ICU currently doesn't recognize equivalence classes of punctuation when parsing
numbers. We received this bug report from a developer:
ICU/CFNumberFormatter does not recognize characters other than U+002D as minus
signs. The Unicode standard has several other minus signs, such as U+2212,
U+FE63, and U+FF0D. This causes problems with Japanese input in particular,
because with many Japanese input methods, hitting the minus sign on the keyboard
next to 0 (not the minus sign on the numeric keypad) results in the character
U+2212 being inserted.
(Deborah again) This probably applies to periods, commas, etc. as well. It's
better to be forgiving when parsing.
Mon Dec 19 11:25:25 2005 weiv changed notes2: assign: "" to "srl", priority: "" to "assess",
Mon Dec 19 11:25:25 2005 weiv moved from incoming to formatting
All of the relevant changes for lenient number parsing, including those for this ticket, are committed under #6109 as part of the changes in r30018. I will soon add some information in #6109 on how and what to review.
#8527 is the umbrella bug describing the overall changes and what revs to look at.
Milestone 4.8RC deleted