def feed_objects(self, objects, getattr=getattr):
""" Feeds a sequence of objects which is converted to CSV.
For each object the set column names are interpreted as
object attributes and used as basis for the CSV data.
None values are converted to empty strings, all other
attributes are added stringified.
"""
columns = self.columns
if not columns:
raise Error, 'no output columns set'
rowlen = len(columns)
# Create an emtpy table
rows = len(objects)
rowindices = Tools.trange(rows)
t = [None] * rows
for i in rowindices:
t[i] = [None] * rowlen
# Fill the table
icols = Tools.irange(columns)
for i in rowindices:
obj = objects[i]
for j, name in icols:
t[i][j] = str(getattr(obj, name))
# Quote and join lines
t = [self.separator.join(self._quote(x)) for x in t]
# Add final CRLF and store CSV text
t.append('')
self.text = self.text + self.lineend.join(t)
class _modinit:
import math
l = Tools.frange(0, 1, MAXLOCALITY)
for i, factor in Tools.irange(l):
_weights[i] = int((math.exp(factor) - 1.0) * 8192)
if _debug:
print i, '. weight =', _weights[i]
def clear_cache(self):
""" Clears the caches used (flushing any data not yet
written).
"""
if self.caching:
#self.flush()
Tools.method_mapply(self.caches, 'clear', ())
def cut(self, NOM=NOM, DENOM=DENOM):
""" Force a cut of the cache's contents.
This will make room for at least one new entry.
"""
if _debug:
print ' Cutting down cache size...'
cachesize = self.cachesize
# Cut the cache down to the entries in recent get history
newdata = {}
known_key = newdata.has_key
data = self.data
for id in self.get_history[-self.locality:]:
if known_key(id):
continue
try:
newdata[id] = data[id]
except KeyError:
pass
cachesize = len(newdata)
if _debug:
print ' Size after cut to recent history:', cachesize
# Check
if cachesize * NOM >= self.max_cachesize * DENOM:
# Calculate weights
d = {}
weights = _weights
d_get = d.get
for i, id in Tools.irange(self.get_history[-self.locality:]):
if not known_key(id):
continue
d[id] = d_get(id, 0) + weights[i]
# Delete all entries left from median
ranking = Tools.sortedby(d.items(), 1)
if _debug:
print ' Ranking:', ranking
for id, weight in ranking[:len(d) / 2]:
if _debug:
print ' Deleting', id, 'with weight =', weight
del newdata[id]
# Check
cachesize = len(newdata)
if cachesize * NOM >= self.max_cachesize * DENOM:
# Ok, so the smart way didn't work...
if _debug:
print ' Did not work, going the hard way...'
newdata.clear()
cachesize = 0
self.data = newdata
self.cachesize = cachesize
self.cuts = self.cuts + 1
def clear_cache(self):
""" Clears the caches used (flushing any data not yet
written).
"""
if self.caching:
#self.flush()
Tools.method_mapply(self.caches,'clear',())
def rpcdecode(url,
prefix='',
decode=1,
splitat=TextTools.splitat,
charsplit=TextTools.charsplit,
len=len,
tuple=tuple,
urldecode=urldecode):
""" Decode a RPC encoded function/method call.
Returns a tuple (name,args,kws) where args is a tuple of
string arguments and kws is a dictionary containing the given
keyword parameters or None. All parameters are returned as
strings; it is up to the caller to decode them into
e.g. integers, etc.
If prefix is given and found it is removed from the name prior
to returning it. decode can be set to false to prevent the url
from being urldecoded prior to processing.
The decode function also supports the syntax 'method' instead
of 'method()' for calls without arguments.
"""
if decode:
url = urldecode(url)
# Decode the method: method[(arg0,arg1,...,kw0=val0,kw1=val1,...)]
name, rawargs = splitat(url, '(')
if rawargs:
# Cut out the pure argument part, ignoring any character after
# the final ')'
rawargs, rest = splitat(rawargs, ')', -1)
# Argument list: split at ','
args = charsplit(rawargs, ',')
if '=' in rawargs:
kws = {}
for i, arg in Tools.reverse(Tools.irange(args)):
if '=' in arg:
k, v = splitat(arg, '=')
kws[k] = v
del args[i]
else:
kws = None
args = tuple(args)
else:
args = ()
kws = None
if prefix:
if name[:len(prefix)] == prefix:
name = name[len(prefix):]
return name, args, kws
else:
return name, args, kws
def rpcdecode(url,prefix='',decode=1,
splitat=TextTools.splitat,charsplit=TextTools.charsplit,
len=len,tuple=tuple,urldecode=urldecode):
""" Decode a RPC encoded function/method call.
Returns a tuple (name,args,kws) where args is a tuple of
string arguments and kws is a dictionary containing the given
keyword parameters or None. All parameters are returned as
strings; it is up to the caller to decode them into
e.g. integers, etc.
