Identify the ordering of converter

visual/logical in .ucm, .cnv, api

• Thu Feb 17 13:51:08 2000 schererm changed notes
• Wed Jun 14 18:07:54 2000 srl changed notes
• Wed Jun 14 18:07:54 2000 srl moved from incoming to conversion
• Mon Dec 4 21:50:29 2000 schererm changed notes
• Mon Dec 4 21:50:29 2000 schererm moved from conversion to feature
• 03/19/02 20:37:57 mark moved from feature to conversion
• 07/31/02 18:44:00 schererm changed notes
• 10/29/02 13:26:38 hshih changed notes2
• 10/29/02 13:26:38 hshih changed notes
• 10/30/02 14:32:07 hshih changed notes2
• 10/30/02 14:32:07 hshih changed notes
• 12/26/02 13:09:08 hshih changed notes2
• 12/26/02 13:09:08 hshih changed notes
• 12/26/02 13:09:56 hshih changed notes2
• 12/26/02 13:09:56 hshih changed notes
• 01/06/03 19:10:24 schererm changed notes
• 01/13/03 00:44:38 grhoten changed notes2
• 05/29/03 15:53:40 hshih changed notes2
• 01/30/04 02:34:05 grhoten changed notes2
• 07/09/04 15:10:45 grhoten changed notes2
• 07/19/04 19:02:31 schererm changed notes2

ucnv_getUnicodeSet does provide a way to get the set of characters in a charset. If the charset doesn't have any right-to-left characters, it's safe to say that the charset is always left-to-right. It's also safe to say that special bidi processing isn't needed when converting between the various Unicode forms.

This type of information is helpful to note when bidi should be performed.

Unfortunately, the table data for charsets like ISO-8859-6, ISO-8859-6-I and ISO-8859-6-E is the same for all three charsets. The issue is the same for ISO-8859-8*. An API is needed to differentiate the visual versus logical order between those aliases. Maybe this could be a converter option that can be used in the alias table.

Take the following as an example:

ibm-1089_P100-1995 { UTR22* }
                        ibm-1089 { IBM* JAVA }
                        ISO-8859-6 { MIME* IANA JAVA* }
                        ISO-8859-6-I { IANA MIME }
                        ISO-8859-6-E { IANA MIME }

This could turn into the following:

ibm-1089_P100-1995 { UTR22* }
                        ibm-1089 { IBM* JAVA }
                        ISO-8859-6 { MIME* IANA JAVA* }
ibm-1089_P100-1995,bidi=e
                        ISO-8859-6-E { IANA* MIME* }
ibm-1089_P100-1995,bidi=i
                        ISO-8859-6-I { IANA* MIME* }

This could provide information to the user through a new function. Maybe it could be called ucnv_getBidiType(). The API could tell users what type of special bidi post-processing or pre-processing is needed depending on the type of conversion.

This is just one possible solution.

IBM's CDRA also has information on bidi processing for specific CCSIDs. That information should be reviewed too.