Rev 3730: Add a TODO discussing how our index requests should evolve. in http://bzr.arbash-meinel.com/branches/bzr/1.8-dev/lighter_log_file

[John Arbash Meinel]

At http://bzr.arbash-meinel.com/branches/bzr/1.8-dev/lighter_log_file

------------------------------------------------------------
revno: 3730
revision-id: john at arbash-meinel.com-20080921140124-kvoi3sdig2owjukf
parent: john at arbash-meinel.com-20080921135843-wnwpal216u3zt2ni
committer: John Arbash Meinel <john at arbash-meinel.com>
branch nick: lighter_log_file
timestamp: Sun 2008-09-21 09:01:24 -0500
message:
  Add a TODO discussing how our index requests should evolve.
-------------- next part --------------
=== modified file 'bzrlib/log.py'
--- a/bzrlib/log.py	2008-09-19 03:50:29 +0000
+++ b/bzrlib/log.py	2008-09-21 14:01:24 +0000
@@ -565,6 +565,10 @@
     # Looking up keys in batches of 1000 can cut the time in half, as well as
     # memory consumption. GraphIndex *does* like to look for a few keys in
     # parallel, it just doesn't like looking for *lots* of keys in parallel.
+    # TODO: This code needs to be re-evaluated periodically as we tune the
+    #       indexing layer. We might consider passing in hints as to the known
+    #       access pattern (sparse/clustered, high success rate/low success
+    #       rate). This particular access is clustered with a low success rate.
     get_parent_map = branch.repository.texts.get_parent_map
     modified_text_revisions = set()
     chunk_size = 1000