# This module includes all exceptions raised by caproto, sentinel objects used
# throughout the package (see detailed comment below), various network-related
# helper functions, and other miscellaneous utilities.
import argparse
import array
import collections
import functools
import inspect
import os
import random
import socket
import sys
import enum
import json
import threading
from collections import namedtuple
from contextlib import contextmanager
from warnings import warn
import weakref
from ._version import get_versions
__version__ = get_versions()['version']
try:
import fcntl
import termios
except ImportError:
fcntl = None
termios = None
# fcntl is unavailale on windows
try:
import netifaces
except ImportError:
netifaces = None
__all__ = ( # noqa F822
'adapt_old_callback_signature',
'apply_arr_filter',
'ChannelFilter',
'get_environment_variables',
'get_address_list',
'get_server_address_list',
'get_beacon_address_list',
'get_netifaces_addresses',
'ensure_bytes',
'random_ports',
'bcast_socket',
'buffer_list_slice',
'incremental_buffer_list_slice',
'send_all',
'async_send_all',
'parse_record_field',
'parse_channel_filter',
'batch_requests',
'CaprotoError',
'ProtocolError',
'LocalProtocolError',
'RemoteProtocolError',
'CaprotoTimeoutError',
'CaprotoKeyError',
'CaprotoAttributeError',
'CaprotoNotImplementedError',
'CaprotoValueError',
'CaprotoTypeError',
'CaprotoConversionError',
'CaprotoRuntimeError',
'CaprotoNetworkError',
'ErrorResponseReceived',
'SendAllRetry',
'RecordModifiers',
'RecordModifier',
'RecordAndField',
'ThreadsafeCounter',
'__version__',
# sentinels dynamically defined and added to globals() below
'CLIENT', 'SERVER', 'RESPONSE', 'REQUEST', 'NEED_DATA',
'SEND_SEARCH_REQUEST', 'AWAIT_SEARCH_RESPONSE',
'SEND_SEARCH_RESPONSE', 'SEND_VERSION_REQUEST',
'AWAIT_VERSION_RESPONSE', 'SEND_VERSION_RESPONSE',
'SEND_CREATE_CHAN_REQUEST', 'AWAIT_CREATE_CHAN_RESPONSE',
'SEND_CREATE_CHAN_RESPONSE', 'CONNECTED', 'MUST_CLOSE',
'CLOSED', 'IDLE', 'FAILED', 'DISCONNECTED')
# This module defines sentinels used in the state machine and elsewhere, such
# as NEED_DATA, CLIENT, SERVER, AWAITING_SEARCH_RESPONSE, etc.
# It also defines some custom Exceptions.
class _SimpleReprEnum(enum.Enum):
def __repr__(self):
return self.name
class Role(_SimpleReprEnum):
CLIENT = enum.auto()
SERVER = enum.auto()
class Direction(_SimpleReprEnum):
RESPONSE = enum.auto()
REQUEST = enum.auto()
class States(_SimpleReprEnum):
SEND_SEARCH_REQUEST = enum.auto()
AWAIT_SEARCH_RESPONSE = enum.auto()
SEND_SEARCH_RESPONSE = enum.auto()
SEND_VERSION_REQUEST = enum.auto()
AWAIT_VERSION_RESPONSE = enum.auto()
SEND_VERSION_RESPONSE = enum.auto()
SEND_CREATE_CHAN_REQUEST = enum.auto()
AWAIT_CREATE_CHAN_RESPONSE = enum.auto()
SEND_CREATE_CHAN_RESPONSE = enum.auto()
CONNECTED = enum.auto()
MUST_CLOSE = enum.auto()
CLOSED = enum.auto()
IDLE = enum.auto()
FAILED = enum.auto()
DISCONNECTED = enum.auto()
# Special old 'sentinel' for needing more data
NEED_DATA = enum.auto()
class ConversionDirection(_SimpleReprEnum):
FROM_WIRE = enum.auto()
TO_WIRE = enum.auto()
globals().update(
{token: getattr(_enum, token)
for _enum in [Role, Direction, States]
for token in dir(_enum)
if not token.startswith('_')
})
class CaprotoError(Exception):
# All exceptions raised by this codebase inherit from this.
...
class ProtocolError(CaprotoError):
# Any error resulting from sending or receiving a command will raise (a
# subclass of) this error and never any other error.
...
