[MERGE] Make bzr pack start optimising

[John Arbash Meinel]

John Arbash Meinel has voted tweak.
Status is now: Conditionally approved
Comment:
So, to summarize and make sure I understand the change correctly....

The bulk of the change is to change _copy_nodes_graph (and _do_copy...) 
to not
figure out how to copy the nodes, but to have that passed in as a
readv_group_iter.

And then you created "_least_readv_node_readv()" to build up a group 
iter that
minimizes the number of readvs you have to make, and had a 
_sort_revisions()
function that uses _least_* in the case of auto-packing (and plain pull, 
right?).
And gets subclassed for the Optmizing packer (which you also introduce).

I probably would prefer a different function name than 
"_sort_revisions".
Maybe: _revision_node_readv, just to match the others.


-    def _execute_pack_operations(self, pack_operations):
+    def _execute_pack_operations(self, pack_operations, 
_packer_class=Packer):
          """Execute a series of pack operations.

          :param pack_operations: A list of [revision_count, 
packs_to_combine].
+        :param _packer_class: The class of packer to use. If not 
supplied
+            Packer will be used.
          :return: None.
          """

^- :param _packer_class: The class of packer to use (default: Packer).


So a small doc change, and a better name for _sort_revisions (maybe at 
least
"sorted", since you are returning a sorted list...)

Otherwise, I'm happy to see this, and it is a good place to bring in 
other pack
optimizations.


For details, see: 
http://bundlebuggy.aaronbentley.com/request/%3C1196735765.15733.182.camel%40lifeless-64%3E