If prefix is given and found it is removed from the name prior
to returning it. decode can be set to false to prevent the url
from being urldecoded prior to processing.
The decode function also supports the syntax 'method' instead
of 'method()' for calls without arguments.
"""
if decode:
url = urldecode(url)
# Decode the method: method[(arg0,arg1,...,kw0=val0,kw1=val1,...)]
name,rawargs = splitat(url,'(')
if rawargs:
# Cut out the pure argument part, ignoring any character after
# the final ')'
rawargs,rest = splitat(rawargs,')',-1)
# Argument list: split at ','
args = charsplit(rawargs,',')
if '=' in rawargs:
kws = {}
for i,arg in Tools.reverse(Tools.irange(args)):
if '=' in arg:
k,v = splitat(arg,'=')
kws[k] = v
del args[i]
else:
kws = None
args = tuple(args)
else:
args = ()
kws = None
if prefix:
if name[:len(prefix)] == prefix:
name = name[len(prefix):]
return name,args,kws
else:
return name,args,kws
def free(self, position, OLD=OLD, HOT=HOT):
""" Deletes an already written record by marking it OLD.
The next garbage collection will make the change permanent
and free the occupied space.
"""
if self.state != HOT:
self.mark(HOT)
file = self.file
file.seek(position + 5)
file.write(OLD)
if self.caching:
Tools.method_mapply(self.caches, 'delete', (position, ))
def feed_list(self, table):
""" Feeds a table (list of rows) which is converted
to CSV.
No more than len(columns) items are written for each
row. All rows are filled up with "" entries to have an
equal number of items. None entries are converted to empty
strings, all other objects are stringified.
"""
columns = self.columns
if columns:
rowlen = len(columns)
else:
# Calculate the max. number of columns in the table
rowlen = max(map(len, table))
# Prepare an empty table
t = [None] * len(table)
_quote = self._quote
# Fill in data
for i, row in Tools.irange(table):
row = _quote(row[:rowlen])
if len(row) < rowlen:
row[len(row):] = ['""'] * (rowlen - len(row))
t[i] = self.separator.join(row)
# Add final CRLF and add as CSV text
t.append('')
self.text = self.text + self.lineend.join(t)
def print_stack(file=_sys.stdout,levels=100,offset=0,locals=0):
# Prepare frames
try:
raise ValueError
except ValueError:
# Go back offset+1 frames...
f = _sys.exc_info()[2].tb_frame
for i in range(offset + 1):
if f.f_back is not None:
f = f.f_back
# Extract frames
frames = []
while f:
frames.append(f)
f = f.f_back
frames.reverse()
# Prepare stack
stack = _traceback.extract_stack()
# Make output
file.write('Stack:\n')
for (frame,(filename, lineno, name, line)) in \
Tools.tuples(frames,stack)[-levels:]:
file.write(' File "%s", line %d, in %s\n' % (filename,lineno,name))
if line:
file.write(' %s\n' % line.strip())
if locals:
print_frame_locals(frame,file,indent=' |',title='')
def print_recursive(obj,file=_sys.stdout,indent='',levels=1,
nonrecursive=(),filter=None):
# Filter out nonrecursive types and objects
try:
if type(obj) in nonrecursive or \
obj in nonrecursive:
return
except:
# Error during compares result in the object not being
# printed
return
# Print the object depending on its interface
if hasattr(obj,'__dict__') and \
obj.__dict__ is not None:
print_dict(obj.__dict__,file,indent,levels,
nonrecursive=nonrecursive, filter=filter)
elif hasattr(obj,'items'):
print_dict(obj,file,indent,levels,1,
nonrecursive=nonrecursive, filter=filter)
elif Tools.issequence(obj) and not is_string(obj):
print_sequence(obj,file,indent,levels,
nonrecursive=nonrecursive)
elif hasattr(obj,'__members__'):
d = {}
for attr in obj.__members__:
d[attr] = getattr(obj,attr)
print_dict(d, file, indent, levels,
nonrecursive=nonrecursive, filter=filter)
def search_bench(word, text):
iterations = Tools.trange(COUNT)
print ('Searching for all occurences of %r using ...' % word)
t0 = time.time()
so = TextTools.TextSearch(word)
for i in iterations:
l = so.findall(text)
t1 = time.time()
count = len(l)
print (' - mx.TextSearch.TextSearch().findall(): %5.3f ms (%i)' %
((t1 - t0) / COUNT * 1000.0, count))
t0 = time.time()
so = re.compile(word)
for i in iterations:
l = so.findall(text)
t1 = time.time()
count = len(l)
print (' - re.compile().findall(): %5.3f ms (%i)' %
((t1 - t0) / COUNT * 1000.0, count))
t0 = time.time()
for i in iterations:
count = text.count(word)
t1 = time.time()
print (' - text.count(): %5.3f ms (%i)' %
((t1 - t0) / COUNT * 1000.0, count))
def print_recursive(obj,file=_sys.stdout,indent='',levels=1,
nonrecursive=(),filter=None):
# Filter out nonrecursive types and objects
try:
if type(obj) in nonrecursive or \
obj in nonrecursive:
return
except:
# Error during compares result in the object not being
# printed
return
# Print the object depending on its interface
if hasattr(obj,'__dict__') and \
obj.__dict__ is not None:
print_dict(obj.__dict__,file,indent,levels,
nonrecursive=nonrecursive, filter=filter)
elif hasattr(obj,'items'):
print_dict(obj,file,indent,levels,1,
nonrecursive=nonrecursive, filter=filter)
elif Tools.issequence(obj) and not is_string(obj):
print_sequence(obj,file,indent,levels,
nonrecursive=nonrecursive)
elif hasattr(obj,'__members__'):
d = {}
for attr in obj.__members__:
d[attr] = getattr(obj,attr)
print_dict(d, file, indent, levels,
nonrecursive=nonrecursive, filter=filter)
def print_stack(file=_sys.stdout, levels=100, offset=0, locals=0):
# Prepare frames
try:
raise ValueError
except ValueError:
# Go back offset+1 frames...
f = _sys.exc_info()[2].tb_frame
for i in range(offset + 1):
if f.f_back is not None:
f = f.f_back
# Extract frames
frames = []
while f:
frames.append(f)
f = f.f_back
frames.reverse()
# Prepare stack
stack = _traceback.extract_stack()
# Make output
file.write('Stack:\n')
for (frame,(filename, lineno, name, line)) in \
Tools.tuples(frames,stack)[-levels:]:
file.write(' File "%s", line %d, in %s\n' % (filename, lineno, name))
if line:
file.write(' %s\n' % line.strip())
if locals:
print_frame_locals(frame, file, indent=' |', title='')
def feed_list(self,table):
""" Feeds a table (list of rows) which is converted
to CSV.
No more than len(columns) items are written for each
row. All rows are filled up with "" entries to have an
equal number of items. None entries are converted to empty
strings, all other objects are stringified.
"""
columns = self.columns
if columns:
rowlen = len(columns)
else:
# Calculate the max. number of columns in the table
rowlen = max(map(len,table))
# Prepare an empty table
t = [None] * len(table)
_quote = self._quote
# Fill in data
for i,row in Tools.irange(table):
row = _quote(row[:rowlen])
if len(row) < rowlen:
row[len(row):] = ['""'] * (rowlen - len(row))
t[i] = self.separator.join(row)
# Add final CRLF and add as CSV text
t.append('')
self.text = self.text + self.lineend.join(t)
def free(self,position,
OLD=OLD,HOT=HOT):
""" Deletes an already written record by marking it OLD.
The next garbage collection will make the change permanent
and free the occupied space.
"""
if self.state != HOT:
self.mark(HOT)
file = self.file
file.seek(position + 5)
file.write(OLD)
if self.caching:
Tools.method_mapply(self.caches,'delete',(position,))
def objects(self, constructor):
""" Builds a list of objects by calling the given constructor
with keywords defined by mapping column names to values for
each input line.
.columns must have been set using .set_columns() or by
processing a given CSV header.
"""
lines = self.lines
keys = self.columns
if keys is None:
raise Error, 'no columns set'
objs = [None] * len(lines)
for i, line in Tools.irange(lines):
kws = dict(Tools.tuples(keys, line))
objs[i] = apply(constructor, (), kws)
return objs
def objects(self,constructor):
""" Builds a list of objects by calling the given constructor
with keywords defined by mapping column names to values for
each input line.
.columns must have been set using .set_columns() or by
processing a given CSV header.
"""
lines = self.lines
keys = self.columns
if keys is None:
raise Error,'no columns set'
objs = [None] * len(lines)
for i,line in Tools.irange(lines):
kws = dict(Tools.tuples(keys, line))
objs[i] = apply(constructor,(),kws)
return objs
def feed_dict(self,table,rows=None):
""" Feeds a table (dict of lists) which is converted
to CSV.
Only the keys set as column names are used to form the CSV
data.
All lists in the dictionary must have equal length or at
least rows number of entries, if rows is given. None
entries are converted to empty strings, all other objects
are stringified.
"""
columns = self.columns
if not columns:
raise Error,'no output columns set'
rowlen = len(columns)
# Create an emtpy table
if not rows:
rows = 0
for column in columns:
nrows = len(table[column])
if nrows > rows:
rows = nrows
rowindices = Tools.trange(rows)
t = [None] * rows
for i in rowindices:
t[i] = [None] * rowlen
# Fill the table
for j,k in Tools.irange(columns):
for i in rowindices:
t[i][j] = table[k][i]
# Quote and join lines
t = [self.separator.join(self._quote(x)) for x in t]
# Add final CRLF and store CSV text
t.append('')
self.text = self.text + self.lineend.join(t)
def feed_dict(self,table,rows=None):
""" Feeds a table (dict of lists) which is converted
to CSV.