[docs]class LocalProtocolError(ProtocolError):
"""
You tried to do something that caproto thinks is illegal.
"""
...
[docs]class RemoteProtocolError(ProtocolError):
"""
Your remote peer tried to do something that caproto thinks is illegal.
"""
...
class ValidationError(CaprotoError):
"""
Could not parse into valid command.
"""
...
class CaprotoTimeoutError(TimeoutError, CaprotoError):
...
class CaprotoAttributeError(AttributeError, CaprotoError):
...
class CaprotoKeyError(KeyError, CaprotoError):
...
class CaprotoNotImplementedError(NotImplementedError, CaprotoError):
...
class CaprotoValueError(ValueError, CaprotoError):
...
class CaprotoConversionError(CaprotoValueError):
...
class CaprotoEnvironmentSetupError(ValueError, CaprotoError):
...
class FilterValidationError(CaprotoValueError):
...
class CaprotoTypeError(TypeError, CaprotoError):
...
class CaprotoRuntimeError(RuntimeError, CaprotoError):
...
class CaprotoNetworkError(OSError, CaprotoError):
...
class ErrorResponseReceived(CaprotoError):
...
_ENVIRONMENT_DEFAULTS = dict(
EPICS_CA_ADDR_LIST='',
EPICS_CA_AUTO_ADDR_LIST='YES',
EPICS_CA_CONN_TMO=30.0,
EPICS_CA_BEACON_PERIOD=15.0,
EPICS_CA_REPEATER_PORT=5065,
EPICS_CA_SERVER_PORT=5064,
EPICS_CA_MAX_ARRAY_BYTES=16384,
EPICS_CA_MAX_SEARCH_PERIOD=300,
EPICS_TS_MIN_WEST=360,
EPICS_CAS_SERVER_PORT=5064,
EPICS_CAS_AUTO_BEACON_ADDR_LIST='YES',
EPICS_CAS_BEACON_ADDR_LIST='',
EPICS_CAS_BEACON_PERIOD=15.0,
EPICS_CAS_BEACON_PORT=5065,
EPICS_CAS_INTF_ADDR_LIST='',
EPICS_CAS_IGNORE_ADDR_LIST='',
)
def get_environment_variables():
'''Get a dictionary of known EPICS environment variables'''
result = dict(os.environ)
for key, default_value in _ENVIRONMENT_DEFAULTS.items():
type_of_env_var = type(default_value)
try:
result[key] = type_of_env_var(result[key])
except KeyError:
result[key] = default_value
except Exception as ex:
raise CaprotoEnvironmentSetupError(f'Environment variable {key} misconfigured: '
f'{ex.__class__.__name__} {ex}') from ex
return result
def _split_address_list(addr_list):
'''Split an address list string into individual items'''
return list(set(addr for addr in addr_list.split(' ') if addr.strip()))
def get_manually_specified_beacon_addresses():
'''Get a list of addresses, as configured by EPICS_CA_ADDR_LIST'''
return _split_address_list(
get_environment_variables()['EPICS_CAS_BEACON_ADDR_LIST'])
def get_manually_specified_client_addresses():
'''Get a list of addresses, as configured by EPICS_CA_ADDR_LIST'''
return _split_address_list(
get_environment_variables()['EPICS_CA_ADDR_LIST'])
def get_address_list():
'''Get channel access client address list based on environment variables
If the address list is set to be automatic, the network interfaces will be
scanned and used to determine the broadcast addresses available.
