[ruby-core:44851] Re: help understanding weird bug

On 05/03 08:56, Rodrigo Rosenfeld Rosas wrote:
> Here is a sample application:
> 
> https://github.com/rosenfeld/weird-sequel-bug
> 
> And here is the strange behavior:
> 
> https://github.com/rosenfeld/weird-sequel-bug/blob/master/example.rb
> 
> children = Sq::Node[1].children
> p children.class
> parent = children.first.parent
> p children.map &:id
> 2.times { p children.map{|c| c.parent = parent; c.id} }
> 
> Output:
> 
> Array
> [2, 3]
> [2, 2]
> [3, 3]
> 
> For some strange reason, changing the parent of the element inside
> children.map{} block will modify the children array itself.
> 
> for example:
> 
> p children.map &:id
> children[0].parent = parent
> p children.map &:id
> 
> outputs:
> [2, 3]
> [3, 2]
> 
> Maybe this is a Sequel bug (I didn't look for the code of
> associations yet), but I'd like to understand how this could happen
> in Ruby.
 
This isn't a Sequel bug or a ruby bug.  When you use c.parent= in
Sequel, it will recognize that parent.children is the reciprocal, and
c.parent= will modify parent.children.

You issue is that you are modifying an array while you are iterating
over it, which is generally a bad idea.

Now, I can probably change it so that it doesn't make such a change
if you pass in a object that is the same as the existing value, but I
don't consider the current implementation incorrect.

> For example, how can I know if this is an ordinary Array or not?
> children.class is Array, so how can I differentiate it from a normal
> Array.

Sequel uses a plain array, it doesn't use a proxy object for
associations.  You could probably use an after_load hook to extend the
array with a module, though I'm not sure what it is you want to do.

> If I change the first line to the one below it works:
> # children = Sq::Node[1].children
> children = Sq::Node[1].children.map{|c| c}

Here you are creating a duplicate of the array, similar to:

  children = Sq::Node[1].children.dup

So you are not modifying the array you are iterating over, and hence,
you don't have that problem.

Jeremy