# Copyright 2002 Ben Escoto # # This file is part of duplicity. # # Duplicity is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License as published by the # Free Software Foundation; either version 3 of the License, or (at your # option) any later version. # # Duplicity is distributed in the hope that it will be useful, but # WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU # General Public License for more details. # # You should have received a copy of the GNU General Public License # along with duplicity; if not, write to the Free Software Foundation, # Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA """Functions for producing signatures and deltas of directories Note that the main processes of this module have two parts. In the first, the signature or delta is constructed of a ROPath iterator. In the second, the ROPath iterator is put into tar block form. """ from __future__ import generators import cStringIO, re, types import tarfile, librsync, log, statistics from path import * from lazy import * # A StatsObj will be written to this whenever DirDelta_WriteSig is # run. stats = None class DiffDirException(Exception): pass def DirSig(path_iter): """Alias for SigTarBlockIter below""" return SigTarBlockIter(path_iter) def DirFull(path_iter): """Return a tarblock full backup of items in path_iter A full backup is just a diff starting from nothing (it may be less elegant than using a standard tar file, but we can be sure that it will be easy to split up the tar and make the volumes the same sizes). """ return DirDelta(path_iter, cStringIO.StringIO("")) def DirFull_WriteSig(path_iter, sig_outfp): """Return full backup like above, but also write signature to sig_outfp""" return DirDelta_WriteSig(path_iter, cStringIO.StringIO(""), sig_outfp) def DirDelta(path_iter, dirsig_fileobj_list): """Produce tarblock diff given dirsig_fileobj_list and pathiter dirsig_fileobj_list should either be a tar fileobj or a list of those, sorted so the most recent is last. """ if type(dirsig_fileobj_list) is types.ListType: sig_iter = combine_path_iters(map(sigtar2path_iter, dirsig_fileobj_list)) else: sig_iter = sigtar2path_iter(dirsig_fileobj_list) return DeltaTarBlockIter(get_delta_iter(path_iter, sig_iter)) def delta_iter_error_handler(exc, new_path, sig_path, sig_tar = None): """Called by get_delta_iter, report error in getting delta""" if new_path: index_string = new_path.get_relative_path() elif sig_path: index_string = sig_path.get_relative_path() else: assert 0, "Both new and sig are None for some reason" log.Log("Error %s getting delta for %s" % (str(exc), index_string), 2) return None def get_delta_path(new_path, sig_path): """Get one delta_path, or None if error""" delta_path = new_path.get_ropath() if not new_path.isreg(): delta_path.difftype = "snapshot" elif not sig_path or not sig_path.isreg(): delta_path.difftype = "snapshot" delta_path.setfileobj(new_path.open("rb")) else: # both new and sig exist and are regular files assert sig_path.difftype == "signature" delta_path.difftype = "diff" sigfp, newfp = sig_path.open("rb"), new_path.open("rb") delta_path.setfileobj(librsync.DeltaFile(sigfp, newfp)) new_path.copy_attribs(delta_path) delta_path.stat.st_size = new_path.stat.st_size return delta_path def log_delta_path(delta_path, new_path = None, stats = None): """Look at delta path and log delta. Add stats if new_path is set""" if delta_path.difftype == "snapshot": if new_path: stats.add_new_file(new_path) log.Log("Generating delta - new file: %s" % (delta_path.get_relative_path(),), 5) else: if new_path: stats.add_changed_file(new_path) log.Log("Generating delta - changed file: %s" % (delta_path.get_relative_path(),), 5) def get_delta_iter(new_iter, sig_iter): """Generate delta iter from new Path iter and sig Path iter. For each delta path of regular file type, path.difftype with be set to "snapshot", "diff". sig_iter will probably iterate ROPaths instead of Paths. """ collated = collate2iters(new_iter, sig_iter) for new_path, sig_path in collated: log.Log("Comparing %s and %s" % (new_path and new_path.index, sig_path and sig_path.index), 6) if (not new_path or not new_path.type) and sig_path and sig_path.type: log.Log("Generating delta - deleted file: %s" % (sig_path.get_relative_path(),), 5) yield ROPath(sig_path.index) elif sig_path and new_path == sig_path: pass # no change, skip else: delta_path = robust.check_common_error(delta_iter_error_handler, get_delta_path, (new_path, sig_path)) if delta_path: # if not, an error must have occurred log_delta_path(delta_path) yield delta_path def sigtar2path_iter(sigtarobj): """Convert