Various CPU optimizations

git-svn-id: http://svn.savannah.nongnu.org/svn/rdiff-backup@126 2b77aa54-bcbc-44c9-a7ec-4f6cf2b41109

trunk/rdiff-backup/rdiff_backup/connection.py

@@ -40,11 +40,7 @@ class LocalConnection(Connection):
 	def __getattr__(self, name):
 		try: return globals()[name]
 		except KeyError:
-			builtins = globals()["__builtins__"]
-			try:
-				if type(builtins) is types.ModuleType:
-					return builtins.__dict__[name]
-				else: return builtins[name]
+			try: return __builtins__.__dict__[name]
 			except KeyError: raise NameError, name
 
 	def __setattr__(self, name, value):

trunk/rdiff-backup/rdiff_backup/highlevel.py

@@ -105,7 +105,7 @@ class HLSourceStruct:
 					if dest_sig.isplaceholder(): yield dest_sig
 					else:
 						diff = Robust.check_common_error(
-							error_handler, RORPIter.diffonce, dest_sig, dsrp)
+							error_handler, RORPIter.diffonce, [dest_sig, dsrp])
 						if diff: yield diff
 				if dsrp: finalizer(dsrp.index, dsrp)
 			finalizer.Finish()
@@ -237,10 +237,10 @@ class HLDestinationStruct:
 			Log("Error: %s processing file %s" % (exc, filename), 2)
 		
 		for indexed_tuple in collated:
-			Log("Processing %s" % str(indexed_tuple), 7)
+			Log(lambda: "Processing %s" % str(indexed_tuple), 7)
 			diff_rorp, dsrp = indexed_tuple
 			dsrp = Robust.check_common_error(error_handler, patch,
-											 diff_rorp, dsrp)
+											 [diff_rorp, dsrp])
 			finalizer(dsrp.index, dsrp)
 		finalizer.Finish()
 
@@ -253,7 +253,7 @@ class HLDestinationStruct:
 
 		try:
 			for indexed_tuple in collated:
-				Log("Processing %s" % str(indexed_tuple), 7)
+				Log(lambda: "Processing %s" % str(indexed_tuple), 7)
 				diff_rorp, dsrp = indexed_tuple
 				if not dsrp: dsrp = cls.get_dsrp(dest_rpath, diff_rorp.index)
 				if diff_rorp and diff_rorp.isplaceholder(): diff_rorp = None
@@ -279,7 +279,7 @@ class HLDestinationStruct:
 
 		try:
 			for indexed_tuple in collated:
-				Log("Processing %s" % str(indexed_tuple), 7)
+				Log(lambda: "Processing %s" % str(indexed_tuple), 7)
 				diff_rorp, dsrp = indexed_tuple
 				index = indexed_tuple.index
 				if not dsrp: dsrp = cls.get_dsrp(dest_rpath, index)

trunk/rdiff-backup/rdiff_backup/lazy.py

@@ -196,9 +196,9 @@ class IterTreeReducer:
 	iterator nature of the connection between hosts and the temporal
 	order in which the files are processed.
 
-	There are four stub functions below: start_process, end_process,
-	branch_process, and check_for_errors.  A class that subclasses
-	this one will probably fill in these functions to do more.
+	There are three stub functions below: start_process, end_process,
+	and branch_process.  A class that subclasses this one will
+	probably fill in these functions to do more.
 
 	It is important that this class be pickable, so keep that in mind
 	when subclassing (this is used to resume failed sessions).
@@ -207,26 +207,49 @@ class IterTreeReducer:
 	def __init__(self, *args):
 		"""ITR initializer"""
 		self.init_args = args
-		self.index = None
-		self.subinstance = None
+		self.base_index = self.index = None
+		self.subinstances = [self]
 		self.finished = None
-		self.caught_exception, self.start_successful = None, None
+		self.caught_exception = None
 
-	def intree(self, index):
-		"""Return true if index is still in current tree"""
-		return self.base_index == index[:len(self.base_index)]
+	def finish_subinstances(self, index):
+		"""Run Finish() on all subinstances index has passed
 
-	def set_subinstance(self):
-		"""Return subinstance of same type as self"""
-		self.subinstance = self.__class__(*self.init_args)
+		When we pass out of a subinstance's tree, delete it and
+		process it with the parent.  The innermost subinstances will
+		be the last in the list.  Return None if we are out of the
+		entire tree, and 1 otherwise.
 
