You can clone with
HTTPS or Subversion.
This fix here is just an update to the sql script for postgresql to pull in the table names with primary keys. This doesn't address the issue for composite primary keys though. It seems either the recommendation would be to not use them, which seems less than ideal, or to change the way these arguments are composed and parsed.
This sounds like a great idea - I can't read the diff though :/
I'm apparently making a mess of this thing. Initially it only shows the single line change, but after I commit it removes all 139 lines just to add a single line. This line below needs to be added under line 24. I'll keep trying.
and tc.constraint_type = 'PRIMARY KEY' \
Okay, I had to create a new pull request for this, but this time the diff is much cleaner.