Ability to override cascade_copy when calling Row->copy

[ghost]

Simple modification adding an optional argument to the Class::Row->copy method allowing you to force overriding of individual relationship's cascade_copy attribute.

[ilmari]

This is a very limited (and limiting, extensibility-wise) API. I suggest a hash ref with { cascade => { rel_one => 1, rel_two => 0 } }, so you can override both ways (rels not mentioned in the cascade hash maintain the value specified in their definition) and add more options in future.

[ilmari]

Thanks for the contribution. However, see my comment on the API above. Also, could you please add some tests for the new behaviour?

[ghost]

Thanks for the feedback. I'll work on implementing the suggested changes when possible. Likely this weekend.

[ilmari]

After discussing this on IRC with @ribasushi, I realised that to work generally in DBIC, it the data structure would have to be able to specify whether to cascade the copy recursively for all relationships, which gets more complicated. It'd have to be something like this, where the value gets passed on to the related ->copy call if it's a hashref:

{
    cascade => {
        rel_one => {            # implicitly true
            nested_rel => 0,
        },
        rel_two => 0,            # can't nest off false
        rel_three => 1,          # further rels get default
    },
}

If you don't have the time or inclination to do all this work (which is perfectly okay), I suggest implementing just what you need as a component that overrides ->copy and localises the cascade_copy attributes as needed.

[ghost]

I'll still take a stab at it. Wanted to get to it this weekend but I was recovering from jet lag.

[ribasushi]

@mikegrb What @ilmari didn't mention was that even with a well though-through API this use-case is too specific to be a part of the DBIC core. That is, while the code in either case is identical, the feature itself should be available in a DBIx::Class:: via an override of copy(). Then a user that needs this would load it in the base Result class via load_components() and be on their way.

Thus I am going to close this PR for the time being. Let me know if you have further question. In particular if you disagree with the above PLEASE don't hesitate to reopen this PR and articulate your point further.