'''
env = get_environment_variables()
addresses = get_manually_specified_client_addresses()
if addresses and env['EPICS_CA_AUTO_ADDR_LIST'].lower() != 'yes':
# Custom address list specified, and EPICS_CA_AUTO_ADDR_LIST=NO
auto_addr_list = []
else:
# No addresses configured or EPICS_CA_AUTO_ADDR_LIST=YES
if netifaces is not None:
auto_addr_list = [bcast for _, bcast in get_netifaces_addresses()]
else:
auto_addr_list = ['255.255.255.255']
return addresses + auto_addr_list
def get_server_address_list():
'''Get the server interfaces based on environment variables
Returns
-------
list of interfaces
'''
intf_addrs = get_environment_variables()['EPICS_CAS_INTF_ADDR_LIST']
if not intf_addrs:
return ['0.0.0.0']
def strip_port(addr):
if ':' in addr:
addr, _, specified_port = addr.partition(':')
warn("Port specified in EPICS_CAS_INTF_ADDR_LIST was ignored.")
return addr
return [strip_port(addr) for addr in _split_address_list(intf_addrs)]
def get_beacon_address_list():
'''Get channel access beacon address list based on environment variables
If the address list is set to be automatic, the network interfaces will be
scanned and used to determine the broadcast addresses available.
Returns
-------
addr_list : list of (addr, beacon_port)
'''
env = get_environment_variables()
auto_addr_list = env['EPICS_CAS_AUTO_BEACON_ADDR_LIST']
addr_list = get_manually_specified_beacon_addresses()
beacon_port = env['EPICS_CAS_BEACON_PORT']
def get_addr_port(addr):
if ':' in addr:
addr, _, specified_port = addr.partition(':')
return (addr, int(specified_port))
return (addr, beacon_port)
if addr_list and auto_addr_list.lower() != 'yes':
# Custom address list and EPICS_CAS_AUTO_BEACON_ADDR_LIST=NO
auto_list = []
else:
auto_list = [('255.255.255.255', beacon_port)]
# Custom address list and EPICS_CAS_AUTO_BEACON_ADDR_LIST=YES
return addr_list + auto_list
def get_netifaces_addresses():
'''Get a list of addresses + broadcast using netifaces
Yields (address, broadcast_address)
'''
if netifaces is None:
raise CaprotoRuntimeError('netifaces unavailable')
for iface in netifaces.interfaces():
interface = netifaces.ifaddresses(iface)
if netifaces.AF_INET in interface:
for af_inet_info in interface[netifaces.AF_INET]:
addr = af_inet_info.get('addr', None)
peer = af_inet_info.get('peer', None)
broadcast = af_inet_info.get('broadcast', None)
if addr is not None and broadcast is not None:
yield (addr, broadcast)
elif peer is not None and broadcast is not None:
yield (peer, broadcast)
elif addr is not None:
yield (addr, addr)
elif peer is not None:
yield (peer, peer)
def ensure_bytes(s):
"""Encode string as bytes with null terminator. Bytes pass through."""
if isinstance(s, bytes):
return s
elif isinstance(s, str):
# be sure to include a null terminator
return s.encode() + b'\0'
else:
raise CaprotoTypeError(f"expected str or bytes, got {s!r} of type "
f"{type(s)}")
def random_ports(num, *, try_first=None):
"""Yield `num` random port numbers, optionally `try_first`."""
if try_first is not None:
yield try_first
for _ in range(num):
yield random.randint(49152, 65535)
def bcast_socket(socket_module=socket):
"""
Make a socket and configure it for UDP broadcast.
Parameters
----------
socket_module: module, optional
Default is the built-in :mod:`socket` module, but anything with the
same interface may be used, such as :mod:`curio.socket`.