-	def process_w_subinstance(self, args):
-		"""Give object to subinstance, if necessary update branch_val"""
-		if not self.subinstance: self.set_subinstance()
-		if not self.subinstance(*args):
-			self.branch_process(self.subinstance)
-			self.set_subinstance()
-			assert self.subinstance(*args)
+		"""
+		subinstances = self.subinstances
+		while 1:
+			to_be_finished = subinstances[-1]
+			base_index = to_be_finished.base_index
+			if base_index != index[:len(base_index)]:
+				# out of the tree, finish with to_be_finished
+				to_be_finished.call_end_proc()
+				del subinstances[-1]
+				if not subinstances: return None
+				subinstances[-1].branch_process(to_be_finished)
+			else: return 1
+
+	def call_end_proc(self):
+		"""Runs the end_process on self, checking for errors"""
+		if self.finished: self.caught_exception = 1
+		if self.caught_exception: self.log_pref_error(self.base_index)
+		Robust.check_common_error(self.on_error, self.end_process)
+		self.finished = 1
+
+	def add_subinstance(self):
+		"""Return subinstance of same type as self, add to subinstances"""
+		subinst = self.__class__(*self.init_args)
+		self.subinstances.append(subinst)
+		return subinst
+
+	def process_w_subinstance(self, index, subinst, args):
+		"""Run start_process on latest subinstance"""
+		Robust.check_common_error(self.on_error, subinst.start_process, args)
+		subinst.base_index = index
 
 	def start_process(self, *args):
 		"""Do some initial processing (stub)"""
@@ -238,31 +261,20 @@ class IterTreeReducer:
 
 	def branch_process(self, subinstance):
 		"""Process a branch right after it is finished (stub)"""
+		assert subinstance.finished
 		pass
 
-	def check_for_errors(self, function, *args):
-		"""start/end_process is called by this function
-
-		Usually it will distinguish between two types of errors.  Some
-		are serious and will be reraised, others are caught and simply
-		invalidate the current instance by setting
-		self.caught_exception.
-
-		"""
-		try: return apply(function, args)
-		except: raise
+	def on_error(self, exc, *args):
+		"""This will be run on any exception in start/end-process"""
+		pass
 
 	def Finish(self):
 		"""Call at end of sequence to tie everything up"""
-		if not self.start_successful or self.finished:
-			self.caught_exception = 1
-		if self.caught_exception: self.log_prev_error(self.index)
-		else:
-			if self.subinstance:
-				self.subinstance.Finish()
-				self.branch_process(self.subinstance)
-			self.check_for_errors(self.end_process)
-			self.finished = 1
+		while 1:
+			to_be_finished = self.subinstances.pop()
+			to_be_finished.call_end_proc()
+			if not self.subinstances: break
+			self.subinstances[-1].branch_process(to_be_finished)
 
 	def log_prev_error(self, index):
 		"""Call function if no pending exception"""
@@ -280,24 +292,22 @@ class IterTreeReducer:
 
 		"""
 		index = args[0]
-		assert type(index) is types.TupleType, type(index)
-
-		if self.index is None:
-			self.check_for_errors(self.start_process, *args)
-			self.start_successful = 1
-			self.index = self.base_index = index
+		if self.base_index is None:
+			self.process_w_subinstance(index, self, args)
+			self.index = index
 			return 1
 
 		if index <= self.index:
 			Log("Warning: oldindex %s >= newindex %s" % (self.index, index), 2)
 			return 1
 
-		if not self.intree(index):
-			self.Finish()
-			return None
-
+		if self.finish_subinstances(index) is None:
+			return None # We are no longer in the main tree
 		if self.caught_exception: self.log_prev_error(index)
-		else: self.process_w_subinstance(args)
+		else:
+			subinst = self.add_subinstance()
+			self.process_w_subinstance(index, subinst, args)
+
 		self.index = index
 		return 1
 
@@ -313,7 +323,4 @@ class ErrorITR(IterTreeReducer):
 		else: filename = "."
 		Log("Error '%s' processing %s" % (exc, filename), 2)
 
