At http://bzr.arbash-meinel.com/branches/bzr/1.10-dev/branch_startup
------------------------------------------------------------
revno: 3825
revision-id: [EMAIL PROTECTED]
parent: [EMAIL PROTECTED]
committer: John Arbash Meinel <[EMAIL PROTECTED]>
branch nick: branch_startup
timestamp: Fri 2008-11-07 17:34:46 -0600
message:
It turns out that we read the pack-names file 3-times because
of inefficiencies in the BTree logic.
1) iter_all_entries() reads the root node, and then proceeds to read all
leaf nodes, but for pack-names, there is only the root node.
2) _read_nodes() reads the whole file when it doesn't know the page size,
but only does so to get the size, and then continues on to read the
requested nodes. Instead, just use what we read.
=== modified file 'bzrlib/btree_index.py'
--- a/bzrlib/btree_index.py 2008-10-29 19:24:01 +0000
+++ b/bzrlib/btree_index.py 2008-11-07 23:34:46 +0000
@@ -879,6 +879,15 @@
"iter_all_entries scales with size of history.")
if not self.key_count():
return
+ if self._row_offsets[-1] == 1:
+ # There is only the root node, and we read that via key_count()
+ if self.node_ref_lists:
+ for key, (value, refs) in sorted(self._root_node.keys.items()):
+ yield (self, key, value, refs)
+ else:
+ for key, (value, refs) in sorted(self._root_node.keys.items()):
+ yield (self, key, value)
+ return
start_of_leaves = self._row_offsets[-2]
end_of_leaves = self._row_offsets[-1]
needed_nodes = range(start_of_leaves, end_of_leaves)
@@ -1246,6 +1255,7 @@
:param nodes: The nodes to read. 0 - first node, 1 - second node etc.
:return: None
"""
+ bytes = None
ranges = []
for index in nodes:
offset = index * _PAGE_SIZE
@@ -1255,11 +1265,9 @@
if self._size:
size = min(_PAGE_SIZE, self._size)
else:
- stream = self._transport.get(self._name)
- start = stream.read(_PAGE_SIZE)
- # Avoid doing this again
- self._size = len(start)
- size = min(_PAGE_SIZE, self._size)
+ bytes = self._transport.get_bytes(self._name)
+ self._size = len(bytes)
+ break
else:
if offset > self._size:
raise AssertionError('tried to read past the end'
@@ -1267,15 +1275,19 @@
% (offset, self._size))
size = min(size, self._size - offset)
ranges.append((offset, size))
- if not ranges:
- return
- if self._file is None:
- data_ranges = self._transport.readv(self._name, ranges)
+ if bytes is not None:
+ data_ranges = [(start, bytes[start:start+_PAGE_SIZE])
+ for start in xrange(0, len(bytes), _PAGE_SIZE)]
else:
- data_ranges = []
- for offset, size in ranges:
- self._file.seek(offset)
- data_ranges.append((offset, self._file.read(size)))
+ if not ranges:
+ return
+ if self._file is None:
+ data_ranges = self._transport.readv(self._name, ranges)
+ else:
+ data_ranges = []
+ for offset, size in ranges:
+ self._file.seek(offset)
+ data_ranges.append((offset, self._file.read(size)))
for offset, data in data_ranges:
if offset == 0:
# extract the header
--
bazaar-commits mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/bazaar-commits
