Strange behaviour with ActiveRecord has_many collections, Rails 4 compared to Rails 3

rubyonrails-talk
1

Hi All,

I have been told to post discussion based topic on here instead of the GitHub Issues.

I thought this ‘could’ be a bug, but just need some clarification really. The issue I posted is here, does anyone have any thoughts?

Again: Strange behaviour with ActiveRecord has_many collections, Rails 4 compared to Rails 3. · Issue #12597 · rails/rails · GitHub

Thanks a lot.

2

To be honest it feels to me like what you were using just happened to work by accident, because by default method_missing on the association sets up a scope and calls the method on the class. If this was indeed accidental behaviour then the change could well be the side effect of another change. Using new in this way isn’t something I’ve come across in quite a few years of using rails.

Fred

3

Hi Fred,

Thanks for your reply.

“Using new in this way isn’t something I’ve come across in quite a few years of using rails.”

How would you initialise a new Task object given you have the @story instance available to you?

@task = Task.new(:story_id => @story.id) ?

If the above, I really don’t see the point of the new and build methods for a has_many association…

For a has_one association, I would be able to do:

@task = @story.build_task

So it makes no sense, that @task = @story.tasks.build doesn’t do the same thing… As ‘build’ seems to just be an alias of ‘new’.

Thanks.

4

I’ve always done @story.tasks.build(…)

The fact that this adds the built instance to the in memory collection has never been a problem for me

Fred

5

Hi Fred,

Thanks for this.

yes I can understand why it has never been a problem. It won’t be too much of a problem for us either, as the collection variable e.g. @tasks, would normally be filtered with .where() or be ordered with .order(), therefore, they would be different instances. But I logged it as a bug, because it just doesn’t seem right.

I can’t think of any reason why one would want the built instance added to the in memory collection…

Anyway, I think I’ve been a bit cheeky on the Github issue, not sure if I’m going to get another response!

Thanks for your reply though.