pykakasi duplicates characters when dealing with unusual unicode characters such as U+00D7 "MULTIPLICATION SIGN"

[curiousjp]

Describe the bug When providing convert() with a string containing Unicode character U+00D7, results list provides an empty dictionary for that character, but then reduplicates the preceding item. The examples given below might provide a clearer explanation.

To Reproduce Steps to reproduce the behavior:

import pykakasi
kakasi = pykakasi.kakasi()
print( kakasi.convert( "三x五" ) )
[{'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}, {'orig': 'x', 'hira': 'x', 'kana': 'x', 'hepburn': 'x', 'kunrei': 'x', 'passport': 'x'}, {'orig': '五', 'hira': 'ご', 'kana': 'ゴ', 'hepburn': 'go', 'kunrei': 'go', 'passport': 'go'}]
print( len( kakasi.convert( "三x五" ) ) )
3

print( kakasi.convert( "三×五" ) )
[{'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}, {'orig': '×', 'hira': '', 'kana': '', 'hepburn': '', 'kunrei': '', 'passport': ''}, {'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}, {'orig': '五', 'hira': 'ご', 'kana': 'ゴ', 'hepburn': 'go', 'kunrei': 'go', 'passport': 'go'}]
print( len( kakasi.convert( "三×五" ) ) )
4

Expected behavior

print( kakasi.convert( "三×五" ) )
[{'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}, {'orig': '×', 'hira': '×', 'kana': '×', 'hepburn': '×', 'kunrei': '×', 'passport': '×'}, {'orig': '五', 'hira': 'ご', 'kana': 'ゴ', 'hepburn': 'go', 'kunrei': 'go', 'passport': 'go'}]
print( len( kakasi.convert( "三×五" ) ) )
3

Environment (please complete the following information):

• OS: Windows Subsystem For Linux (Ubuntu), hosted on Windows 10
• Python 3.8.5
• pykakasi version: 2.2.1 (via pip)

Additional context Problem discovered while parsing ebook titles.

[miurahr]

152 adds an implement to support Latin-1 characters.

Kakasi is originally born in Japan in EUC_JP encoding, so Latin-1 characters are basically out-of-scope.

[miurahr]

A fix is released as v2.3.0b1.

[zhangfeiran]

I installed v2.3.0b, but the duplication bug still exists for some other Unicode characters:

import pykakasi
kakasi = pykakasi.kakasi()
print(kakasi.convert('三'))
# [{'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}]
print(kakasi.convert('三｡'))
# [{'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}, {'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}]
print(kakasi.convert('｢三｣'))
# [{'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}, {'orig': '三', 'hira': 'さん', 'kana': 'サン', 'hepburn': 'san', 'kunrei': 'san', 'passport': 'san'}]

｡ ｢ ｣ is U+FF61, U+FF62, U+FF63, coincidentally.