-	def check_for_errors(self, function, *args):
-		"""Catch some non-fatal errors"""
-		return Robust.check_common_error(self.on_error, function, *args)
 

trunk/rdiff-backup/rdiff_backup/log.py

@@ -81,7 +81,21 @@ class Logger:
 								   message)
 
 	def __call__(self, message, verbosity):
-		"""Log message that has verbosity importance"""
+		"""Log message that has verbosity importance
+
+		message can be a string, which is logged as-is, or a function,
+		which is then called and should return the string to be
+		logged.  We do it this way in case producing the string would
+		take a significant amount of CPU.
+		
+		"""
+		if verbosity > self.verbosity and verbosity > self.term_verbosity:
+			return
+
+		if not type(message) is types.StringType:
+			assert type(message) is types.FunctionType
+			message = message()
+
 		if verbosity <= self.verbosity: self.log_to_file(message)
 		if verbosity <= self.term_verbosity:
 			self.log_to_term(message, verbosity)

trunk/rdiff-backup/rdiff_backup/profiled_rdb.py

@@ -13,8 +13,7 @@ import profile, pstats
 profile.run("Globals.Main.Main(%s)" % repr(sys.argv[1:]), "profile-output")
 p = pstats.Stats("profile-output")
 p.sort_stats('time')
-p.print_stats(20)
-p.sort_stats('cumulative')
-p.print_stats(20)
+p.print_stats(40)
+#p.print_callers(20)
 
 

trunk/rdiff-backup/rdiff_backup/robust.py

@@ -234,7 +234,7 @@ class Robust:
 			tf.setdata()
 		return Robust.make_tf_robustaction(init, tf, rp)
 
-	def check_common_error(error_handler, function, *args):
+	def check_common_error(error_handler, function, args = []):
 		"""Apply function to args, if error, run error_handler on exception
 
 		This only catches certain exceptions which seems innocent

trunk/rdiff-backup/rdiff_backup/rorpiter.py

@@ -69,7 +69,7 @@ class RORPIter:
 					if rp.isflaglinked(): rorp.flaglinked()
 					else:
 						fp = Robust.check_common_error(
-							error_handler, Rdiff.get_signature, rp)
+							error_handler, Rdiff.get_signature, (rp,))
 						if fp: rorp.setfile(fp)
 						else: continue
 				yield rorp

trunk/rdiff-backup/rdiff_backup/rpath.py

@@ -163,7 +163,7 @@ class RPathStatic:
 	def rename(rp_source, rp_dest):
 		"""Rename rp_source to rp_dest"""
 		assert rp_source.conn is rp_dest.conn
-		Log("Renaming %s to %s" % (rp_source.path, rp_dest.path), 7)
+		Log(lambda: "Renaming %s to %s" % (rp_source.path, rp_dest.path), 7)
 		rp_source.conn.os.rename(rp_source.path, rp_dest.path)
 		rp_dest.data = rp_source.data
 		rp_source.data = {'type': None}
@@ -422,7 +422,7 @@ class RPath(RORPath):
 		self.conn = connection
 		self.index = index
 		self.base = base
-		self.path = apply(os.path.join, (base,) + self.index)
+		if base is not None: self.path = "/".join((base,) + index)
 		self.file = None
 		if data or base is None: self.data = data
 		else: self.setdata()
@@ -445,7 +445,7 @@ class RPath(RORPath):
 		"""Reproduce RPath from __getstate__ output"""
 		self.conn = Globals.local_connection
 		self.index, self.base, self.data = rpath_state
-		self.path = apply(os.path.join, (self.base,) + self.index)
+		self.path = "/".join((self.base,) + self.index)
 
 	def setdata(self):
 		"""Create the data dictionary"""
@@ -456,8 +456,7 @@ class RPath(RORPath):
 		data = {}
 		mode = statblock[stat.ST_MODE]
 
-		if stat.S_ISREG(mode):
-			type = 'reg'
+		if stat.S_ISREG(mode): type = 'reg'
 		elif stat.S_ISDIR(mode): type = 'dir'
 		elif stat.S_ISCHR(mode):
 			type = 'dev'
@@ -509,7 +508,7 @@ class RPath(RORPath):
 	def quote_path(self):
 		"""Set path from quoted version of index"""
 		quoted_list = [FilenameMapping.quote(path) for path in self.index]
-		self.path = apply(os.path.join, [self.base] + quoted_list)
+		self.path = "/".join([self.base] + quoted_list)
 		self.setdata()
 
 	def chmod(self, permissions):
@@ -526,7 +525,7 @@ class RPath(RORPath):
 
 	def setmtime(self, modtime):
 		"""Set only modtime (access time to present)"""
-		Log("Setting time of %s to %d" % (self.path, modtime), 7)
+		Log(lambda: "Setting time of %s to %d" % (self.path, modtime), 7)
 		self.conn.os.utime(self.path, (time.time(), modtime))
 		self.data['mtime'] = modtime
 
