mercurial-scm/hg: mercurial/revlog.py comparison

comparison mercurial/revlog.py @ 1680:c21b54f7f7b8

Merge with crew

author	Benoit Boissinot <benoit.boissinot@ens-lyon.org>
date	Wed, 01 Feb 2006 19:18:15 +0100
parents	b345cc4c22c0 daff3ef0de8d
children

comparison

equal deleted inserted replaced

-:675ca845c2f8
+:c21b54f7f7b8
 """
 self.indexfile = indexfile
 self.datafile = datafile
 self.opener = opener
 self.cache = None
+self.chunkcache = None
 try:
 i = self.opener(self.indexfile).read()
 except IOError, inst:
 if inst.errno != errno.ENOENT:
 raise
 i = ""
+if i and i[:4] != "\0\0\0\0":
+raise RevlogError(_("incompatible revlog signature on %s") %
+self.indexfile)
 if len(i) > 10000:
 # big index, let's parse it on demand
 parser = lazyparser(i, self)
 self.index = lazyindex(parser)
 l = len(i) / s
 self.index = [None] * l
 m = [None] * l
 n = 0
-for f in xrange(0, len(i), s):
+for f in xrange(0, l * s, s):
 # offset, size, base, linkrev, p1, p2, nodeid
 e = struct.unpack(indexformat, i[f:f + s])
 m[n] = (e[6], n)
 self.index[n] = e
 n += 1
 def patches(self, t, pl):
 """apply a list of patches to a string"""
 return mdiff.patches(t, pl)
+def chunk(self, rev):
+start, length = self.start(rev), self.length(rev)
+end = start + length
+def loadcache():
+cache_length = max(4096 * 1024, length) # 4Mo
+df = self.opener(self.datafile)
+df.seek(start)
+self.chunkcache = (start, df.read(cache_length))
+if not self.chunkcache:
+loadcache()
+cache_start = self.chunkcache[0]
+cache_end = cache_start + len(self.chunkcache[1])
+if start >= cache_start and end <= cache_end:
+# it is cached
+offset = start - cache_start
+else:
+loadcache()
+offset = 0
+#def checkchunk():
+#    df = self.opener(self.datafile)
+#    df.seek(start)
+#    return df.read(length)
+#assert s == checkchunk()
+return decompress(self.chunkcache[1][offset:offset + length])
 def delta(self, node):
 """return or calculate a delta between a node and its predecessor"""
 r = self.rev(node)
 b = self.base(r)
 if r == b:
 return self.diff(self.revision(self.node(r - 1)),
 self.revision(node))
 else:
-f = self.opener(self.datafile)
+return self.chunk(r)
-f.seek(self.start(r))
-data = f.read(self.length(r))
-return decompress(data)
 def revision(self, node):
 """return an uncompressed revision of a given"""
 if node == nullid: return ""
 if self.cache and self.cache[0] == node: return self.cache[2]
 # look up what we need to read
 text = None
 rev = self.rev(node)
-start, length, base, link, p1, p2, node = self.index[rev]
+base = self.base(rev)
-end = start + length
-if base != rev: start = self.start(base)
 # do we have useful data cached?
 if self.cache and self.cache[1] >= base and self.cache[1] < rev:
 base = self.cache[1]
-start = self.start(base + 1)
 text = self.cache[2]
-last = 0
+else:
+text = self.chunk(base)
-f = self.opener(self.datafile)
-f.seek(start)
-data = f.read(end - start)
-if text is None:
-last = self.length(base)
-text = decompress(data[:last])
 bins = []
 for r in xrange(base + 1, rev + 1):
-s = self.length(r)
+bins.append(self.chunk(r))
-bins.append(decompress(data[last:last + s]))
-last = last + s
 text = mdiff.patches(text, bins)
+p1, p2 = self.parents(node)
 if node != hash(text, p1, p2):
 raise RevlogError(_("integrity check failed on %s:%d")
 % (self.datafile, rev))
 self.cache = (node, rev, text)
 gy = y.next()
 else:
 #print "next x"
 gx = x.next()
-def group(self, nodelist, lookup, infocollect = None):
