2 from bup import _helpers
3 from bup.helpers import *
5 BLOB_MAX = 8192*4 # 8192 is the "typical" blob size for bupsplit
6 BLOB_READ_SIZE = 1024*1024
8 progress_callback = None
11 GIT_MODE_FILE = 0100644
12 GIT_MODE_TREE = 040000
13 GIT_MODE_SYMLINK = 0120000
14 assert(GIT_MODE_TREE != 40000) # 0xxx should be treated as octal
16 # The purpose of this type of buffer is to avoid copying on peek(), get(),
17 # and eat(). We do copy the buffer contents on put(), but that should
18 # be ok if we always only put() large amounts of data at a time.
26 self.data = buffer(self.data, self.start) + s
29 def peek(self, count):
30 return buffer(self.data, self.start, count)
36 v = buffer(self.data, self.start, count)
41 return len(self.data) - self.start
44 def readfile_iter(files, progress=None):
45 for filenum,f in enumerate(files):
50 progress(filenum, len(b))
51 b = f.read(BLOB_READ_SIZE)
58 def _splitbuf(buf, basebits, fanbits):
60 b = buf.peek(buf.used())
61 (ofs, bits) = _helpers.splitbuf(b)
67 level = (bits-basebits)//fanbits # integer division
69 yield buffer(b, 0, ofs), level
72 while buf.used() >= BLOB_MAX:
74 yield buf.get(BLOB_MAX), 0
77 def _hashsplit_iter(files, progress):
78 assert(BLOB_READ_SIZE > BLOB_MAX)
79 basebits = _helpers.blobbits()
80 fanbits = int(math.log(fanout or 128, 2))
82 for inblock in readfile_iter(files, progress):
84 for buf_and_level in _splitbuf(buf, basebits, fanbits):
87 yield buf.get(buf.used()), 0
90 def _hashsplit_iter_keep_boundaries(files, progress):
91 for real_filenum,f in enumerate(files):
93 def prog(filenum, nbytes):
94 # the inner _hashsplit_iter doesn't know the real file count,
95 # so we'll replace it here.
96 return progress(real_filenum, nbytes)
99 for buf_and_level in _hashsplit_iter([f], progress=prog):
103 def hashsplit_iter(files, keep_boundaries, progress):
105 return _hashsplit_iter_keep_boundaries(files, progress)
107 return _hashsplit_iter(files, progress)
111 def split_to_blobs(makeblob, files, keep_boundaries, progress):
113 for (blob, level) in hashsplit_iter(files, keep_boundaries, progress):
115 total_split += len(blob)
116 if progress_callback:
117 progress_callback(len(blob))
118 yield (sha, len(blob), level)
121 def _make_shalist(l):
124 total = sum(size for mode,sha,size, in l)
125 vlen = len('%x' % total)
127 for (mode, sha, size) in l:
128 shalist.append((mode, '%0*x' % (vlen,ofs), sha))
131 return (shalist, total)
134 def _squish(maketree, stacks, n):
136 while i < n or len(stacks[i]) >= MAX_PER_TREE:
137 while len(stacks) <= i+1:
139 if len(stacks[i]) == 1:
140 stacks[i+1] += stacks[i]
142 (shalist, size) = _make_shalist(stacks[i])
143 tree = maketree(shalist)
144 stacks[i+1].append((GIT_MODE_TREE, tree, size))
149 def split_to_shalist(makeblob, maketree, files,
150 keep_boundaries, progress=None):
151 sl = split_to_blobs(makeblob, files, keep_boundaries, progress)
155 for (sha,size,level) in sl:
156 shal.append((GIT_MODE_FILE, sha, size))
157 return _make_shalist(shal)[0]
160 for (sha,size,level) in sl:
161 stacks[0].append((GIT_MODE_FILE, sha, size))
162 _squish(maketree, stacks, level)
163 #log('stacks: %r\n' % [len(i) for i in stacks])
164 _squish(maketree, stacks, len(stacks)-1)
165 #log('stacks: %r\n' % [len(i) for i in stacks])
166 return _make_shalist(stacks[-1])[0]
169 def split_to_blob_or_tree(makeblob, maketree, files,
170 keep_boundaries, progress=None):
171 shalist = list(split_to_shalist(makeblob, maketree,
172 files, keep_boundaries, progress))
173 if len(shalist) == 1:
174 return (shalist[0][0], shalist[0][2])
175 elif len(shalist) == 0:
176 return (GIT_MODE_FILE, makeblob(''))
178 return (GIT_MODE_TREE, maketree(shalist))
181 def open_noatime(name):
182 fd = _helpers.open_noatime(name)
184 return os.fdopen(fd, 'rb', 1024*1024)