@@ -594,7 +593,7 @@ class RPath(RORPath):
 		
 	def isowner(self):
 		"""Return true if current process is owner of rp or root"""
-		uid = self.conn.Globals.get('process_uid')
+		uid = self.conn.os.getuid()
 		return uid == 0 or uid == self.data['uid']
 
 	def isgroup(self):
@@ -666,8 +665,7 @@ class RPath(RORPath):
 	def append_path(self, ext, new_index = ()):
 		"""Like append, but add ext to path instead of to index"""
 		assert not self.index # doesn't make sense if index isn't ()
-		return self.__class__(self.conn, os.path.join(self.base, ext),
-							  new_index)
+		return self.__class__(self.conn, "/".join((self.base, ext)), new_index)
 
 	def new_index(self, index):
 		"""Return similar RPath but with new index"""

trunk/rdiff-backup/rdiff_backup/selection.py

@@ -117,13 +117,15 @@ class Select:
 			if dsrpath.isdir():
 				for dsrp in self.iterate_in_dir(dsrpath, rec_func, sel_func):
 					yield dsrp
-		elif s == 2 and dsrpath.isdir(): # Directory is merely scanned
-			iid = self.iterate_in_dir(dsrpath, rec_func, sel_func)
-			try: first = iid.next()
-			except StopIteration: return # no files inside; skip dsrp
-			yield dsrpath
-			yield first
-			for dsrp in iid: yield dsrp
+		elif s == 2:
+			if dsrpath.isdir(): # Directory is merely scanned
+				iid = self.iterate_in_dir(dsrpath, rec_func, sel_func)
+				try: first = iid.next()
+				except StopIteration: return # no files inside; skip dsrp
+				yield dsrpath
+				yield first
+				for dsrp in iid: yield dsrp
+		else: assert 0, "Invalid selection result %s" % (str(s),)
 
 	def iterate_in_dir(self, dsrpath, rec_func, sel_func):
 		"""Iterate the dsrps in directory dsrpath."""
@@ -137,7 +139,7 @@ class Select:
 		else:
 			for filename in Robust.listrp(dsrpath):
 				new_dsrp = Robust.check_common_error(
-					error_handler, dsrpath.append, filename)
+					error_handler, dsrpath.append, [filename])
 				if new_dsrp:
 					for dsrp in rec_func(new_dsrp, rec_func, sel_func):
 						yield dsrp

trunk/rdiff-backup/rdiff_backup/statistics.py

@@ -32,21 +32,19 @@ class StatsObj:
 					   ('ChangedSourceSize', 1), ('ChangedMirrorSize', 1),
 					   ('IncrementFiles', None), ('IncrementFileSize', 1))
 
-	# Set all stats to None, indicating info not available
-	for attr in stat_attrs: locals()[attr] = None
-
 	# This is used in get_byte_summary_string below
 	byte_abbrev_list = ((1024*1024*1024*1024, "TB"),
 						(1024*1024*1024, "GB"),
 						(1024*1024, "MB"),
 						(1024, "KB"))
 
+	def __init__(self):
+		"""Set attributes to None"""
+		for attr in self.stat_attrs: self.__dict__[attr] = None
+
 	def get_stat(self, attribute):
 		"""Get a statistic"""
-		try: return self.__dict__[attribute]
-		except KeyError:
-			# this may be a hack, but seems no good way to get attrs in python
-			return eval("self.%s" % attribute)
+		return self.__dict__[attribute]
 
 	def set_stat(self, attr, value):
 		"""Set attribute to given value"""
@@ -54,7 +52,7 @@ class StatsObj:
 
 	def increment_stat(self, attr):
 		"""Add 1 to value of attribute"""
-		self.__dict__[attr] = self.get_stat(attr) + 1
+		self.__dict__[attr] += 1
 
 	def get_stats_line(self, index, use_repr = 1):
 		"""Return one line abbreviated version of full stats string"""
@@ -217,8 +215,12 @@ class StatsITR(IterTreeReducer, StatsObj):
 	This is subclassed by the mirroring and incrementing ITRs.
 
 	"""
-	# zero out file statistics
-	for attr in StatsObj.stat_file_attrs: locals()[attr] = 0
+	def __init__(self, *args):
+		"""StatsITR initializer - zero out statistics"""
+		attr_dict = self.__dict__
+		for attr in StatsObj.stat_file_attrs: attr_dict[attr] = 0
+		self.ElapsedTime = self.Filename = None
+		IterTreeReducer.__init__(self, *args)
 
 	def start_stats(self, mirror_dsrp):
 		"""Record status of mirror dsrp
@@ -272,8 +274,7 @@ class StatsITR(IterTreeReducer, StatsObj):
 	def add_file_stats(self, subinstance):
 		"""Add all file statistics from subinstance to current totals"""
 		for attr in self.stat_file_attrs:
-			self.set_stat(attr,
-						  self.get_stat(attr) + subinstance.get_stat(attr))
+			self.__dict__[attr] += subinstance.__dict__[attr]
 
