Belongs to polymorphic association assignment with new records doesn't update foreign id or type

Patch on github:

http://github.com/timcharper/rai...

Fixes the following scenarios:

• I have validates_inclusion_of on the type field for a polymorphic

belongs_to association. I assign a new record to the model's polymorphic

relationship of the proper type. validation fails because the type field

has not been updated.

• I replace the value for a polymorphic association to a new record of

another class. The type field still says its the previous class, and the

id field points to the previous record as well.

-----

[7:22pm] technoweenie: timcharper: it probably shouldnt set the record.id if record is new
[7:22pm] technoweenie: i'm not sure what record.id will be, nil I guess?  maybe it doesnt matter
[7:22pm] timcharper: yeah - that's what I'm thinking
[7:22pm] timcharper: it should at least clear it
[7:22pm] technoweenie: i guess in rare cases someone might pre-set the id  and would want it set
[7:22pm] timcharper: i was debating between that or a ternary
[7:23pm] technoweenie: record.new_record? ? nil : record.id
[7:23pm] timcharper: yeah
[7:23pm] timcharper: but was thinking the case would be in the minority that they set the id before saving the record
[7:23pm] timcharper: and if so, they probably did so intentionally, and would likely want the foreign key to receive that value as well
[7:24pm] technoweenie: yup
[7:24pm] technoweenie: as long as it doesnt blow up for the rest of us
[7:24pm] technoweenie: blow up, or set bad values
[7:25pm] timcharper: can you think of how it would ?  I can't think of anything
[7:25pm] technoweenie: nope
[7:25pm] technoweenie: i guess if you're a moron and try to set it wrong, you deserve what you get
[7:26pm] timcharper: yeah, that sounds about right
[7:26pm] timcharper: submitting the ticket
[7:26pm] technoweenie: and if its committed, and there is some wacky case where it blows up, it just means we were missing a regression test anyway