signature tar file object open for reading into path iter""" tf = tarfile.TarFile("Arbitrary Name", "r", sigtarobj) tf.debug = 2 for tarinfo in tf: for prefix in ["signature/", "snapshot/", "deleted/"]: if tarinfo.name.startswith(prefix): # strip prefix and from name and set it to difftype name, difftype = tarinfo.name[len(prefix):], prefix[:-1] break else: raise DiffDirException("Bad tarinfo name %s" % (tarinfo.name,)) index = tuple(name.split("/")) if not index[-1]: index = index[:-1] # deal with trailing /, "" ropath = ROPath(index) ropath.difftype = difftype if difftype == "signature" or difftype == "snapshot": ropath.init_from_tarinfo(tarinfo) if ropath.isreg(): ropath.setfileobj(tf.extractfile(tarinfo)) yield ropath sigtarobj.close() def collate2iters(riter1, riter2): """Collate two iterators. The elements yielded by each iterator must be have an index variable, and this function returns pairs (elem1, elem2), (elem1, None), or (None, elem2) two elements in a pair will have the same index, and earlier indicies are yielded later than later indicies. """ relem1, relem2 = None, None while 1: if not relem1: try: relem1 = riter1.next() except StopIteration: if relem2: yield (None, relem2) for relem2 in riter2: yield (None, relem2) break index1 = relem1.index if not relem2: try: relem2 = riter2.next() except StopIteration: if relem1: yield (relem1, None) for relem1 in riter1: yield (relem1, None) break index2 = relem2.index if index1 < index2: yield (relem1, None) relem1 = None elif index1 == index2: yield (relem1, relem2) relem1, relem2 = None, None else: # index2 is less yield (None, relem2) relem2 = None def combine_path_iters(path_iter_list): """Produce new iterator by combining the iterators in path_iter_list This new iter will iterate every path that is in path_iter_list in order of increasing index. If multiple iterators in path_iter_list yield paths with the same index, combine_path_iters will discard all paths but the one yielded by the last path_iter. This is used to combine signature iters, as the output will be a full up-to-date signature iter. """ path_iter_list = path_iter_list[:] # copy before destructive reverse path_iter_list.reverse() def get_triple(iter_index): """Represent the next element as a triple, to help sorting""" try: path = path_iter_list[iter_index].next() except StopIteration: return None return (path.index, iter_index, path) def refresh_triple_list(triple_list): """Update all elements with path_index same as first element""" path_index = triple_list[0][0] iter_index = 0 while iter_index < len(triple_list): old_triple = triple_list[iter_index] if old_triple[0] == path_index: new_triple = get_triple(old_triple[1]) if new_triple: triple_list[iter_index] = new_triple iter_index += 1 else: del triple_list[iter_index] else: break # assumed triple_list sorted, so can exit now triple_list = filter(lambda x: x, map(get_triple, range(len(path_iter_list)))) while triple_list: triple_list.sort() yield triple_list[0][2] refresh_triple_list(triple_list) def DirDelta_WriteSig(path_iter, sig_infp_list, newsig_outfp): """Like DirDelta but also write signature into sig_fileobj Like DirDelta, sig_infp_list can be a tar fileobj or a sorted list of those. A signature will only be written to newsig_outfp if it is different from (the combined) sig_infp_list. """ global stats stats = statistics.StatsDeltaProcess() if type(sig_infp_list) is types.ListType: sig_path_iter = get_combined_path_iter(sig_infp_list) else: sig_path_iter = sigtar2path_iter(sig_infp_list) delta_iter = get_delta_iter_w_sig(path_iter, sig_path_iter, newsig_outfp) return DeltaTarBlockIter(delta_iter) def get_combined_path_iter(sig_infp_list): """Return path iter combining signatures in list of open sig files""" return combine_path_iters(map(sigtar2path_iter, sig_infp_list)) def get_delta_iter_w_sig(path_iter, sig_path_iter, sig_fileobj): """Like get_delta_iter but also write signatures to sig_fileobj""" collated = collate2iters(path_iter, sig_path_iter) sigTarFile = tarfile.TarFile("arbitrary", "w", sig_fileobj) for new_path, sig_path in collated: log.Log("Comparing %s and %s" % (new_path and new_path.index, sig_path and sig_path.index), 6) if not new_path or not new_path.type: # file doesn't exist if sig_path and sig_path.exists(): # but signature says it did log.Log("Generating delta - deleted file: %s" % (sig_path.get_relative_path(),), 5) ti = ROPath(sig_path.index).get_tarinfo() ti.name = "deleted/" + "/".join(sig_path.index) sigTarFile.addfile(ti) stats.add_deleted_file() yield ROPath(sig_path.index) elif not sig_path or new_path != sig_path: # Must calculate new signature and create delta delta_path = robust.check_common_error( delta_iter_error_handler, get_delta_path_w_sig, (new_path, sig_path, sigTarFile)) if delta_path: log_delta_path(delta_path, new_path, stats) yield delta_path else: stats.Errors += 1 else: stats.add_unchanged_file(new_path) stats.close() sigTarFile.close() def get_delta_path_w_sig(new_path, sig_path, sigTarFile): """Return new delta_path which, when read, writes sig to sig_fileobj""" assert new_path ti = new_path.get_tarinfo() index = new_path.index delta_path = new_path.get_ropath() log.Log("Getting delta of %s and %s" % (new_path, sig_path), 7) def callback(sig_string): """Callback activated when FileWithSignature read to end""" ti.size = len(sig_string) ti.name = "signature/" + "/".join(index) sigTarFile.addfile(ti, cStringIO.StringIO(sig_string)) if new_path.isreg() and sig_path and sig_path.difftype == "signature": delta_path.difftype = "diff" old_sigfp = sig_path.open("rb") newfp = FileWithSignature(new_path.open("rb"), callback, new_path.getsize()) delta_path.setfileobj(librsync.DeltaFile(old_sigfp, newfp)) else: delta_path.difftype = "snapshot" ti.name = "snapshot/" + "/".join(index) if not new_path.isreg(): sigTarFile.addfile(ti) else: delta_path.setfileobj(FileWithSignature(new_path.open("rb"), callback, new_path.getsize())) new_path.copy_attribs(delta_path) delta_path.stat.st_size = new_path.stat.st_size return delta_path class FileWithSignature: """File-like object which also computes signature as it is read""" blocksize = 32 * 1024 def __init__(self, infile, callback, filelen, *extra_args): """FileTee initializer The object will act like infile, but whenever it is read it add infile's data to a SigGenerator object. When the file has been read to the end the callback will be called with the calculated signature, and any extra_args if given. filelen is used to calculate the block size of the signature. """ self.infile, self.callback = infile, callback self.sig_gen = librsync.SigGenerator(get_block_size(filelen)) self.activated_callback = None self.extra_args = extra_args def read(self, length = -1): buf = self.infile.read(length) self.sig_gen.update(buf) return buf def close(self): # Make sure all of infile read if not self.activated_callback: while self.read(self.blocksize): pass self.activated_callback = 1 self.callback(self.sig_gen.getsig(), *self.extra_args) return self.infile.close() class TarBlock: """Contain information to add next file to tar""" def __init__(self, index, data): """TarBlock initializer - just store data""" self.index = index self.data = data class TarBlockIter: """A bit like an iterator, yield tar blocks given input iterator Unlike an iterator, however, control over the maximum size of a tarblock is available by passing an argument to next(). Also the get_footer() is available. """ def __init__(self, input_iter): """TarBlockIter initializer""" self.input_iter = input_iter self.offset = 0l # total length of data read self.process_waiting = None # process_continued has more blocks self.previous_index = None # holds index of last block returned self.remember_next = None # see remember_next_index() self.remember_value = None # We need to instantiate a dummy TarFile just to get access to # some of the functions like _get_full_headers. self.tf = tarfile.TarFromIterator(None) def tarinfo2tarblock(self, index, tarinfo, file_data = ""): """Make tarblock out of tarinfo and file data""" tarinfo.size = len(file_data) headers = self.tf._get_full_headers(tarinfo) blocks, remainder = divmod(tarinfo.size, tarfile.BLOCKSIZE) if remainder > 0: filler_data = "\0" * (tarfile.BLOCKSIZE - remainder) else: filler_data = "" return TarBlock(index, "%s%s%s" % (headers, file_data, filler_data)) def process(self, val, size): """Turn next value of input_iter into a TarBlock""" assert not self.process_waiting XXX # this should be overridden in subclass def process_continued(self, size): """Get more tarblocks If processing val above would produce more than one TarBlock, get the rest of them by calling process_continue. """ assert self.process_waiting XXX # Override this too in subclass def next(self, size = 1024 * 1024): """Return next block, no bigger than size, and update offset""" if self.process_waiting: result = self.process_continued(size) else: # Below a StopIteration exception will just be passed upwards result = self.process(self.input_iter.next(), size) self.offset += len(result.data) self.previous_index = result.index if self.remember_next: self.remember_value = result.index self.remember_next = None return