 
 class Stats:

trunk/rdiff-backup/src/connection.py

@@ -40,11 +40,7 @@ class LocalConnection(Connection):
 	def __getattr__(self, name):
 		try: return globals()[name]
 		except KeyError:
-			builtins = globals()["__builtins__"]
-			try:
-				if type(builtins) is types.ModuleType:
-					return builtins.__dict__[name]
-				else: return builtins[name]
+			try: return __builtins__.__dict__[name]
 			except KeyError: raise NameError, name
 
 	def __setattr__(self, name, value):

trunk/rdiff-backup/src/highlevel.py

@@ -105,7 +105,7 @@ class HLSourceStruct:
 					if dest_sig.isplaceholder(): yield dest_sig
 					else:
 						diff = Robust.check_common_error(
-							error_handler, RORPIter.diffonce, dest_sig, dsrp)
+							error_handler, RORPIter.diffonce, [dest_sig, dsrp])
 						if diff: yield diff
 				if dsrp: finalizer(dsrp.index, dsrp)
 			finalizer.Finish()
@@ -237,10 +237,10 @@ class HLDestinationStruct:
 			Log("Error: %s processing file %s" % (exc, filename), 2)
 		
 		for indexed_tuple in collated:
-			Log("Processing %s" % str(indexed_tuple), 7)
+			Log(lambda: "Processing %s" % str(indexed_tuple), 7)
 			diff_rorp, dsrp = indexed_tuple
 			dsrp = Robust.check_common_error(error_handler, patch,
-											 diff_rorp, dsrp)
+											 [diff_rorp, dsrp])
 			finalizer(dsrp.index, dsrp)
 		finalizer.Finish()
 
@@ -253,7 +253,7 @@ class HLDestinationStruct:
 
 		try:
 			for indexed_tuple in collated:
-				Log("Processing %s" % str(indexed_tuple), 7)
+				Log(lambda: "Processing %s" % str(indexed_tuple), 7)
 				diff_rorp, dsrp = indexed_tuple
 				if not dsrp: dsrp = cls.get_dsrp(dest_rpath, diff_rorp.index)
 				if diff_rorp and diff_rorp.isplaceholder(): diff_rorp = None
@@ -279,7 +279,7 @@ class HLDestinationStruct:
 
 		try:
 			for indexed_tuple in collated:
-				Log("Processing %s" % str(indexed_tuple), 7)
+				Log(lambda: "Processing %s" % str(indexed_tuple), 7)
 				diff_rorp, dsrp = indexed_tuple
 				index = indexed_tuple.index
 				if not dsrp: dsrp = cls.get_dsrp(dest_rpath, index)

trunk/rdiff-backup/src/lazy.py

@@ -196,9 +196,9 @@ class IterTreeReducer:
 	iterator nature of the connection between hosts and the temporal
 	order in which the files are processed.
 
-	There are four stub functions below: start_process, end_process,
-	branch_process, and check_for_errors.  A class that subclasses
-	this one will probably fill in these functions to do more.
+	There are three stub functions below: start_process, end_process,
+	and branch_process.  A class that subclasses this one will
+	probably fill in these functions to do more.
 
 	It is important that this class be pickable, so keep that in mind
 	when subclassing (this is used to resume failed sessions).
@@ -207,26 +207,49 @@ class IterTreeReducer:
 	def __init__(self, *args):
 		"""ITR initializer"""
 		self.init_args = args
-		self.index = None
-		self.subinstance = None
+		self.base_index = self.index = None
+		self.subinstances = [self]
 		self.finished = None
-		self.caught_exception, self.start_successful = None, None
+		self.caught_exception = None
 
-	def intree(self, index):
-		"""Return true if index is still in current tree"""
-		return self.base_index == index[:len(self.base_index)]
+	def finish_subinstances(self, index):
+		"""Run Finish() on all subinstances index has passed
 
-	def set_subinstance(self):
-		"""Return subinstance of same type as self"""
-		self.subinstance = self.__class__(*self.init_args)
+		When we pass out of a subinstance's tree, delete it and
+		process it with the parent.  The innermost subinstances will
+		be the last in the list.  Return None if we are out of the
+		entire tree, and 1 otherwise.
 
