postgres collation differences. osx v ubuntu

Question

So, i\'ve recently come to realize that collation is a huge deal on postgres, and that many comments refer to OSX / locale support as \"broken\", which hasn\'t enlightened me. f

Answer 1

In the Default Unicode Collation Element Table you can see these two entries:

002C  ; [*0220.0020.0002] # COMMA
002D  ; [*020D.0020.0002] # HYPHEN-MINUS

Here, the primary weight of COMMA is greater than the primary weight of HYPHEN-MINUS, therefore HYPHEN-MINUS sorts before COMMA.

Note that this is the expected sort order according to the Unicode Collation Algorithm with the default weights. If you expect sort order by ASCII byte values, you get a different order. And there are other valid orders. But if the locale is named "en_US.UTF-8" (or "en_US.utf8", same thing), then you'd probably expect Unicode order. But that's between you and your operating system vendor.