from __future__ import absolute_import, print_function
from collections import namedtuple
-from errno import ELOOP, ENOENT, ENOTDIR
+from errno import EINVAL, ELOOP, ENOENT, ENOTDIR
from itertools import chain, dropwhile, groupby, tee
from random import randrange
from stat import S_IFDIR, S_IFLNK, S_IFREG, S_ISDIR, S_ISLNK, S_ISREG
_, obj_t, size = next(it)
return ofs + sum(len(b) for b in it)
+def _skip_chunks_before_offset(tree, offset):
+ prev_ent = next(tree, None)
+ if not prev_ent:
+ return tree
+ ent = None
+ for ent in tree:
+ ent_ofs = int(ent[1], 16)
+ if ent_ofs > offset:
+ return chain([prev_ent, ent], tree)
+ if ent_ofs == offset:
+ return chain([ent], tree)
+ prev_ent = ent
+ return [prev_ent]
+
def _tree_chunks(repo, tree, startofs):
"Tree should be a sequence of (name, mode, hash) as per tree_decode()."
assert(startofs >= 0)
# name is the chunk's hex offset in the original file
- tree = dropwhile(lambda (_1, name, _2): int(name, 16) < startofs, tree)
- for mode, name, oid in tree:
+ for mode, name, oid in _skip_chunks_before_offset(tree, startofs):
ofs = int(name, 16)
skipmore = startofs - ofs
if skipmore < 0:
return self._size
def seek(self, ofs):
- if ofs < 0:
- raise IOError(errno.EINVAL, 'Invalid argument')
- if ofs > self._compute_size():
- raise IOError(errno.EINVAL, 'Invalid argument')
+ if ofs < 0 or ofs > self._compute_size():
+ raise IOError(EINVAL, 'Invalid seek offset: %d' % ofs)
self.ofs = ofs
def tell(self):
return self.ofs
def read(self, count=-1):
+ size = self._compute_size()
+ if self.ofs >= size:
+ return ''
if count < 0:
- count = self._compute_size() - self.ofs
+ count = size - self.ofs
if not self.reader or self.reader.ofs != self.ofs:
self.reader = _ChunkReader(self._repo, self.oid, self.ofs)
try:
Item = namedtuple('Item', ('meta', 'oid'))
Chunky = namedtuple('Chunky', ('meta', 'oid'))
+FakeLink = namedtuple('FakeLink', ('meta', 'target'))
Root = namedtuple('Root', ('meta'))
Tags = namedtuple('Tags', ('meta'))
RevList = namedtuple('RevList', ('meta', 'oid'))
def is_valid_cache_key(x):
"""Return logically true if x looks like it could be a valid cache key
(with respect to structure). Current valid cache entries:
- (repo-id, path, parent, want_meta, dref) -> resolution
- commit_oid -> commit
- commit_oid + ':r' -> rev-list
- i.e. rev-list -> {'.', commit, '2012...', next_commit, ...}
+ res:... -> resolution
+ itm:OID -> Commit
+ rvl:OID -> {'.', commit, '2012...', next_commit, ...}
"""
# Suspect we may eventually add "(container_oid, name) -> ...", and others.
x_t = type(x)
- if x_t is tuple:
- return len(x) == 5
if x_t is bytes:
- if len(x) == 20:
+ tag = x[:4]
+ if tag in ('itm:', 'rvl:') and len(x) == 24:
return True
- if len(x) == 22 and x.endswith(b':r'):
+ if tag == 'res:':
return True
def cache_get(key):
global _cache
- assert is_valid_cache_key(key)
+ if not is_valid_cache_key(key):
+ raise Exception('invalid cache key: ' + repr(key))
return _cache.get(key)
def cache_notice(key, value):
global _cache, _cache_keys, _cache_max_items
- assert is_valid_cache_key(key)
+ if not is_valid_cache_key(key):
+ raise Exception('invalid cache key: ' + repr(key))
if key in _cache:
return
if len(_cache) < _cache_max_items:
When need_meta is true don't return a cached item that only has a
mode."""
# tree might be stored independently, or as '.' with its entries.
- item = cache_get(oid)
+ commit_key = b'itm:' + oid
+ item = cache_get(commit_key)
if item:
if not need_meta:
return item
if isinstance(item.meta, Metadata):
return item
- entries = cache_get(oid + b':r')
+ entries = cache_get(b'rvl:' + oid)
if entries:
return entries['.']
target = item.meta.symlink_target
if target:
return target
+ elif isinstance(item, FakeLink):
+ return item.target
return _readlink(repo, item.oid)
def _compute_item_size(repo, item):
meta = _find_treeish_oid_metadata(repo, commit.oid)
if meta:
commit = commit._replace(meta=meta)
- cache_notice(oid, commit)
+ commit_key = b'itm:' + oid
+ cache_notice(commit_key, commit)
return commit
def _revlist_item_from_oid(repo, oid, require_meta):
if item:
return item
item = Commit(meta=default_dir_mode, oid=tree_oid, coid=coid)
- cache_notice(item.coid, item)
+ commit_key = b'itm:' + coid
+ cache_notice(commit_key, item)
return item
def cache_commit(repo, oid):
revs = None # Don't disturb the tees
rev_names = _reverse_suffix_duplicates(_name_for_rev(x) for x in rev_names)
rev_items = (_item_for_rev(x) for x in rev_items)
- latest = None
+ tip = None
for item in rev_items:
- latest = latest or item
name = next(rev_names)
+ tip = tip or (name, item)
entries[name] = item
- entries['latest'] = latest
- cache_notice(latest.coid + b':r', entries)
+ entries['latest'] = FakeLink(meta=default_symlink_mode, target=tip[0])
+ revlist_key = b'rvl:' + tip[1].coid
+ cache_notice(revlist_key, entries)
return entries
def revlist_items(repo, oid, names):
# For now, don't worry about the possibility of the contents being
# "too big" for the cache.
- entries = cache_get(oid + b':r')
+ revlist_key = b'rvl:' + oid
+ entries = cache_get(revlist_key)
if not entries:
entries = cache_commit(repo, oid)
yield x
def _resolve_path(repo, path, parent=None, want_meta=True, deref=False):
- cache_key = (repo.id(), tuple(path), parent, bool(want_meta), bool(deref))
+ cache_key = b'res:%d%d%d:%s\0%s' \
+ % (bool(want_meta), bool(deref), repo.id(),
+ ('/'.join(x[0] for x in parent) if parent else ''),
+ '/'.join(path))
resolution = cache_get(cache_key)
if resolution:
return resolution
meta.mode = m
meta.uid = meta.gid = meta.atime = meta.mtime = meta.ctime = 0
if S_ISLNK(m):
- target = _readlink(repo, item.oid)
+ if isinstance(item, FakeLink):
+ target = item.target
+ else:
+ target = _readlink(repo, item.oid)
meta.symlink_target = target
meta.size = len(target)
elif include_size:
projects / bup.git / blobdiff