-	def process_w_subinstance(self, args):
-		"""Give object to subinstance, if necessary update branch_val"""
-		if not self.subinstance: self.set_subinstance()
-		if not self.subinstance(*args):
-			self.branch_process(self.subinstance)
-			self.set_subinstance()
-			assert self.subinstance(*args)
+		"""
+		subinstances = self.subinstances
+		while 1:
+			to_be_finished = subinstances[-1]
+			base_index = to_be_finished.base_index
+			if base_index != index[:len(base_index)]:
+				# out of the tree, finish with to_be_finished
+				to_be_finished.call_end_proc()
+				del subinstances[-1]
+				if not subinstances: return None
+				subinstances[-1].branch_process(to_be_finished)
+			else: return 1
+
+	def call_end_proc(self):
+		"""Runs the end_process on self, checking for errors"""
+		if self.finished: self.caught_exception = 1
+		if self.caught_exception: self.log_pref_error(self.base_index)
+		Robust.check_common_error(self.on_error, self.end_process)
+		self.finished = 1
+
+	def add_subinstance(self):
+		"""Return subinstance of same type as self, add to subinstances"""
+		subinst = self.__class__(*self.init_args)
+		self.subinstances.append(subinst)
+		return subinst
+
+	def process_w_subinstance(self, index, subinst, args):
+		"""Run start_process on latest subinstance"""
+		Robust.check_common_error(self.on_error, subinst.start_process, args)
+		subinst.base_index = index
 
 	def start_process(self, *args):
 		"""Do some initial processing (stub)"""
@@ -238,31 +261,20 @@ class IterTreeReducer:
 
 	def branch_process(self, subinstance):
 		"""Process a branch right after it is finished (stub)"""
+		assert subinstance.finished
 		pass
 
-	def check_for_errors(self, function, *args):
-		"""start/end_process is called by this function
-
-		Usually it will distinguish between two types of errors.  Some
-		are serious and will be reraised, others are caught and simply
-		invalidate the current instance by setting
-		self.caught_exception.
-
-		"""
-		try: return apply(function, args)
-		except: raise
+	def on_error(self, exc, *args):
+		"""This will be run on any exception in start/end-process"""
+		pass
 
 	def Finish(self):
 		"""Call at end of sequence to tie everything up"""
-		if not self.start_successful or self.finished:
-			self.caught_exception = 1
-		if self.caught_exception: self.log_prev_error(self.index)
-		else:
-			if self.subinstance:
-				self.subinstance.Finish()
-				self.branch_process(self.subinstance)
-			self.check_for_errors(self.end_process)
-			self.finished = 1
+		while 1:
+			to_be_finished = self.subinstances.pop()
+			to_be_finished.call_end_proc()
+			if not self.subinstances: break
+			self.subinstances[-1].branch_process(to_be_finished)
 
 	def log_prev_error(self, index):
 		"""Call function if no pending exception"""
@@ -280,24 +292,22 @@ class IterTreeReducer:
 
 		"""
 		index = args[0]
-		assert type(index) is types.TupleType, type(index)
-
-		if self.index is None:
-			self.check_for_errors(self.start_process, *args)
-			self.start_successful = 1
-			self.index = self.base_index = index
+		if self.base_index is None:
+			self.process_w_subinstance(index, self, args)
+			self.index = index
 			return 1
 
 		if index <= self.index:
 			Log("Warning: oldindex %s >= newindex %s" % (self.index, index), 2)
 			return 1
 
-		if not self.intree(index):
-			self.Finish()
-			return None
-
+		if self.finish_subinstances(index) is None:
+			return None # We are no longer in the main tree
 		if self.caught_exception: self.log_prev_error(index)
-		else: self.process_w_subinstance(args)
+		else:
+			subinst = self.add_subinstance()
+			self.process_w_subinstance(index, subinst, args)
+
 		self.index = index
 		return 1
 
@@ -313,7 +323,4 @@ class ErrorITR(IterTreeReducer):
 		else: filename = "."
 		Log("Error '%s' processing %s" % (exc, filename), 2)
 
-	def check_for_errors(self, function, *args):
-		"""Catch some non-fatal errors"""
-		return Robust.check_common_error(self.on_error, function, *args)
 

trunk/rdiff-backup/src/log.py

@@ -81,7 +81,21 @@ class Logger:
 								   message)
 
 	def __call__(self, message, verbosity):
-		"""Log message that has verbosity importance"""
+		"""Log message that has verbosity importance
+
+		message can be a string, which is logged as-is, or a function,
+		which is then called and should return the string to be
+		logged.  We do it this way in case producing the string would
+		take a significant amount of CPU.
+		
+		"""
+		if verbosity > self.verbosity and verbosity > self.term_verbosity:
+			return
+
+		if not type(message) is types.StringType:
+			assert type(message) is types.FunctionType
+			message = message()
+
 		if verbosity <= self.verbosity: self.log_to_file(message)
 		if verbosity <= self.term_verbosity:
 			self.log_to_term(message, verbosity)

trunk/rdiff-backup/src/profiled_rdb.py

@@ -13,8 +13,7 @@ import profile, pstats
 profile.run("Globals.Main.Main(%s)" % repr(sys.argv[1:]), "profile-output")
 p = pstats.Stats("profile-output")
 p.sort_stats('time')
-p.print_stats(20)
-p.sort_stats('cumulative')
-p.print_stats(20)
+p.print_stats(40)
+#p.print_callers(20)
 