Only the keys set as column names are used to form the CSV
data.
All lists in the dictionary must have equal length or at
least rows number of entries, if rows is given. None
entries are converted to empty strings, all other objects
are stringified.
"""
columns = self.columns
if not columns:
raise Error,'no output columns set'
rowlen = len(columns)
# Create an emtpy table
if not rows:
rows = 0
for column in columns:
nrows = len(table[column])
if nrows > rows:
rows = nrows
rowindices = Tools.trange(rows)
t = [None] * rows
for i in rowindices:
t[i] = [None] * rowlen
# Fill the table
for j,k in Tools.irange(columns):
for i in rowindices:
t[i][j] = table[k][i]
# Quote and join lines
t = [self.separator.join(self._quote(x)) for x in t]
# Add final CRLF and store CSV text
t.append('')
self.text = self.text + self.lineend.join(t)
def _quote(self, line, str=str):
""" CSV style quote the given line of text.
"""
nline = ['""'] * len(line)
for i, item in Tools.irange(line):
if item is not None:
text = str(item)
else:
text = ''
nline[i] = '"%s"' % text.replace('"', '""')
return nline
def _unquote(self, line):
""" Unquote a CSV style quoted line of text.
Internal method. Do not use directly.
"""
for i, text in Tools.irange(line):
if text[:1] == '"' and text[-1:] == '"':
text = text[1:-1]
line[i] = text.replace('""', '"')
return line
def _unquote(self,line):
""" Unquote a CSV style quoted line of text.
Internal method. Do not use directly.
"""
for i,text in Tools.irange(line):
if text[:1] == '"' and text[-1:] == '"':
text = text[1:-1]
line[i] = text.replace('""','"')
return line
def dictionary(self):
""" Return the current data as dictionary of lists of strings,
with one entry for each column.
.columns must have been set using .set_columns() or by
processing a given CSV header.
"""
table = {}
lines = self.lines
keys = self.columns
if keys is None:
raise Error, 'no columns set'
rows = len(lines)
for k in keys:
table[k] = [None] * rows
for i, key in Tools.irange(keys):
column = table[key]
for j, row in Tools.irange(lines):
if len(row) > i:
column[j] = row[i]
return table
def feed_objects(self,objects,
getattr=getattr):
""" Feeds a sequence of objects which is converted to CSV.
For each object the set column names are interpreted as
object attributes and used as basis for the CSV data.
None values are converted to empty strings, all other
attributes are added stringified.
"""
columns = self.columns
if not columns:
raise Error,'no output columns set'
rowlen = len(columns)
# Create an emtpy table
rows = len(objects)
rowindices = Tools.trange(rows)
t = [None] * rows
for i in rowindices:
t[i] = [None] * rowlen
# Fill the table
icols = Tools.irange(columns)
for i in rowindices:
obj = objects[i]
for j,name in icols:
t[i][j] = str(getattr(obj, name))
# Quote and join lines
t = [self.separator.join(self._quote(x)) for x in t]
# Add final CRLF and store CSV text
t.append('')
self.text = self.text + self.lineend.join(t)
def dictionary(self):
""" Return the current data as dictionary of lists of strings,
with one entry for each column.
.columns must have been set using .set_columns() or by
processing a given CSV header.
"""
table = {}
lines = self.lines
keys = self.columns
if keys is None:
raise Error,'no columns set'
rows = len(lines)
for k in keys:
table[k] = [None] * rows
for i, key in Tools.irange(keys):
column = table[key]
for j, row in Tools.irange(lines):
if len(row) > i:
column[j] = row[i]
return table
def filter_header(self, header, lower=TextTools.lower):
""" Filter the given header line.
The base class converts the column names to all lowercase
and removes any whitespace included in the header.
This method is only called in case the header was read
from the data provided to the object.
"""
l = [''] * len(header)
for i, column in Tools.irange(header):
l[i] = ''.join(lower(column).split())
return l
def _quote(self, line,
str=str):
""" CSV style quote the given line of text.
"""
nline = ['""'] * len(line)
for i,item in Tools.irange(line):
if item is not None:
text = str(item)
else:
text = ''
nline[i] = '"%s"' % text.replace('"','""')
return nline
def print_sequence(obj,
file=_sys.stdout,
indent='',
levels=2,
nonrecursive=()):
l = []
unfold = 0
try:
length = len(obj)
except (AttributeError, ValueError, TypeError):
return
for i in Tools.trange(min(length, _VALUE_LEN_LIMIT)):
try:
value = obj[i]
except:
break
try:
r = repr(value)
except:
r = '*repr()-error*'
# Truncate
if len(r) > _VALUE_LEN_LIMIT:
r = r[:_VALUE_LEN_LIMIT] + '...'