"""
sock = socket_module.socket(socket_module.AF_INET, socket_module.SOCK_DGRAM)
try:
sock.setsockopt(socket_module.SOL_SOCKET, socket_module.SO_REUSEADDR, 1)
except AttributeError:
if sys.platform != 'win32':
raise
# WIN32_TODO: trio compatibility
import socket as system_socket
sock.setsockopt(socket_module.SOL_SOCKET,
system_socket.SO_REUSEADDR, 1)
# sock.setsockopt(socket_module.SOL_SOCKET,
# system_socket.SO_EXCLUSIVEADDRUSE, 1)
sock.setsockopt(socket_module.SOL_SOCKET, socket_module.SO_BROADCAST, 1)
# for BSD/Darwin only
try:
socket_module.SO_REUSEPORT
except AttributeError:
...
else:
sock.setsockopt(socket_module.SOL_SOCKET, socket_module.SO_REUSEPORT, 1)
return sock
if 'pypy' in sys.implementation.name:
def _cast_buffers_to_byte(buffers):
def inner(b):
try:
return memoryview(b).cast('b')
except TypeError:
target_type = b.dtype.str.replace('>', '<')
return memoryview(b.astype(target_type)).cast('b')
return tuple(inner(b) for b in buffers)
else:
def _cast_buffers_to_byte(buffers):
return tuple(memoryview(b).cast('b') for b in buffers)
def buffer_list_slice(*buffers, offset):
'Helper function for slicing a list of buffers'
if offset < 0:
raise CaprotoValueError('Negative offset')
buffers = _cast_buffers_to_byte(buffers)
start = 0
for bufidx, buf in enumerate(buffers):
end = start + len(buf)
if offset < end:
offset -= start
return (buf[offset:], ) + buffers[bufidx + 1:]
start = end
raise CaprotoValueError('Offset beyond end of buffers '
'(total length={} offset={})'.format(end, offset))
def incremental_buffer_list_slice(*buffers):
'Incrementally slice a list of buffers'
buffers = _cast_buffers_to_byte(buffers)
total_size = sum(len(b) for b in buffers)
total_sent = 0
while total_sent < total_size:
sent = yield buffers
total_sent += sent
if total_sent == total_size:
break
buffers = buffer_list_slice(*buffers, offset=sent)
class SendAllRetry(CaprotoError):
...
def send_all(buffers_to_send, send_func):
'''Incrementally slice a list of buffers, and send it using `send_func`
Parameters
----------
buffers_to_send : (buffer1, buffer2, ...)
Buffers are expected to be memoryviews or similar
send_func : callable
Function to call with list of buffers to send
Expected to return number of bytes sent or raise SendAllRetry otherwise
'''
if not buffers_to_send:
return
gen = incremental_buffer_list_slice(*buffers_to_send)
# prime the generator
gen.send(None)
while buffers_to_send:
try:
while True:
try:
sent = send_func(buffers_to_send)
break
except OSError:
buffers_to_send = buffers_to_send[:len(buffers_to_send) // 2]
except SendAllRetry:
continue
try:
buffers_to_send = gen.send(sent)
except StopIteration:
# finished sending
break
async def async_send_all(buffers_to_send, async_send_func):
'''Incrementally slice a list of buffers, and send it using `send_func`
Parameters
----------
buffers_to_send : (buffer1, buffer2, ...)
Buffers are expected to be memoryviews or similar
async_send_func : callable
Async function to call with list of buffers to send
Expected to return number of bytes sent or raise SendAllRetry otherwise
'''
if not buffers_to_send:
return
gen = incremental_buffer_list_slice(*buffers_to_send)
# prime the generator
gen.send(None)
while buffers_to_send:
try:
while True:
try:
sent = await async_send_func(buffers_to_send)
break
except OSError:
buffers_to_send = buffers_to_send[:len(buffers_to_send) // 2]
except SendAllRetry:
continue
try:
buffers_to_send = gen.send(sent)
except StopIteration:
# finished sending
break
RecordAndField = namedtuple('RecordAndField',
['record_dot_field', 'record',
'field', 'modifiers'])
class RecordModifiers(enum.Flag):
long_string = enum.auto()
filtered = enum.auto()
filter_timestamp = enum.auto()
filter_deadband = enum.auto()
filter_array = enum.auto()
filter_synchronize = enum.auto()
RecordModifier = namedtuple('RecordModifier',
['modifiers', 'filter_'])
def parse_record_field(pvname):
'''Parse a record[.FIELD][{FILTER}]
Returns
-------
RecordAndField(record_dot_field, record, field, modifiers)
'''
if '.' not in pvname:
return RecordAndField(pvname, pvname, None, None)
record, field = pvname.split('.', 1)
if field.startswith('{'):
field, filter_ = None, field
modifiers = RecordModifiers.filtered
elif not field:
# valid case of "{record}."
return RecordAndField(record, record, None, None)
else:
if '{' in field:
field, filter_ = field.split('{', 1)
filter_ = '{' + filter_
modifiers = RecordModifiers.filtered
elif '[' in field and field.endswith(']'):
field, filter_ = field.split('[', 1)
filter_ = '[' + filter_
modifiers = RecordModifiers.filtered
else:
filter_ = None
modifiers = None
if field.endswith('$'):
if modifiers is not None:
modifiers |= RecordModifiers.long_string
else:
modifiers = RecordModifiers.long_string
field = field[:-1]
# NOTE: VAL is equated to 'record' at a higher level than this.
if field:
record_field = f'{record}.{field}'
else:
record_field = record
if modifiers:
modifiers = RecordModifier(modifiers, filter_=filter_)
return RecordAndField(record_field, record, field, modifiers)
ChannelFilter = namedtuple('ChannelFilter', 'ts dbnd arr sync')
# TimestampFilter is just True or None, no need for namedtuple.
DeadbandFilter = namedtuple('DeadbandFilter', 'm d')
ArrayFilter = namedtuple('ArrayFilter', 's i e')
SyncFilter = namedtuple('SyncFilter', 'm s')
sync_modes = set(['before', 'first', 'while', 'last', 'after', 'unless'])
def parse_channel_filter(filter_text):
"Parse and validate filter_text into a ChannelFilter."
# https://epics.anl.gov/base/R3-15/5-docs/filters.html
if not filter_text:
return ChannelFilter(ts=False, dbnd=None, sync=None, arr=None)
# If there is a shorthand array filter, that is the only filter allowed, so
# we parse that and return, shortcircuiting the rest.
if filter_text.startswith('[') and filter_text.endswith(']'):
arr = parse_arr_shorthand_filter(filter_text)
return ChannelFilter(ts=False, dbnd=None, sync=None, arr=arr)
try:
filter_ = json.loads(filter_text)
except Exception as exc:
raise FilterValidationError(
f"Unable to parse channel filter text as JSON: "
f"{filter_text}") from exc
# If there is a shorthand sync filter, that is the only filter allowed.
# Rewrite the filter to expand the shorthand and go through the normal
# codepath.
intersection = sync_modes & set(filter_)
if intersection:
if len(filter_) > 1:
raise FilterValidationError(
f"Found shorthand sync filter key {next(iter(intersection))} "
f"This must be the only key if present but additional keys "
f"were found: {filter_}")
(mode, state), = filter_.items()
filter_ = {"sync": {"m": mode, "s": state}}
valid_filters = {'ts', 'arr', 'sync', 'dbnd'}
filter_keys = set(filter_)
invalid_keys = filter_keys - valid_filters
if invalid_keys:
raise FilterValidationError(f'Unsupported filters: {invalid_keys}')
# Validate and normalize the filter, expanding "shorthand" notation and
# applying defaults.
return ChannelFilter(arr=parse_arr_filter(filter_.get('arr')),
dbnd=parse_dbnd_filter(filter_.get('dbnd')),
ts=parse_ts_filter(filter_.get('ts')),
sync=parse_sync_filter(filter_.get('sync')))
def parse_arr_shorthand_filter(filter_text):
elements = []
for elem in filter_text[1:-1].split(':'):
if not elem:
elements.append(None)
else:
elements.append(int(elem))
if len(elements) == 1:
arr = ArrayFilter(s=elements[0], i=1, e=elements[0])
elif len(elements) == 2:
arr = ArrayFilter(s=elements[0], i=1, e=elements[1])
elif len(elements) == 3:
arr = ArrayFilter(s=elements[0], i=elements[1], e=elements[2])
return arr
def parse_ts_filter(val):
if val is None:
return None
# Empty object means 'true' according to the spec:
# https://epics.anl.gov/base/R3-15/3-docs/filters.html
if val == {}:
return True
# We'll accept `true` and any other truth-y value also....
return bool(val)
def parse_dbnd_filter(val):
if val is None:
return None
if 'rel' in val:
invalid_keys = set(val.keys()) - set(['rel'])
if invalid_keys:
raise FilterValidationError(
f"Unsupported keys in 'dbnd': {invalid_keys}. When 'rel' "
f"shorthand is used, no other keys may be used.")
return DeadbandFilter(m='rel', d=float(val['rel']))
if 'abs' in val:
invalid_keys = set(val.keys()) - set(['abs'])
if invalid_keys:
raise FilterValidationError(
f"Unsupported keys in 'dbnd': {invalid_keys}. When 'abs' "
f"shorthand is used, no other keys may be used.")
return DeadbandFilter(m='abs', d=float(val['abs']))
else:
invalid_keys = set(val.keys()) - set('dm')
if invalid_keys:
raise FilterValidationError(
f"Unsupported keys in 'dbnd': {invalid_keys}")
if set('md') != set(val.keys()):
raise FilterValidationError(
f"'dbnd' must include 'rel' or 'abs' or both 'd' and 'm'. "
f"Found keys {set(val.keys())}.")
return DeadbandFilter(m=float(val['m']), d=float(val['d']))
def parse_arr_filter(val):
if val is None:
return None
invalid_keys = set(val.keys()) - set('sie')
if invalid_keys:
raise FilterValidationError(f"Unsupported keys in 'arr': "
f"{invalid_keys}")
return ArrayFilter(s=int(val.get('s', 0)),
i=int(val.get('i', 1)),
e=int(val.get('e', -1)))
def parse_sync_filter(val):
if val is None:
return None
if set('ms') != set(val.keys()):
raise FilterValidationError(
f"'sync' must include both 'm' and 's'. "
f"Found keys {set(val.keys())}.")
if val['m'] not in sync_modes:
raise FilterValidationError(f"Unsupported mode in 'sync': "
f"{val['m']}")
if not isinstance(val['s'], str):
raise FilterValidationError(f"Unsupported type in 'sync': "
f"value 's' must be a string. "
f"Found {repr(val['s'])}.")
return SyncFilter(m=val['m'], s=val['s'])
def apply_arr_filter(arr_filter, values):
# Apply array Channel Filter.
if arr_filter is None:
return values
start, stop, step = arr_filter.s, arr_filter.e, arr_filter.i
# Cope with CA slice conventions being different from
# Python's. It specifies an interval closed on both ends,
# whereas Python's open open on the right end.
if stop is not None:
if stop == -1:
stop = None
else:
stop += 1
return values[start:stop:step]
def batch_requests(request_iter, max_length):
'''Batch a set of items with length, thresholded on sum of item length
Yields
------
batch : collections.deque
Batch of items from request_iter, where:
sum(len(b) for b in batch) < max_length
NOTE: instance of deque is reused, cleared at each iteration
'''
size = 0
batch = collections.deque()
for command in request_iter:
_len = len(command)
if size + _len > max_length:
yield batch
batch.clear()
size = 0
batch.append(command)
size += _len
if batch:
yield batch
class ThreadsafeCounter:
'''A thread-safe counter with a couple features:
1. Loops around at 2 ** 16
2. Optionally ensures no clashing with existing IDs
Note: if necessary, use the counter lock to keep the dont_clash_with object
in sync.