 

trunk/rdiff-backup/src/robust.py

@@ -234,7 +234,7 @@ class Robust:
 			tf.setdata()
 		return Robust.make_tf_robustaction(init, tf, rp)
 
-	def check_common_error(error_handler, function, *args):
+	def check_common_error(error_handler, function, args = []):
 		"""Apply function to args, if error, run error_handler on exception
 
 		This only catches certain exceptions which seems innocent

trunk/rdiff-backup/src/rorpiter.py

@@ -69,7 +69,7 @@ class RORPIter:
 					if rp.isflaglinked(): rorp.flaglinked()
 					else:
 						fp = Robust.check_common_error(
-							error_handler, Rdiff.get_signature, rp)
+							error_handler, Rdiff.get_signature, (rp,))
 						if fp: rorp.setfile(fp)
 						else: continue
 				yield rorp

trunk/rdiff-backup/src/rpath.py

@@ -163,7 +163,7 @@ class RPathStatic:
 	def rename(rp_source, rp_dest):
 		"""Rename rp_source to rp_dest"""
 		assert rp_source.conn is rp_dest.conn
-		Log("Renaming %s to %s" % (rp_source.path, rp_dest.path), 7)
+		Log(lambda: "Renaming %s to %s" % (rp_source.path, rp_dest.path), 7)
 		rp_source.conn.os.rename(rp_source.path, rp_dest.path)
 		rp_dest.data = rp_source.data
 		rp_source.data = {'type': None}
@@ -422,7 +422,7 @@ class RPath(RORPath):
 		self.conn = connection
 		self.index = index
 		self.base = base
-		self.path = apply(os.path.join, (base,) + self.index)
+		if base is not None: self.path = "/".join((base,) + index)
 		self.file = None
 		if data or base is None: self.data = data
 		else: self.setdata()
@@ -445,7 +445,7 @@ class RPath(RORPath):
 		"""Reproduce RPath from __getstate__ output"""
 		self.conn = Globals.local_connection
 		self.index, self.base, self.data = rpath_state
-		self.path = apply(os.path.join, (self.base,) + self.index)
+		self.path = "/".join((self.base,) + self.index)
 
 	def setdata(self):
 		"""Create the data dictionary"""
@@ -456,8 +456,7 @@ class RPath(RORPath):
 		data = {}
 		mode = statblock[stat.ST_MODE]
 
-		if stat.S_ISREG(mode):
-			type = 'reg'
+		if stat.S_ISREG(mode): type = 'reg'
 		elif stat.S_ISDIR(mode): type = 'dir'
 		elif stat.S_ISCHR(mode):
 			type = 'dev'
@@ -509,7 +508,7 @@ class RPath(RORPath):
 	def quote_path(self):
 		"""Set path from quoted version of index"""
 		quoted_list = [FilenameMapping.quote(path) for path in self.index]
-		self.path = apply(os.path.join, [self.base] + quoted_list)
+		self.path = "/".join([self.base] + quoted_list)
 		self.setdata()
 
 	def chmod(self, permissions):
@@ -526,7 +525,7 @@ class RPath(RORPath):
 
 	def setmtime(self, modtime):
 		"""Set only modtime (access time to present)"""
-		Log("Setting time of %s to %d" % (self.path, modtime), 7)
+		Log(lambda: "Setting time of %s to %d" % (self.path, modtime), 7)
 		self.conn.os.utime(self.path, (time.time(), modtime))
 		self.data['mtime'] = modtime
 