# Write value
l.append((value, r))
# Only unfold sequences that have non-string items or string items
# with more than on character
if not is_string(value) or len(value) > 1:
unfold = 1
if len(obj) > _VALUE_LEN_LIMIT:
l.append(('...', '...truncated...'))
# Unfold value object
if unfold:
for i, (value, rvalue) in irange(l):
file.write('%s%-15s = %s\n' % (indent, '[%i]' % i, rvalue))
if levels > 1:
print_recursive(value,
file,
indent + ' ',
levels - 1,
nonrecursive=nonrecursive)
def list(self):
""" Return the current data as list of lists, each having
self.width string entries.
Missing entries are set to None.
"""
width = self.width
lines = self.lines
table = [None] * len(lines)
for i, row in Tools.irange(lines):
row = row[:]
if len(row) < width:
row[len(row):] = [None] * (width - len(row))
table[i] = row
return table
def _quote(self, line,
str=str):
""" CSV style quote the given line of text.
"""
nline = ['""'] * len(line)
for i,item in Tools.irange(line):
if item is None:
text = ''
elif isinstance(item, unicode):
text = item.encode(self.encoding)
else:
text = str(item)
nline[i] = '"%s"' % text.replace('"','""')
return nline
def list(self):
""" Return the current data as list of lists, each having
self.width string entries.
Missing entries are set to None.
"""
width = self.width
lines = self.lines
table = [None] * len(lines)
for i, row in Tools.irange(lines):
row = row[:]
if len(row) < width:
row[len(row):] = [None]*(width-len(row))
table[i] = row
return table
def filter_header(self, header,
lower=TextTools.lower):
""" Filter the given header line.
The base class converts the column names to all lowercase
and removes any whitespace included in the header.
This method is only called in case the header was read
from the data provided to the object.
"""
l = [''] * len(header)
for i,column in Tools.irange(header):
l[i] = ''.join(lower(column).split())
return l
def __str__(self):
lines = self.list()
desc = self.description()
width = 0
output = []
write = output.append
for col in desc:
write('%-*s|' % (col[1], col[0]))
write('\n')
for col in desc:
write('=' * col[1] + '+')
write('\n')
for line in lines:
for i, item in Tools.irange(line):
write('%-*s|' % (desc[i][1], item))
write('\n')
return ''.join(output)
def __str__(self):
lines = self.list()
desc = self.description()
width = 0
output = []
write = output.append
for col in desc:
write('%-*s|' % (col[1],col[0]))
write('\n')
for col in desc:
write('=' * col[1] + '+')
write('\n')
for line in lines:
for i,item in Tools.irange(line):
write('%-*s|' % (desc[i][1],item))
write('\n')
return ''.join(output)
def print_sequence(obj,file=_sys.stdout,indent='',levels=2,
nonrecursive=()):
l = []
unfold = 0
try:
length = len(obj)
except (AttributeError, ValueError, TypeError):
return
for i in Tools.trange(min(length,_VALUE_LEN_LIMIT)):
try:
value = obj[i]
except:
break
try:
r = repr(value)
except:
r = '*repr()-error*'
# Truncate
if len(r) > _VALUE_LEN_LIMIT:
r = r[:_VALUE_LEN_LIMIT] + '...'
# Write value
l.append((value,r))
# Only unfold sequences that have non-string items or string items
# with more than on character
if not is_string(value) or len(value) > 1:
unfold = 1
if len(obj) > _VALUE_LEN_LIMIT:
l.append(('...','...truncated...'))
# Unfold value object
if unfold:
for i,(value,rvalue) in irange(l):
file.write('%s%-15s = %s\n' % (indent, '[%i]' % i, rvalue))
if levels > 1:
print_recursive(value,file,indent + ' ',levels-1,
nonrecursive=nonrecursive)
def description(self, header=1):
""" Return a list of tuples (column name, max length) found in the
data.
If header is true (default), the column names themselves
are included in the calculation.
"""
lines = self.lines
columns = self.columns
width = len(columns)
if header:
lengths = []
for column in columns:
lengths.append(len(column))
else:
lengths = [0] * width
for row in self.lines:
for i, o in Tools.irange(row[:width]):
if len(o) > lengths[i]:
lengths[i] = len(o)
return map(None, columns, lengths)
def description(self, header=1):
""" Return a list of tuples (column name, max length) found in the
data.
If header is true (default), the column names themselves
are included in the calculation.
"""
lines = self.lines
columns = self.columns
width = len(columns)
if header:
lengths = []
for column in columns:
lengths.append(len(column))
else:
lengths = [0] * width
for row in self.lines:
for i,o in Tools.irange(row[:width]):
if len(o) > lengths[i]:
lengths[i] = len(o)
return map(None,columns,lengths)
class WorkerProcess(object):
""" Worker process encapsulation.
