Dynamically accessing record elements using EXECUTE

I would like to write a function that can take in a table name and a pk value and make a copy of that row into the same table, but with a new value for the PK. 

I originally had a working function that would create a temp table like the passed-in table, including defaults, and I found the name of the pk column and set that column to default. Since we use nextval('...') as the default value for our PKs, it would assign the new PK before we copied the row back into the table.

However, we found that cloning ~2000 rows in one transaction did not work, because creating and dropping the temp table once for each cloned row caused postgres to run out of shared memory and abort the transaction (it said I might have to increase max_locks_per_transaction, but I didn't understand why).

So, I decided to try to write it using record variables instead, to see if that would fix the problem. Below is the function I came up with, but it does not work because the record variable my_row cannot be accessed from within the EXECUTE. 

How can I modify my function to assign the default value of the PK column into the corresponding column of the my_row record variable, and then insert the contents of that record back into the original table? 

Alternatively, how can I keep Postgres from running out of memory with the temp table method?

The function:

Thank you!
--