<−>>    CLDR's minus is NOT U+002D HYPHEN-MINUS
  U+002D HYPHEN-MINUS                 <<->>

WHAT TO ASSERT
--------------
1. Reading this file and extracting each << >> block yields exactly the code point named on the
   same line — no normalisation has folded U+00A0 to U+0020 or U+2019 to U+0027.
2. Your number parser accepts every group separator listed and returns 1234567 for a string built
   from any of them.
3. Your digit mapper turns each non-latn digit set back into 0-9.
4. A round trip through your storage layer (JSON, CSV, a database column) preserves U+202F and
   U+00A0 rather than collapsing either to a plain space.