These work a lot like server processes, except that they are
managed by daemon process as child processes.
The implementation uses two contexts:
- the server daemon context in which .start_server() and
.stop_server() are called
- the worker process contect in which .main() is run
The .main() method has to be overridden to implement the
worker process logic.
"""
# Note: This code is similar to ServerDaemon, but for worker
# processes we don't fork twice since we want the workers to be
# child processes of the server process.
# Name of the worker
name = 'Worker Process'
# PID of the worker process; set in both the server and the worker
# process context
pid = 0
# Started flag. Set by .start_worker()/.stop_worker() in the
# server context.
started = False
# Exit status code. Set by .worker_exited() in the server context.
exit_status = 0
# mxLog object to use. Inherited from the ServerDaemon if None
log = None
# Log id to use in the worker process. Inherited from the
# ServerDaemon if None
log_id = None
# Process name to use for the worker process. Note: this is not
# guaranteed to work. Inherited from the ServerDaemon if None
process_name = None
# Startup time of the worker processes in seconds. The
# .start_worker() method will wait this number of second for the
# worker process to start up.
worker_startup_time = 2
# Shutdown time of the worker processes in seconds. The
# .stop_worker() method will wait this number of second for the
# worker processes to terminate.
worker_shutdown_time = 2
# Kill time of the worker processes in seconds. The .stop_worker()
# method will wait this number of second for the worker processes
# to terminate after having received the KILL signal.
worker_kill_time = 1
# Range of file descriptors to close after the fork; all open fds
# except of stdin, stdout, stderr
close_file_descriptors = tuple(range(3, 99))
def __init__(self, server_daemon):
# Inherit settings from the server
if self.log is None:
self.log = server_daemon.log
if self.log is None:
self.log = Log.LogNothing
if self.log_id is None:
self.log_id = server_daemon.log_id
if self.process_name is None:
self.process_name = server_daemon.process_name
def __repr__(self):
return '%s(%s with PID %s)' % (self.__class__.__name__, self.name,
self.pid)
def setup_worker(self, **parameters):
""" Prepare the worker startup and adjust the parameters to be
passed on to the worker's .main() method.
This method is called by .start_worker() before forking
off a child process in order to give the WorkerProcess
implementation a chance to adjust itself to the
parameters.
It has to return a copy of the parameters keyword argument
dictionary.
This method is called in the context of the server.
"""
return parameters.copy()
def start_worker(self, **parameters):
""" Start the worker process and pass the given keyword parameters
to the .main() method.
"""
# Prepare startup
parameters = self.setup_worker(**parameters)
assert parameters is not None, \
'.setup_worker() did not return a parameters dictionary'
# Flush file descriptors
sys.stderr.flush()
sys.stdout.flush()
# Create a socket pair
server_socket, worker_socket = socket.socketpair(
socket.AF_UNIX, socket.SOCK_STREAM)
# Fork a child process, errors will be reported to the caller
pid = os.fork()
if pid != 0:
### Server process context ...
# Close our end of the socket pair
server_socket.close()
# Wait for the child to start up
worker_socket.settimeout(self.worker_startup_time)
try:
ok = worker_socket.recv(1)
except socket.timeout:
ok = None
worker_socket.close()
if not ok:
# Terminate the child, if it didn't startup in time
self.log(
self.log.ERROR, '%s: '
'Collecting worker process PID %s due to startup failure',
self.name, pid)
try:
self._kill_worker(pid)
except WorkerNotStoppedError:
pass
# Report the failure
raise WorkerNotStartedError(
'%s: Worker process with PID %s did not start up' %
(self.name, pid))
# Remember the worker process pid and return it
self.pid = pid
self.started = True
self.exit_status = 0
return pid
### Worker process context ...
# Close our end of the socket pair
worker_socket.close()
# Close all open fds except of stdin, stdout, stderr
self.log.close()
server_socket_fd = server_socket.fileno()
for i in self.close_file_descriptors:
if i == server_socket_fd:
# We'll close that manually later on
continue
try:
os.close(i)
except (IOError, OSError), reason:
pass
# Reopen the log file
self.log.open()
if self.log_id:
self.log.setup(log_id=self.log_id)
# Redirect stdout and stderr to the log file
self.log.redirect_stdout()
self.log.redirect_stderr()
# Try to rename the process
if self.process_name:
try:
Tools.setproctitle(self.process_name)
except AttributeError:
pass
# Set the PID of the worker process
self.pid = os.getpid()
# Let the server process know that we've started up
server_socket.send('1')
server_socket.close()
# Run the .main() method
rc = 0
try:
try:
self.log(self.log.INFO, '%s: Worker process PID %s %s',
self.name, self.pid, '-' * 40)
if _debug > 1:
self.log.object(
self.log.DEBUG,
'%s: Using the following startup parameters:' %
self.name, parameters)
# Run the worker's .main() method
main_rc = self.main(**parameters)
# Return the exit code, if it's an integer
if main_rc is not None and isinstance(main_rc, int):
rc = main_rc
except Exception:
# Something unexpected happened... log the problem and exit
self.log.traceback(self.log.ERROR, '%s: '
'Unexpected worker process error:',
self.name)
rc = 1
finally:
self.cleanup_worker()
# Exit process
os._exit(rc)
class ServerDaemon(object):
""" Server daemon encapsulation.
