Add self.name method to abstract class about pg_party

[rkrage]

Hi @rogozin, thanks for reporting this.

Just curious, is there a specific reason you're calling model_name here? The in_partition method currently returns an ActiveRecord::Relation. The anonymous class wasn't really meant to be leaked and interacted with (outside of chaining query methods).

But I think you're right, delegating name to the parent class seems like a reasonable thing to do.

[rogozin]

Hi @rkrage, thanks for your answer 👍

Frankly I'm using activerecord-import with ON CONFILCT expression, so in fact I'm calling

TMainDatum.import huge_data_array, on_duplicate_key_update: { target_and_column_options } 

And of course it won't work because there are PG partition restrictions for on_conflict usage .

Eventually I'm targeting specific partition like this:

TMainDatum.in_partition('t_main_data_2017_34').import huge_data_array, on_duplicate_key_update: { target_and_column_options }

and it works well, but the only exception as I've described above :)

[rkrage]

Ah, okay. Honestly, I didn't think a whole lot about use cases for in_partition. Maybe it makes sense to change the API a little bit then - have it return the raw child class so users can interact with it directly?

I have a pull request up that fixes the bug (#14), but I'll try to revise it a bit to reflect this API change.

Sound good?

[rkrage]

Side note, if it sounds like in_partition will be a widely used method, we might want to consider adding a bit of validation (check that the partition is actually a child partition of the parent), and add some class caching (to reduce the amount of child classes that are dynamically created).

Sounds like that work should be in a separate PR though.

[rkrage]

@rogozin the fix should be included in version 0.6.0