During some post 3.2 release testing we have determined that if you are upgrading a GLASS database from 3.1.x to 3.2.x, you need to drop and rebuild String-based indexes as well as evaluate any string-based sorted collections.
As part of improving our Unicode support in 3.2 we determined that it was not possible to sort/compare instances of *String classes with instances of Unicode* ... UNLESS ... we used a collator for all sorting and comparisons... see pages 44-45 of the 3.2 release notes[1]
Because the majority of GLASS applications use Seaside and Unicode strings are commonplace and even desired, I decided that during the Seaside upgrade process, the repository would be set to use Unicode comparison mode.
Unicode comparison mode means that all comparison and sorting for all CharacterCollection classes is done using the default collator.
Depending upon the codepoints of the strings that you are sorting/comparing and the collator that you are using the collation sequence and even equality status can change (yes, there are certain codepoints and sequences that can affect equality).
Rather than quibble about which sequences are safe, it is simply prudent to rebuild all String based indexes and SortedCollections.
If for some reason, it is not practical to rebuild indexes and SortedCollections there are alternative approaches than you can try that might be preferable but they are not without their drawbacks ... If anyone feels they are in this position, we can discuss these other alternatives in more detail...
Dale
_______________________________________________
Glass mailing list
[hidden email]
http://lists.gemtalksystems.com/mailman/listinfo/glass