result def get_previous_index(self): """Return index of last tarblock, or None if no previous index""" return self.previous_index def remember_next_index(self): """When called, remember the index of the next block iterated""" self.remember_next = 1 self.remember_value = None def recall_index(self): """Retrieve index remembered with remember_next_index""" return self.remember_value # or None if no index recalled def get_footer(self): """Return closing string for tarfile, reset offset""" blocks, remainder = divmod(self.offset, tarfile.RECORDSIZE) self.offset = 0l return '\0' * (tarfile.RECORDSIZE - remainder) # remainder can be 0 def __iter__(self): return self class SigTarBlockIter(TarBlockIter): """TarBlockIter that yields blocks of a signature tar from path_iter""" def process(self, path, size): """Return associated signature TarBlock from path Here size is just ignored --- let's hope a signature isn't too big. Also signatures are stored in multiple volumes so it doesn't matter. """ ti = path.get_tarinfo() if path.isreg(): sfp = librsync.SigFile(path.open("rb"), get_block_size(path.getsize())) sigbuf = sfp.read() sfp.close() ti.name = "signature/" + "/".join(path.index) return self.tarinfo2tarblock(path.index, ti, sigbuf) else: ti.name = "snapshot/" + "/".join(path.index) return self.tarinfo2tarblock(path.index, ti) class DeltaTarBlockIter(TarBlockIter): """TarBlockIter that yields parts of a deltatar file Unlike SigTarBlockIter, the argument to __init__ is a delta_path_iter, so the delta information has already been calculated. """ def process(self, delta_ropath, size): """Get a tarblock from delta_ropath""" def add_prefix(tarinfo, prefix): """Add prefix to the name of a tarinfo file""" if tarinfo.name == ".": tarinfo.name = prefix + "/" else: tarinfo.name = "%s/%s" % (prefix, tarinfo.name) ti = delta_ropath.get_tarinfo() index = delta_ropath.index # Return blocks of deleted files or fileless snapshots if not delta_ropath.type or not delta_ropath.fileobj: if not delta_ropath.type: add_prefix(ti, "deleted") else: assert delta_ropath.difftype == "snapshot" add_prefix(ti, "snapshot") return self.tarinfo2tarblock(index, ti) # Now handle single volume block case fp = delta_ropath.open("rb") # Below the 512 is the usual length of a tar header data, last_block = self.get_data_block(fp, size - 512) if stats: stats.RawDeltaSize += len(data) if last_block: if delta_ropath.difftype == "snapshot": add_prefix(ti, "snapshot") elif delta_ropath.difftype == "diff": add_prefix(ti, "diff") else: assert 0, "Unknown difftype" return self.tarinfo2tarblock(index, ti, data) # Finally, do multivol snapshot or diff case full_name = "multivol_%s/%s" % (delta_ropath.difftype, ti.name) ti.name = full_name + "/1" self.process_prefix = full_name self.process_fp = fp self.process_ropath = delta_ropath self.process_waiting = 1 self.process_next_vol_number = 2 return self.tarinfo2tarblock(index, ti, data) def get_data_block(self, fp, max_size): """Return pair (next data block, boolean last data block)""" read_size = min(64*1024, max_size) buf = fp.read(read_size) if len(buf) < read_size: if fp.close(): raise DiffDirException("Error closing file") return (buf, 1) else: return (buf, None) def process_continued(self, size): """Return next volume in multivol diff or snapshot""" assert self.process_waiting ropath = self.process_ropath ti, index = ropath.get_tarinfo(), ropath.index ti.name = "%s/%d" % (self.process_prefix, self.process_next_vol_number) data, last_block = self.get_data_block(self.process_fp, size-512) if last_block: self.process_prefix = None self.process_fp = None self.process_ropath = None self.process_waiting = None self.process_next_vol_number = None else: self.process_next_vol_number += 1 return self.tarinfo2tarblock(index, ti, data) def write_block_iter(block_iter, out_obj): """Write block_iter to filename, path, or file object""" if isinstance(out_obj, Path): fp = open(out_obj.name, "wb") elif type(out_obj) is types.StringType: fp = open(out_obj, "wb") else: fp = out_obj for block in block_iter: fp.write(block.data) fp.write(block_iter.get_footer()) assert not fp.close() if isinstance(out_obj, Path): out_obj.setdata() def get_block_size(file_len): """Return a reasonable block size to use on files of length file_len If the block size is too big, deltas will be bigger than is necessary. If the block size is too small, making deltas and patching can take a really long time. """ if file_len < 1024000: return 512 # set minimum of 512 bytes else: # Split file into about 2000 pieces, rounding to 512 return min(long((file_len/(2000*512))*512), 2048)