This class provides an easy way to setup a Unix server daemon
that uses a single process. It may still spawn off additional
processes, but this encapsulation only manages the main
process.
The implementation runs two contexts:
- the control context in which .start_server() and .stop_server()
are called
- the server process contect in which .main() is run
"""
# Name of the server
name = 'Server Daemon'
# PID of the process
pid = 0
# Location of the PID file of the parent process
pid_file = 'server.pid'
# umask to set for the forked server process
umask = 022
# Root dir to change to for the forked server process
root_dir = ''
# Range of file descriptors to close after the fork; all open fds
# except of stdin, stdout, stderr
close_file_descriptors = tuple(range(3, 99))
# mxLog object to use
log = Log.log
# Log id to use in the forked server process
log_id = ''
# Process name to use for the forked server process. Note: this is
# not guaranteed to work
process_name = ''
# Server startup time in seconds. The .start_server()
# method will wait at most this number of seconds for the main
# server process to initialize and enter the .main() method. This
# includes forking overhead, module import times, etc. It does not
# cover the startup time that the server may need to become usable
# for external applications. The startup time can be configured
# with .server_startup_time
server_startup_time = 2
# Startup initialization time of the server in seconds. The
# .start_server() method will unconditionally wait this number of
# seconds after having initialized the server in order to give the
# .main() method a chance to setup any resources it may need to
# initialize.
server_startup_init_time = 0
# Server shutdown time in seconds. The .stop_server()
# method will wait at most this number of seconds for the main
# server process to terminate after sending it a TERM signal.
server_shutdown_time = 2
# Kill time of the server processes in seconds. The .stop_server()
# method will wait this number of second for the worker processes
# to terminate after having received the KILL signal.
server_kill_time = 1
# Shutdown cleanup time of the server in seconds. The
# .stop_server() method will unconditionally wait this number of
# seconds after having terminated the main server process in order
# to give possibly additionally spawned processes a chance to
# terminate cleanly as well.
server_shutdown_cleanup_time = 0
###
def setup_server(self, **parameters):
""" Prepare the server startup and adjust the parameters to be
passed on to the server's .main() method.
This method is called by .start_server() before forking
off a child process in order to give the WorkerProcess
implementation a chance to adjust itself to the
parameters.
It has to return a copy of the parameters keyword argument
dictionary.
This method is called in the context of the server.
"""
return parameters.copy()
def _kill_server(self, pid):
""" Kill a server process pid and collect it.
Returns the process exit status or -1 in case this cannot
be determined.
Raises a ServerNotStoppedError in case the process cannot
be stopped.
"""
try:
return kill_process(pid,
shutdown_time=self.server_shutdown_time,
kill_time=self.server_kill_time,
log=self.log,
log_prefix='%s: ' % self.name)
except ProcessNotStoppedError:
# Did not work out...
raise ServerNotStoppedError(
'%s: Server process with PID %s did not stop' %
(self.name, pid))
def start_server(self, **parameters):
""" Starts the server.
Keyword parameters are passed on to the forked process'
.main() method.
Returns the PID of the started server daemon.
Raises a ServerAlreadyRunningError if the server is
already running. Raises a ServerNotStartedError in case
the daemon could not be started.
"""
# Verify if we have a running server process
pid = self.server_status()
if pid is not None:
raise ServerAlreadyRunningError(
'Server is already running (PID %s)' % pid)
# Prepare startup
parameters = self.setup_server(**parameters)
assert parameters is not None, \
'.setup_server() did not return a parameters dictionary'
# Flush the standard file descriptors
sys.stderr.flush()
sys.stdout.flush()
# Fork a child process, errors will be reported to the caller
pid = os.fork()
if pid != 0:
### Parent process
# Collect the first child
if _debug:
self.log(
self.log.DEBUG, '%s: '
'Waiting for the first child with PID %s to terminate',
self.name, pid)
os.waitpid(pid, 0)
# Wait a few seconds until the server has started
if _debug:
self.log(self.log.DEBUG, '%s: '
'Waiting for the server process to startup',
self.name)
for i in xrange(int(self.server_startup_time * 100) + 1):
spid = self.server_status()
if spid is not None:
break
time.sleep(0.01)
else:
# Server did not startup in time: terminate the first
# child
self.log(self.log.ERROR,
'%s: Server process failed to startup', self.name)
try:
self._kill_server(pid)
except ServerNotStoppedError:
