From: Jim Cain <jecain@...>
Date: 2007-04-03T04:30:33+09:00
Subject: Re: Using transactions with secondary indexes in bdb

On 4/2/07, ara.t.howard@noaa.gov <ara.t.howard@noaa.gov> wrote:
> there are transaction examples in the bdb src - have you looked at that?
>

I looked at the one example file "txn.rb" originally, but I didn't put
it all together then. Upon a second look, it's making a bit more
sense. In that example, he has:

module BDB
   class ThreadHash < Hash
      def start(val, num)
         Thread.start do
            self.env.begin(self) do |txn, b1|
               b1.delete_if do |k, v|
                  print "\t#{num} #{k} -- #{v}\n"
                  k == val
               end
               p "Thread #{num} Pass"
               Thread.pass
               b1.each do |k, v|
                  print "\t#{num} #{k} -- #{v}\n"
               end
               if num == 1
                  txn.abort
               else
                  txn.commit
               end
            end
            p "End Thread #{num}"
         end
      end
   end
end

And then later on he calls #start for each ThreadHash object. I
suppose the business part of that is:

self.env.begin(self) do |txn, b1|
# ...
end

So he's passing in the ThreadHash object as the only argument to
#begin, but the docs have this for Env#begin:

begin(flags = 0, db, ...) { |txn, db, ...| ...}

This says that "flags" is the first argument. Maybe that's where my
confusion started.

Cheers,
Jim