Question

I have a prepared insert statement in Sequel (using Oracle).

prepared_statement = DB[:table_name].prepare(:insert, :name, :value=>:$value)

When I call it the row gets added just fine.

prepared_statement.call :value=>'Some value'

I have a trigger and a sequence set up so the ID will be auto generated. I would like to get back the row (or the id) I just added, but I can’t see how. I can’t use insert because value is a CLOB and may be greater than 4000 characters.

Answer 1

In JRuby, using the JDBC adapter you can override the insert and pass in the returning clause. The tricky part is that you don’t always know what the primary key is at this level so you may have to use ROWID or request all of the columns back.

You end up with something that looks similar to this:

module Sequel
  module JDBC
    class Database
      def execute_insert_with_returning(conn, sql, opts = {})        
        columns = opts[:key_columns] || ["ROWID"]
        q = "{ call #{sql} returning #{columns.join(',')} into #{columns.collect {|_| '?'}.join(',')} }"
        stmt = conn.prepare_call(q)
        raise "Unable to prepare call for insert" if stmt.nil?

        begin
          columns.each_with_index do |_, index|
            stmt.registerOutParameter(index+1, JavaSQL::Types::VARCHAR)
          end
          return nil if 0 == stmt.executeQuery

          values = (1..columns.count).inject({}) do |memo, index|
            key = columns[index-1].downcase.to_sym rescue nil
            memo[key] = stmt.get_string(index) unless key.nil?
            memo
          end
          values
        ensure
          stmt.close
        end
      end # #execute_insert_with_returning

      alias execute_without_specialized_insert execute
      def execute(sql, opts={}, &block)
        if opts[:type] == :insert
          synchronize(opts[:server]) do |conn|
            execute_insert_with_returning conn, sql, opts
          end
        else
          execute_without_specialized_insert sql, opts, &block
        end
      end # #execute
    end # Database
  end # JDBC
end # Sequel

I’ve done something pretty much like this and it works pretty good. I think we had to override the Sequel::Model as well so it passes the primary key in as opts[:key_columns] but I may be remembering incorrectly.

This is a bit of a load bearing kludge that gets the job done. It would be more elegant to specialize it to the Oracle JDBC adapter and to ensure that all of the error handling code is present from the original execute statement. Given the time I’d love to get something better and give it back to the Sequel project.