pass
# Report the problem; XXX Note that the second child
# may still startup after this first has already
# terminated.
raise ServerNotStartedError('%s did not start up' % self.name)
if self.server_startup_init_time:
time.sleep(self.server_startup_init_time)
return spid
### This is the first child process
# Daemonize process
os.setpgrp()
if self.root_dir:
os.chdir(self.root_dir)
if self.umask:
os.umask(self.umask)
try:
# Try to become a session leader
os.setsid()
except OSError:
# We are already the process session leader
pass
# Close all open fds except of stdin, stdout, stderr
self.log.close()
for i in self.close_file_descriptors:
try:
os.close(i)
except (IOError, OSError), reason:
pass
# Fork again to become a separate daemon process
pid = os.fork()
if pid != 0:
# We need to terminate the "middle" process at this point, since we
# don't want to continue with two instances of the original caller.
# We must not call any cleanup handlers here.
os._exit(0)
### This is the second child process: the server daemon
# Turn the daemon into a process group leader
os.setpgrp()
# Reopen the log file
self.log.open()
if self.log_id:
self.log.setup(log_id=self.log_id)
# Redirect stdout and stderr to the log file
self.log.redirect_stdout()
self.log.redirect_stderr()
# Try to rename the process
if self.process_name:
try:
Tools.setproctitle(self.process_name)
except AttributeError:
pass
# Save the PID of the server daemon process
self.pid = os.getpid()
self.save_server_pid(self.pid)
# We need to remove the PID file on exit
rc = 0
try:
try:
self.log(self.log.INFO, '%s: Server process PID %s %s',
self.name, self.pid, '*' * 60)
# Run the server's .main() method
main_rc = self.main(**parameters)
# Return the exit code, if it's an integer
if main_rc is not None and isinstance(main_rc, int):
rc = main_rc
except SystemExit, exc:
# Normal shutdown
rc = exc.code
self.log(self.log.INFO, '%s: Shutting down with status: %s',
self.name, rc)
except Exception:
# Something unexpected happened... log the problem and exit
self.log.traceback(self.log.ERROR,
'%s: Unexpected server error:', self.name)
rc = 1
def _set_log_id(self, value):
self._custom_log_id = value
log_id = property(_get_log_id, _set_log_id)
### LogNothing disabled logging completely
class LogNothing(Log):
# Don't log anything
ignore_level = SYSTEM_LOG_NOTHING
def setup(self, *args, **kws):
Log.setup(*args, **kws)
self.ignore_level = SYSTEM_LOG_NOTHING
def open(self, flags='a'):
self.open_log_file = None
def close(self):
pass
###
# Create a main log object
if __debug__ and Tools.debugging():
log = Log(SYSTEM_LOG_EVERYTHING)
else:
log = Log(SYSTEM_INFO + 1)
def cut(self,
NOM=NOM,DENOM=DENOM):
""" Force a cut of the cache's contents.
This will make room for at least one new entry.
"""
if _debug:
print ' Cutting down cache size...'
cachesize = self.cachesize
# Cut the cache down to the entries in recent get history
newdata = {}
known_key = newdata.has_key
data = self.data
for id in self.get_history[-self.locality:]:
if known_key(id):
continue
try:
newdata[id] = data[id]
except KeyError:
pass
cachesize = len(newdata)
if _debug:
print ' Size after cut to recent history:',cachesize
# Check
if cachesize * NOM >= self.max_cachesize * DENOM:
# Calculate weights
d = {}
weights = _weights
d_get = d.get
for i,id in Tools.irange(self.get_history[-self.locality:]):
if not known_key(id):
continue
d[id] = d_get(id,0) + weights[i]
# Delete all entries left from median
ranking = Tools.sortedby(d.items(),1)
if _debug:
print ' Ranking:',ranking
for id,weight in ranking[:len(d)/2]:
if _debug:
print ' Deleting',id,'with weight =',weight
del newdata[id]
# Check
cachesize = len(newdata)
if cachesize * NOM >= self.max_cachesize * DENOM:
# Ok, so the smart way didn't work...
if _debug:
print ' Did not work, going the hard way...'
newdata.clear()
cachesize = 0
self.data = newdata
self.cachesize = cachesize
self.cuts = self.cuts + 1
log_id = property(_get_log_id, _set_log_id)
### LogNothing disabled logging completely
class LogNothing(Log):
# Don't log anything
ignore_level = SYSTEM_LOG_NOTHING
def setup(self, *args, **kws):
Log.setup(*args, **kws)
self.ignore_level = SYSTEM_LOG_NOTHING
def open(self, flags='a'):
self.open_log_file = None
def close(self):
pass
###
# Create a main log object
if __debug__ and Tools.debugging():
log = Log(SYSTEM_LOG_EVERYTHING)
else:
log = Log(SYSTEM_INFO + 1)
