Lighthouse has a new layout. Prefer the old one? Return to the old layout, and switch back any time from the link at the top of each page.

This project is archived and is in readonly mode.

Invalid has_one association causes parent object save to fail

#338

This ticket on the old Rails trac supplied a patch that changes the behaviour of has_one; now, if the associated record is invalid, the parent object will not save. This was supposedly to make it consistent with has_many and was introduced in this changeset

I'm am opening this ticket because I consider this a bug and an undocumented feature change and I'm struggling to understand why this was introduced; it has broken part of an application that I'm attempting to upgrade to Rails 2.1 and I have spent the best part of a few hours trying to track this down.

The reason I consider this a bug is because there is already a way of producing this behaviour explicitly using validates_associated. Why, if this can be done manually (when required) is this being done automatically?

The last post in the old Trac ticket demonstrates a catch-22 situation which is what lead to my own test failures. Here is an example that fails as a result of this new behaviour:

class Snippet
  has_one :wiki_entry
end

class WikiEntry
  belongs_to :snippet
  validates_presence_of :snippet_id <= this is the important bit!
end

s = Snippet.new
s.build_wiki_entry
s.save # => FAIL!

WikiEntry fails validation because it requires a snippet_id. It cannot get a snippet_id until it's parent Snippet has been saved. But we cannot save the parent Snippet because it's child WikIEntry is valid; catch-22.

Reported by Luke Redpath · June 5th, 2008 @ 11:18 AM

State: resolved
Milestone: none
Assigned to: Pratik Pratik
Importance: none

Activity

  1. Luke Redpath
    Luke Redpath

    One further note, this appears to be related to this ticket

    It appears some kind of fix has been checked in (a :validate option to disable this behaviour) but I still consider this to be a bug unless this behaviour is either disabled by default or properly documented.

    June 5th, 2008 @ 11:21 AM

  2. Drew Blas
    Drew Blas

    +1

    I'd like to see this removed entirely. We already had a perfectly good method for the validation. This undocumented and entirely unexpected behavior is much more bug than feature.

    June 9th, 2008 @ 10:35 PM

  3. James Cox
  4. Pratik
    Pratik
    • State changed from new to resolved

    http://github.com/rails/rails/co... disabled it by default.

    June 11th, 2008 @ 12:34 PM

  5. Luke Redpath
    Luke Redpath
    • Assigned user set to Pratik

    There seems to be some inconsistency in the documentation in that commit; sometimes it says false by default, sometimes it is true by default. Is this just a documentation error?

    June 11th, 2008 @ 01:23 PM

  6. Pratik
    Pratik

    It's false by default for has_one and belongs_to. Made another commits to fix docs/bheavior.

    Thanks

    June 11th, 2008 @ 01:33 PM

  7. Jeremy McNevin