'''
MAX_ID = 2 ** 16
def __init__(self, *, initial_value=-1, dont_clash_with=None):
if dont_clash_with is None:
dont_clash_with = set()
self.value = initial_value
self.lock = threading.RLock()
self.dont_clash_with = dont_clash_with
def __call__(self):
'Get next ID, wrapping around at 2**16, ensuring no clashes'
with self.lock:
value = self.value + 1
while value in self.dont_clash_with or value >= self.MAX_ID:
value = 0 if value >= self.MAX_ID else value + 1
self.value = value
return self.value
if sys.platform == 'win32' or fcntl is None:
def socket_bytes_available(sock, *, default_buffer_size=4096, # noqa
available_buffer=None):
# No support for fcntl/termios on win32
return default_buffer_size
else:
def socket_bytes_available(sock, *, default_buffer_size=4096,
available_buffer=None):
'''Return bytes available to receive on socket
Parameters
----------
sock : socket.socket
default_buffer_size : int, optional
Default recv buffer size, should the platform not support the call or
the call fails for unknown reasons
available_buffer : array.array, optional
Array used for call to fcntl; can be specified to avoid reallocating
many times
'''
if available_buffer is None:
available_buffer = array.array('i', [0])
ok = fcntl.ioctl(sock, termios.FIONREAD, available_buffer) >= 0
return (max((available_buffer[0], default_buffer_size))
if ok else default_buffer_size)
@contextmanager
def named_temporary_file(*args, delete=True, **kwargs):
'''NamedTemporaryFile wrapper that works around issues in windows'''
# See: https://bugs.python.org/issue14243
from tempfile import NamedTemporaryFile
with NamedTemporaryFile(*args, delete=False, **kwargs) as f:
try:
yield f
finally:
if delete:
os.unlink(f.name)
class ShowVersionAction(argparse.Action):
# a special action that allows the usage --version to override
# any 'required args' requirements, the same way that --help does
def __init__(self,
option_strings,
dest=argparse.SUPPRESS,
default=argparse.SUPPRESS,
help=None):
super().__init__(
option_strings=option_strings,
dest=dest,
default=default,
nargs=0,
help=help)
def __call__(self, parser, namespace, values, option_string=None):
print(__version__)
parser.exit()
@functools.lru_cache(maxsize=128)
def safe_getsockname(sock):
"""
Call sock.getsockname() and, on Windows, return ('0.0.0.0', 0) if an error is raised.
This is a workaround to a critical issue affecting Windows. A better
solution should be found but requires more discussion. See
https://github.com/caproto/caproto/issues/514
and issues/PRs linked from there.
"""
try:
return sock.getsockname()
except Exception:
if sys.platform != 'win32':
raise
# This is what the Linux OS returns for a unconnected socket that has
# not yet been sent from.
return ('0.0.0.0', 0)
def adapt_old_callback_signature(func, weakref_set):
"""
If func has signature func(response), wrap in signature func(sub, response)
Parameters
----------
func: callable
Expected signature ``f(response)`` or ``f(sub, response)``
weakref_set: set
Will be used to store state.
Returns
-------
func: callable
Signature ``f(sub, response)``
"""
# Handle func with signature func(respons) for back-compat.
sig = inspect.signature(func)
try:
# Does this function accept two positional arguments?
sig.bind(None, None)
except TypeError:
warn(
"The signature of a subscription callback is now expected to "
"be func(sub, response). The signature func(response) is "
"supported, but support will be removed in a future release "
"of caproto.")
raw_func = func
raw_func_weakref = weakref.ref(raw_func)
def func(sub, response):
# Avoid closing over raw_func itself here or it will never be
# garbage collected.
raw_func = raw_func_weakref()
if raw_func is not None:
# Do nothing with sub because the user-provided func cannot
# accept it.
raw_func(response)
# Ensure func does not get garbage collected until raw_func does.
def called_when_raw_func_is_released(w):
# The point of this function is to hold one hard ref to func
# until raw_func is garbage collected.
func
# Clean up after ourselves.
weakref_set.remove(w)
w = weakref.ref(raw_func, called_when_raw_func_is_released)
# Hold a hard reference to w. Its callback removes it from this set.
weakref_set.add(w)
return func