+def group(self, nodelist, lookup, infocollect=None):
 """calculate a delta group
 Given a list of changeset revs, return a set of deltas and
 metadata corresponding to nodes. the first delta is
 parent(nodes[0]) -> nodes[0] the receiver is guaranteed to
 have this parent as it has all history before these
 changesets. parent is parent[0]
 """
 revs = [self.rev(n) for n in nodelist]
-needed = dict.fromkeys(revs, 1)
 # if we don't have any revisions touched by these changesets, bail
 if not revs:
 yield struct.pack(">l", 0)
 return
 # add the parent of the first rev
 p = self.parents(self.node(revs[0]))[0]
 revs.insert(0, self.rev(p))
-# for each delta that isn't contiguous in the log, we need to
-# reconstruct the base, reconstruct the result, and then
-# calculate the delta. We also need to do this where we've
-# stored a full version and not a delta
-for i in xrange(0, len(revs) - 1):
-a, b = revs[i], revs[i + 1]
-if a + 1 != b or self.base(b) == b:
-for j in xrange(self.base(a), a + 1):
-needed[j] = 1
-for j in xrange(self.base(b), b + 1):
-needed[j] = 1
-# calculate spans to retrieve from datafile
-needed = needed.keys()
-needed.sort()
-spans = []
-oo = -1
-ol = 0
-for n in needed:
-if n < 0: continue
-o = self.start(n)
-l = self.length(n)
-if oo + ol == o: # can we merge with the previous?
-nl = spans[-1][2]
-nl.append((n, l))
-ol += l
-spans[-1] = (oo, ol, nl)
-else:
-oo = o
-ol = l
-spans.append((oo, ol, [(n, l)]))
-# read spans in, divide up chunks
-chunks = {}
-for span in spans:
-# we reopen the file for each span to make http happy for now
-f = self.opener(self.datafile)
-f.seek(span[0])
-data = f.read(span[1])
-# divide up the span
-pos = 0
-for r, l in span[2]:
-chunks[r] = decompress(data[pos: pos + l])
-pos += l
 # helper to reconstruct intermediate versions
 def construct(text, base, rev):
-bins = [chunks[r] for r in xrange(base + 1, rev + 1)]
+bins = [self.chunk(r) for r in xrange(base + 1, rev + 1)]
 return mdiff.patches(text, bins)
 # build deltas
-deltas = []
 for d in xrange(0, len(revs) - 1):
 a, b = revs[d], revs[d + 1]
-n = self.node(b)
+na = self.node(a)
+nb = self.node(b)
 if infocollect is not None:
-infocollect(n)
+infocollect(nb)
 # do we need to construct a new delta?
 if a + 1 != b or self.base(b) == b:
-if a >= 0:
+ta = self.revision(na)
-base = self.base(a)
+tb = self.revision(nb)
-ta = chunks[self.base(a)]
-ta = construct(ta, base, a)
-else:
-ta = ""
-base = self.base(b)
-if a > base:
-base = a
-tb = ta
-else:
-tb = chunks[self.base(b)]
-tb = construct(tb, base, b)
 d = self.diff(ta, tb)
 else:
-d = chunks[b]
+d = self.chunk(b)
-p = self.parents(n)
+p = self.parents(nb)
-meta = n + p[0] + p[1] + lookup(n)
+meta = nb + p[0] + p[1] + lookup(nb)
 l = struct.pack(">l", len(meta) + len(d) + 4)
 yield l
 yield meta
 yield d
 def checksize(self):
 expected = 0
 if self.count():
 expected = self.end(self.count() - 1)
 try:
 f = self.opener(self.datafile)
 f.seek(0, 2)
 actual = f.tell()
-return expected - actual
+dd = actual - expected
 except IOError, inst:
-if inst.errno == errno.ENOENT:
+if inst.errno != errno.ENOENT:
-return 0
+raise
-raise
+dd = 0
+try:
+f = self.opener(self.indexfile)
+f.seek(0, 2)
+actual = f.tell()
+s = struct.calcsize(indexformat)
+i = actual / s
+di = actual - (i * s)
+except IOError, inst:
+if inst.errno != errno.ENOENT:
+raise
+di = 0
+return (dd, di)

Mercurial > public > mercurial-scm > hg

comparison mercurial/revlog.py @ 1680:c21b54f7f7b8