In general, you probably don't want to include both the definition of New Years and New Years observed (though you can but you have to be careful of the ordering if you add both... this is talked about in the documentation on Date::Manip::Holidays).

To define the true New Years, just add (to your holidays):

  Jan 1  = New Years
[download]

To define an observed New Years (where you observe it on Friday Dec 31 if Jan 1 falls on Saturday, or on Mon Jan 2 if Jan 1 falls on a Sunday), use the following:

   1*1:0:1:0:0:0*DWD               = New Year's Day
[download]

NOTE: this is straight from the documentation on Date::Manip::Holidays which is a better source of information on defining holidays than the Date::Manip::Recur document.

Other definitions are also possible, but these cover the most commonly desired definitions.

Excellent, the solution right from the author. Thanks, SBECK!

Unfortunately, it appears that I didn't get the right answer. Going back and rereading it (more carefully this time around), the desired date of New Years is:

  1/1  if 1/1 is Mon-Fri OR Sat
  1/2  if 1/1 is Sun
[download]

That's much more complicated. I found something that I thought would work...

 1*1:0:1:0:0:0*NBD,BD1,IBD,FD1   = New Year's Day (Saturday)
 1*1:0:1:0:0:0*NBD,BD1,NBD,FD2   = New Year's Day (Sunday observed Mon
+day)
 1*1:0:1:0:0:0*IBD               = New Year's Day (real)
[download]

but it is failing on some years. I actually think it might be a bug. I'll check into it tomorrow.

I have definitely decided that I need to add a couple more modifiers similar to IBD and NBD which will check if it is a given day of the week. That would make this much simpler. I hadn't been aware of the complexity of the holiday definition used here.