VersionedProperty is garbled when reading some old files
Spotted while working on #432, I got character.jsp to tell me that U+0041 went
| Identifier_Type | 9.0:Not_Character | 10.0:Recommended |
|---|
Which seems wrong, and indeed is; the default value changed between 9.0 and 10.0, from Recommended to Not_Character.
Likewise for U+4E00:
| Block | 2.0..3.0: 9FFF | 3.1..15.1: CJK_Unified_Ideographs |
|---|
(The format of Blocks.txt changed after 3.0, from first; last; <block> to first..last; <block>.)
Broadened this issue to cover the wider issue of files changing format. Cleverness will be needed at some point.
Documentation needs a bit of work!
Blocks: There is provision for variant formats in the data files; look at IndexUnicodeProperties, eg:
# Unicode 13 moves kTotalStrokes to Unihan_IRGSources.txt.
# The line with the new location (i.e., the line without version number)
# must occur in this file before the line with the old location.
Unihan_DictionaryLikeData ; kTotalStrokes ; v12.1
For (unique) radical changes like that, it may be better to special case in the code. Ugg.
Identifier_Type: We could extend ExtraPropertyValueAliases.txt in the same way; a trailing Age means that age or below.