@@ -594,7 +593,7 @@ class RPath(RORPath):
 		
 	def isowner(self):
 		"""Return true if current process is owner of rp or root"""
-		uid = self.conn.Globals.get('process_uid')
+		uid = self.conn.os.getuid()
 		return uid == 0 or uid == self.data['uid']
 
 	def isgroup(self):
@@ -666,8 +665,7 @@ class RPath(RORPath):
 	def append_path(self, ext, new_index = ()):
 		"""Like append, but add ext to path instead of to index"""
 		assert not self.index # doesn't make sense if index isn't ()
-		return self.__class__(self.conn, os.path.join(self.base, ext),
-							  new_index)
+		return self.__class__(self.conn, "/".join((self.base, ext)), new_index)
 
 	def new_index(self, index):
 		"""Return similar RPath but with new index"""

trunk/rdiff-backup/src/selection.py

@@ -117,13 +117,15 @@ class Select:
 			if dsrpath.isdir():
 				for dsrp in self.iterate_in_dir(dsrpath, rec_func, sel_func):
 					yield dsrp
-		elif s == 2 and dsrpath.isdir(): # Directory is merely scanned
-			iid = self.iterate_in_dir(dsrpath, rec_func, sel_func)
-			try: first = iid.next()
-			except StopIteration: return # no files inside; skip dsrp
-			yield dsrpath
-			yield first
-			for dsrp in iid: yield dsrp
+		elif s == 2:
+			if dsrpath.isdir(): # Directory is merely scanned
+				iid = self.iterate_in_dir(dsrpath, rec_func, sel_func)
+				try: first = iid.next()
+				except StopIteration: return # no files inside; skip dsrp
+				yield dsrpath
+				yield first
+				for dsrp in iid: yield dsrp
+		else: assert 0, "Invalid selection result %s" % (str(s),)
 
 	def iterate_in_dir(self, dsrpath, rec_func, sel_func):
 		"""Iterate the dsrps in directory dsrpath."""
@@ -137,7 +139,7 @@ class Select:
 		else:
 			for filename in Robust.listrp(dsrpath):
 				new_dsrp = Robust.check_common_error(
-					error_handler, dsrpath.append, filename)
+					error_handler, dsrpath.append, [filename])
 				if new_dsrp:
 					for dsrp in rec_func(new_dsrp, rec_func, sel_func):
 						yield dsrp

trunk/rdiff-backup/src/statistics.py

@@ -32,21 +32,19 @@ class StatsObj:
 					   ('ChangedSourceSize', 1), ('ChangedMirrorSize', 1),
 					   ('IncrementFiles', None), ('IncrementFileSize', 1))
 
-	# Set all stats to None, indicating info not available
-	for attr in stat_attrs: locals()[attr] = None
-
 	# This is used in get_byte_summary_string below
 	byte_abbrev_list = ((1024*1024*1024*1024, "TB"),
 						(1024*1024*1024, "GB"),
 						(1024*1024, "MB"),
 						(1024, "KB"))
 
+	def __init__(self):
+		"""Set attributes to None"""
+		for attr in self.stat_attrs: self.__dict__[attr] = None
+
 	def get_stat(self, attribute):
 		"""Get a statistic"""
-		try: return self.__dict__[attribute]
-		except KeyError:
-			# this may be a hack, but seems no good way to get attrs in python
-			return eval("self.%s" % attribute)
+		return self.__dict__[attribute]
 
 	def set_stat(self, attr, value):
 		"""Set attribute to given value"""
@@ -54,7 +52,7 @@ class StatsObj:
 
 	def increment_stat(self, attr):
 		"""Add 1 to value of attribute"""
-		self.__dict__[attr] = self.get_stat(attr) + 1
+		self.__dict__[attr] += 1
 
 	def get_stats_line(self, index, use_repr = 1):
 		"""Return one line abbreviated version of full stats string"""
@@ -217,8 +215,12 @@ class StatsITR(IterTreeReducer, StatsObj):
 	This is subclassed by the mirroring and incrementing ITRs.
 
 	"""
-	# zero out file statistics
-	for attr in StatsObj.stat_file_attrs: locals()[attr] = 0
+	def __init__(self, *args):
+		"""StatsITR initializer - zero out statistics"""
+		attr_dict = self.__dict__
+		for attr in StatsObj.stat_file_attrs: attr_dict[attr] = 0
+		self.ElapsedTime = self.Filename = None
+		IterTreeReducer.__init__(self, *args)
 
 	def start_stats(self, mirror_dsrp):
 		"""Record status of mirror dsrp
@@ -272,8 +274,7 @@ class StatsITR(IterTreeReducer, StatsObj):
 	def add_file_stats(self, subinstance):
 		"""Add all file statistics from subinstance to current totals"""
 		for attr in self.stat_file_attrs:
-			self.set_stat(attr,
-						  self.get_stat(attr) + subinstance.get_stat(attr))
+			self.__dict__[attr] += subinstance.__dict__[attr]
 
 
 class Stats: