pytrms 0.9.2__py3-none-any.whl → 0.9.5__py3-none-any.whl

pytrms/clients/mqtt.py

@@ -1,797 +1,820 @@
-import os
-import time
-import json
-import queue
-from collections import deque, namedtuple
-from datetime import datetime
-from functools import wraps
-from itertools import cycle, chain, zip_longest
-from threading import Condition, RLock
-
-from . import _logging
-from . import _par_id_file
-from .._base import itype, MqttClientBase
-
-
-log = _logging.getLogger(__name__)
-
-__all__ = ['MqttClient', 'MqttClientBase']
-
-
-with open(_par_id_file) as f:
-    from pandas import read_csv, isna
-
-    _par_id_info = read_csv(f, sep='\t').drop(0).set_index('Name')
-    if isna(_par_id_info.at['MPV_1', 'Access']):
-        log.warning(f'filling in read-properties still missing in {os.path.basename(_par_id_file)}')
-        _par_id_info.at['MPV_1', 'Access'] = 'RW'
-        _par_id_info.at['MPV_2', 'Access'] = 'RW'
-        _par_id_info.at['MPV_3', 'Access'] = 'RW'
-
-
-
-## >>>>>>>>    adaptor functions    <<<<<<<< ##
-
-def _build_header():
-    ts = datetime.now()
-    header = {
-        "TimeStamp": {
-            "Str": ts.isoformat(),
-            "sec": ts.timestamp() + 2082844800,  # convert to LabVIEW time
-        },
-    }
-    return header
-
-def _build_data_element(value, unit="-"):
-    elm = {
-        "Datatype": "",
-        "Index": -1,
-        "Value": str(value),
-        "Unit": str(unit),
-    }
-    if isinstance(value, bool):
-        # Note: True is also instance of int! Therefore, we must check it first:
-        elm.update({"Datatype": "BOOL", "Value": str(value).lower()})
-    elif isinstance(value, str):
-        elm.update({"Datatype": "STRING"})
-    elif isinstance(value, int):
-        elm.update({"Datatype": "I32"})
-    elif isinstance(value, float):
-        elm.update({"Datatype": "DBL"})
-    else:
-        raise NotImplemented("unknown datatype")
-
-    return elm
-
-def _build_write_command(parID, value, future_cycle=None):
-    cmd = {
-        "ParaID": str(parID),
-        "Value": str(value),
-        "Datatype": "",
-        "CMDMode": "Set",
-        "Index": -1,
-    }
-    if future_cycle is not None:
-        cmd.update({
-            "SchedMode": "OverallCycle",
-            "Schedule": str(future_cycle),
-        })
-    if isinstance(value, bool):
-        # Note: True is also instance of int!
-        cmd.update({"Datatype": "BOOL", "Value": str(value).lower()})
-    elif isinstance(value, str):
-        cmd.update({"Datatype": "STRING"})
-    elif isinstance(value, int):
-        cmd.update({"Datatype": "I32"})
-    elif isinstance(value, float):
-        cmd.update({"Datatype": "DBL"})
-    else:
-        raise NotImplemented("unknown datatype")
-
-    return cmd
-
-
-## >>>>>>>>    parsing functions    <<<<<<<< ##
-
-class ParsingError(Exception):
-    pass
-
-
-def _parse_data_element(elm):
-    '''
-    raises: ParsingError, KeyError
-    '''
-    # make a Python object of a DataElement
-    if elm["Datatype"] == "BOOL":
-        return bool(elm["Value"])
-    elif elm["Datatype"] == "DBL":
-        return float(elm["Value"])
-    elif elm["Datatype"] == "SGL":
-        return float(elm["Value"])
-    elif elm["Datatype"] == "I32":
-        return int(elm["Value"])
-    elif elm["Datatype"] == "I16":
-        return int(elm["Value"])
-    elif elm["Datatype"] == "STRING":
-        return str(elm["Value"])
-    raise ParsingError("unknown datatype: " + str(elm["Datatype"]))
-
-def _parse_fullcycle(byte_string, need_add_data=False):
-    '''Parses 'timecycle', 'intensity', 'mass_cal' and 'add_data' from bytes.
-
-    Important: the byteorder of the parsed arrays will be big-endian! This
-     may be aligned if needed with the `.byteswap()`-method on the array,
-     but is not automatically performed to avoid any extra copy.
-
-    @params
-    - need_add_data if `False`, the 'mass_cal' and 'add_data' returned will be None
-
-    Parsing the AddData-cluster is much slower than parsing the intensity-array!
-     This may be skipped to improve performance, but is necessary for loading
-     the 'mass_cal' anyway. For orientation:
-
-    performance (on a Intel Core i5, 8th Gen Ubuntu Linux):
-      < 2 ms  when `need_add_data=False` (default)
-      6-7 ms  when needing to parse the AddData-cluster (else)
-    
-    @returns a namedtuple ('timecycle', 'intensity', 'mass_cal', 'add_data')
-    '''
-    import numpy as np
-
-    _f32 = np.dtype(np.float32).newbyteorder('>')
-    _f64 = np.dtype(np.float64).newbyteorder('>')
-    _i16 = np.dtype(np.int16).newbyteorder('>')
-    _i32 = np.dtype(np.int32).newbyteorder('>')
-    _i64 = np.dtype(np.int64).newbyteorder('>')
-    _chr = np.dtype(np.int8).newbyteorder('>')
-
-    offset = 0
-
-    def rd_single(dtype=_i32):
-        nonlocal offset
-        _arr = np.frombuffer(byte_string, dtype=dtype, count=1, offset=offset)
-        offset += _arr.nbytes
-        return _arr[0]
-    
-    def rd_arr1d(dtype=_f32, count=None):
-        nonlocal offset
-        if count is None:
-            count = rd_single()
-        arr = np.frombuffer(byte_string, dtype=dtype, count=count, offset=offset)
-        offset += arr.nbytes
-        return arr
-    
-    def rd_arr2d(dtype=_f32):
-        nonlocal offset
-        n = rd_single()
-        m = rd_single()
-        arr = np.frombuffer(byte_string, dtype=dtype, count=n*m, offset=offset)
-        offset += arr.nbytes
-        return arr.reshape((n, m))
-
-    def rd_string():
-        nonlocal offset
-        return rd_arr1d(dtype=_chr).tobytes().decode('latin-1').lstrip('\x00')
-    
-    tc_cluster      = rd_arr1d(dtype=_f64, count=4)
-    run__, cpx__    = rd_arr1d(dtype=_f64, count=2)  # (discarded)
-    # SpecData #
-    intensity       = rd_arr1d(dtype=_f32)
-    sum_inty        = rd_arr1d(dtype=_f32)  # (discarded)
-    mon_peaks       = rd_arr2d(dtype=_f32)  # (discarded)
-    
-    if not need_add_data:
-        # skip costly parsing of Trace- and Add-Data cluster:
-        return itype.fullcycle_t(itype.timecycle_t(*tc_cluster), intensity, None, None)
-
-    # TraceData #  (as yet discarded)
-    tc_cluster2     = rd_arr1d(dtype=_f64, count=6)
-    twoD_raw        = rd_arr2d(dtype=_f32)
-    sum_raw         = rd_arr1d(dtype=_f32)
-    sum_corr        = rd_arr1d(dtype=_f32)
-    sum_conz        = rd_arr1d(dtype=_f32)
-    calc_traces     = rd_arr1d(dtype=_f32)
-    n_calc_trcs     = rd_single()
-    for i in range(n_calc_trcs):
-        calc_names  = rd_arr1d(dtype=_chr)
-    peak_centrs     = rd_arr1d(dtype=_f32)
-    # AddData #
-    add_data = dict()
-    n_add_data      = rd_single()
-    for i in range(n_add_data):
-        grp_name    = rd_string()
-        descr = []
-        for i in range(rd_single()):
-            descr.append(rd_string())
-        units = []
-        for i in range(rd_single()):
-            units.append(rd_string())
-        data        = rd_arr1d(dtype=_f32)
-        view        = rd_arr1d(dtype=_chr)
-        n_lv_times  = rd_single()
-        offset += 16 * n_lv_times  # skipping LabVIEW timestamp
-        add_data[grp_name] = [itype.add_data_item_t(*tup) for tup in zip_longest(data, descr, units, view)]
-
-    # MassCal #
-    mc_masses       = rd_arr1d(dtype=_f64)
-    mc_tbins        = rd_arr1d(dtype=_f64)
-    cal_paras       = rd_arr1d(dtype=_f64)
-    segmnt_cal_pars = rd_arr2d(dtype=_f64)
-    mcal_mode       = rd_single(dtype=_i16)
-    mass_cal = itype.masscal_t(mcal_mode, mc_masses, mc_tbins, cal_paras, segmnt_cal_pars)
-
-    return itype.fullcycle_t(itype.timecycle_t(*tc_cluster), intensity, mass_cal, add_data)
-
-
-class CalcConzInfo:
-
-    def __init__(self):
-        self.tables = {
-            "primary_ions": list(),
-            "transmission": list(),
-        }
-
-    @staticmethod
-    def load_json(json_string):
-        cc = CalcConzInfo()
-        j = json.loads(json_string)
-        delm = j["DataElement"]
-        for li in delm["Value"]["PISets"]["PiSets"]:
-            if not li["PriIonSetName"]:
-                log.info(f'loaded ({len(cc.tables["primary_ions"])}) primary-ion settings')
-                break
-
-            masses = map(float, filter(lambda x: x > 0, li["PriIonSetMasses"]))
-            values = map(float, li["PriIonSetMultiplier"])
-            cc.tables["primary_ions"].append(itype.table_setting_t(str(li["PriIonSetName"]), list(zip(masses, values))))
-
-        for li in j["DataElement"]["Value"]["TransSets"]["Transsets"]:
-            if not li["Name"]:
-                log.info(f'loaded ({len(cc.tables["transmission"])}) transmission settings')
-                break
-
-            masses = map(float, filter(lambda x: x > 0, li["Mass"]))
-            values = map(float, li["Value"])
-            # float(li["Voltage"])  # (not used)
-            cc.tables["transmission"].append(itype.table_setting_t(str(li["Name"]), list(zip(masses, values))))
-
-        return cc
-
-
-## >>>>>>>>    callback functions    <<<<<<<< ##
-
-def follow_calc_conz_info(client, self, msg):
-    if not msg.payload:
-        # empty payload will clear a retained topic
-        self._calcconzinfo = MqttClient._calcconzinfo
-        return
-
-    if not self._calcconzinfo[0] is _NOT_INIT:
-        # nothing to do..
-        return
-
-    log.debug(f"updating tm-/pi-table from {msg.topic}...")
-    self._calcconzinfo.append(CalcConzInfo.load_json(msg.payload.decode('latin-1')))
-
-follow_calc_conz_info.topics = ["PTR/Act/PTR_CalcConzInfo"]
-
-def follow_schedule(client, self, msg):
-    with follow_schedule._lock:
-        if msg.topic.endswith("SRV_ScheduleClear"):
-            self._sched_cmds.clear()
-            return
-
-        if msg.topic.endswith("SRV_Schedule"):
-            if not msg.payload:
-                log.warn("empty ACQ_SRV_Schedule payload has cleared retained topic")
-                self._sched_cmds.clear()
-                return
-
-            if msg.retain:
-                # Note: we either have received a message that has been
-                #  retained because of a new connection..
-                payload = json.loads(msg.payload.decode())
-                self._sched_cmds.clear()
-                self._sched_cmds.extend(payload["CMDs"])
-            else:
-                #  ..or the schedule as maintained by IoniTOF has changed,
-                #  which we handle ourselves below:
-                pass
-
-        if msg.topic.startswith("IC_Command"):
-            if not msg.payload:
-                log.error("empty IC_Command! has topic been cleared?")
-                return
-
-            # these are the freshly added scheduling requests:
-            payload = json.loads(msg.payload.decode())
-            self._sched_cmds.extend(payload["CMDs"])
-
-follow_schedule.topics = [
-    "DataCollection/Act/ACQ_SRV_Schedule",
-    "DataCollection/Set/ACQ_SRV_ScheduleClear",
-    "IC_Command/Write/Scheduled"
-]
-follow_schedule._lock = RLock()
-
-def follow_state(client, self, msg):
-    if not msg.payload:
-        # empty payload will clear a retained topic
-        self._server_state = MqttClient._server_state
-        return
-
-    payload = json.loads(msg.payload.decode())
-    state = payload["DataElement"]["Value"]
-    log.debug(f"[{self}] new server-state: " + str(state))
-    # replace the current state with the new element:
-    self._server_state.append(state)
-    meas_running = (state == "ACQ_Aquire")  # yes, there's a typo, plz keep it :)
-    just_started = (meas_running and not msg.retain)
-    if meas_running:
-        # signal the relevant thread(s) that we need an update:
-        self._calcconzinfo.append(_NOT_INIT)
-    if just_started:
-        # invalidate the source-file until we get a new one:
-        self._sf_filename.append(_NOT_INIT)
-
-follow_state.topics = ["DataCollection/Act/ACQ_SRV_CurrentState"]
-
-def follow_sourcefile(client, self, msg):
-    if not msg.payload:
-        # empty payload will clear a retained topic
-        self._sf_filename = MqttClient._sf_filename
-        return
-
-    payload = json.loads(msg.payload.decode())
-    path = payload["DataElement"]["Value"]
-    log.debug(f"[{self}] new source-file: " + str(path))
-    # replace the current path with the new element:
-    self._sf_filename.append(path)
-
-follow_sourcefile.topics = ["DataCollection/Act/ACQ_SRV_SetFullStorageFile"]
-
-def follow_act_set_values(client, self, msg):
-    if not msg.payload:
-        # empty payload will clear a retained topic
-        return
-
-    try:
-        server, kind, parID = msg.topic.split('/')
-        if server == "DataCollection":
-            # Note: this topic doesn't strictly follow the convention and is handled separately
-            return
-
-        if server == "Sequencer":
-            # Note: this is a separate program and will be ignored (has its own AUTO_-numbers et.c.)
-            return
-
-        if parID == "PTR_CalcConzInfo":
-            # another "special" topic handled in 'follow_calc_conz_info' ...
-            return
-
-        if parID not in _par_id_info.index:
-            log.warning(f"unknown par-ID in [{msg.topic}]")
-            return
-
-        payload = json.loads(msg.payload.decode())
-        if kind == "Act":
-            self.act_values[parID] = _parse_data_element(payload["DataElement"])
-        if kind == "Set":
-            self.set_values[parID] = _parse_data_element(payload["DataElement"])
-    except json.decoder.JSONDecodeError as exc:
-        log.error(f"{exc.__class__.__name__}: {exc} :: while processing [{msg.topic}] ({msg.payload})")
-        raise
-    except KeyError as exc:
-        log.error(f"{exc.__class__.__name__}: {exc} :: while processing [{msg.topic}] ({msg.payload})")
-        pass
-    except ParsingError as exc:
-        log.error(f"while parsing [{parID}] :: {str(exc)}")
-        pass
-
-follow_act_set_values.topics = ["+/Act/+", "+/Set/+"]
-
-def follow_cycle(client, self, msg):
-    if not msg.payload:
-        # empty payload will clear a retained topic
-        return
-
-    payload = json.loads(msg.payload.decode())
-    current = int(payload["DataElement"]["Value"])
-    # replace the current timecycle with the new element:
-    self._overallcycle.append(current)
-
-follow_cycle.topics = ["DataCollection/Act/ACQ_SRV_OverallCycle"]
-
-# collect all follow-functions together:
-_subscriber_functions = [fun for name, fun in list(vars().items())
-    if callable(fun) and name.startswith('follow_')]
-
-
-_NOT_INIT = object()
-
-
-class MqttClient(MqttClientBase):
-    """a simplified client for the Ionicon MQTT API.
-
-    > mq = MqttClient()
-    > mq.write('TCP_MCP_B', 3400)
-    ValueError()
-
-    """
-
-    _sched_cmds   = deque([_NOT_INIT], maxlen=None)
-    _server_state = deque([_NOT_INIT], maxlen=1)
-    _calcconzinfo = deque([_NOT_INIT], maxlen=1)
-    _sf_filename  = deque([""],        maxlen=1)
-    _overallcycle = deque([0],         maxlen=1)
-    act_values    = dict()
-    set_values    = dict()
-
-    set_value_limit = {
-        "TCP_MCP_B": 3200.0,
-    }
-    
-    @property
-    def is_connected(self):
-        '''Returns `True` if connection to IoniTOF could be established.'''
-        return (super().is_connected
-            and self._server_state[0] is not _NOT_INIT
-            and (len(self._sched_cmds) == 0 or self._sched_cmds[0] is not _NOT_INIT))
-
-    @property
-    def is_running(self):
-        '''Returns `True` if IoniTOF is currently acquiring data.'''
-        return self.current_server_state == 'ACQ_Aquire'  # yes, there's a typo, plz keep it :)
-
-    @property
-    def current_schedule(self):
-        '''Returns a list with the upcoming write commands in ascending order.'''
-        if not self.is_connected:
-            return []
-
-        current_cycle = self._overallcycle[0]
-        filter_fun = lambda cmd: float(cmd["Schedule"]) > current_cycle
-        sorted_fun = lambda cmd: float(cmd["Schedule"])
-
-        return sorted(filter(filter_fun, self._sched_cmds), key=sorted_fun)
-
-    @property
-    def current_server_state(self):
-        '''Returns the state of the acquisition-server. One of:
-
-        - "ACQ_Idle"
-        - "ACQ_JustStarted"
-        - "ACQ_Aquire"
-        - "ACQ_Stopping"
-
-        or "<unknown>" if there's no connection to IoniTOF.
-        '''
-        if self.is_connected:
-            return self._server_state[0]
-        return "<unknown>"
-
-    @property
-    def current_sourcefile(self):
-        '''Returns the path to the hdf5-file that is currently being written.
-        
-        Returns an empty string if no measurement is running.
-        '''
-        if not self.is_running:
-            return ""
-
-        if self._sf_filename[0] is not _NOT_INIT:
-            return self._sf_filename[0]
-
-        # Note: '_NOT_INIT' is set by us on start of acquisition, so we'd expect
-        #  to receive the source-file-topic after a (generous) timeout:
-        timeout_s = 15
-        started_at = time.monotonic()
-        while time.monotonic() < started_at + timeout_s:
-            if self._sf_filename[0] is not _NOT_INIT:
-                return self._sf_filename[0]
-    
-            time.sleep(10e-3)
-        else:
-            raise TimeoutError(f"[{self}] unable to retrieve source-file after ({timeout_s = })");
-
-    @property
-    def current_cycle(self):
-        '''Returns the current 'AbsCycle' (/'OverallCycle').'''
-        if self.is_running:
-            return self._overallcycle[0]
-        return 0
-
-    def __init__(self, host='127.0.0.1', port=1883):
-        # this sets up the mqtt connection with default callbacks:
-        super().__init__(host, port, _subscriber_functions, None, None, None)
-        log.debug(f"connection check ({self.is_connected}) :: {self._server_state = } / {self._sched_cmds = }");
-
-    def disconnect(self):
-        super().disconnect()
-        log.debug(f"[{self}] has disconnected")
-        # reset internal queues to their defaults:
-        self._sched_cmds   = MqttClient._sched_cmds
-        self._server_state = MqttClient._server_state
-        self._calcconzinfo = MqttClient._calcconzinfo
-        self._sf_filename  = MqttClient._sf_filename
-        self._overallcycle = MqttClient._overallcycle
-        self.act_values    = MqttClient.act_values
-        self.set_values    = MqttClient.set_values
-
-    def get(self, parID, kind="set"):
-        '''Return the last known value for the given `parID`.
-
-        - kind: one of 'set'/'act' (default: 'set')
-
-        A `KeyError` will be raised if the given `parID` is unknown!
-        '''
-        if not self.is_connected:
-            raise Exception(f"[{self}] no connection to instrument");
-
-        _lut = self.act_values if kind.lower() == "act" else self.set_values
-        is_read_only = ('W' not in _par_id_info.loc[parID].Access)  # may raise KeyError!
-        if _lut is self.set_values and is_read_only:
-            raise ValueError(f"'{parID}' is read-only, did you mean `kind='act'`?")
-
-        if not parID in _lut:
-            # Note: The values should need NO! time to be populated from the MQTT topics,
-            #  because all topics are published as *retained* by the PTR-server.
-            #  However, a short timeout is respected before raising a `KeyError`:
-            time.sleep(200e-3)
-            rv = _lut.get(parID)
-            if rv is not None:
-                return rv
-
-            # still not found? give some useful hints for the user not to go crazy:
-            error_hint = (
-                "act" if parID in self.act_values else
-                "set" if parID in self.set_values else
-                "")
-            raise KeyError(str(parID) + (' (did you mean `kind="%s"`?)' % error_hint) if error_hint else "")
-        return _lut[parID]
-
-    def get_table(self, table_name):
-        timeout_s = 10
-        started_at = time.monotonic()
-        try:
-            while time.monotonic() < started_at + timeout_s:
-                # confirm change of state:
-                if not self._calcconzinfo[0] is _NOT_INIT:
-                    return self._calcconzinfo[0].tables[table_name]
-    
-                time.sleep(10e-3)
-            else:
-                raise TimeoutError(f"[{self}] unable to retrieve calc-conz-info from PTR server");
-        except KeyError as exc:
-            raise KeyError(str(exc) + f", possible values: {list(CalcConzInfo.tables.keys())}")
-
-    def set(self, parID, new_value, unit='-'):
-        '''Set a 'new_value' to 'parID' in the DataCollection.'''
-        if not self.is_connected:
-            raise Exception(f"[{self}] no connection to instrument");
-
-        raise NotImplementedError("DataCollection/Set, did you mean .write(parID)?")
-
-        topic, qos, retain = "DataCollection/Set/" + str(parID), 1, True
-        log.info(f"setting '{parID}' ~> [{new_value}]")
-        payload = {
-            "Header":      _build_header(),
-            "DataElement": _build_data_element(new_value, unit),
-        }
-        return self.publish_with_ack(topic, json.dumps(payload), qos=qos, retain=retain)
-
-    def filter_schedule(self, parID):
-        '''Returns a list with the upcoming write commands for 'parID' in ascending order.'''
-        return (cmd for cmd in self.current_schedule if cmd["ParaID"] == str(parID))
-
-    def write(self, parID, new_value):
-        '''Write a 'new_value' to 'parID' directly.'''
-        if not self.is_connected:
-            raise Exception(f"[{self}] no connection to instrument");
-
-        if not 'W' in _par_id_info.loc[parID].Access:  # may raise KeyError!
-            raise ValueError(f"'{parID}' is read-only")
-
-        if parID in __class__.set_value_limit and new_value > __class__.set_value_limit[parID]:
-            raise ValueError("set value limit of {__class__.set_value_limit[parID]} on '{parID}'")
-
-        topic, qos, retain = "IC_Command/Write/Direct", 1, False
-        log.info(f"writing '{parID}' ~> [{new_value}]")
-        cmd = _build_write_command(parID, new_value)
-        payload = {
-            "Header": _build_header(),
-            "CMDs": [ cmd, ]
-        }
-        return self.publish_with_ack(topic, json.dumps(payload), qos=qos, retain=retain)
-
-    def schedule(self, parID, new_value, future_cycle):
-        '''Schedule a 'new_value' to 'parID' for the given 'future_cycle'.
-
-        If 'future_cycle' is in fact in the past, the behaviour is defined by IoniTOF
-        (most likely the command is ignored). To be sure, the '.current_cycle' should
-        be checked before and after running the '.schedule' command programmatically!
-        '''
-        if not self.is_connected:
-            raise Exception(f"[{self}] no connection to instrument");
-
-        if not 'W' in _par_id_info.loc[parID].Access:  # may raise KeyError!
-            raise ValueError(f"'{parID}' is read-only")
-
-        if parID in __class__.set_value_limit and new_value > __class__.set_value_limit[parID]:
-            raise ValueError("set value limit of {__class__.set_value_limit[parID]} on '{parID}'")
-
-        if (future_cycle == 0 and not self.is_running):
-            # Note: ioniTOF40 doesn't handle scheduling for the 0th cycle!
-            if parID == "AME_ActionNumber":
-                # a) the action-number will trigger a script for the 0th cycle, so
-                #    we *must* be scheduling it!
-                self.write("AME_ActionNumber", new_value)
-            elif parID.startswith("AME_"):
-                # b) the AME-numbers cannot (currently) be set (i.e. written), but since
-                #    they are inserted just *before* the cycle, this will work just fine:
-                future_cycle = 1
-            else:
-                # c) in all other cases, let's assume the measurement will start soon
-                #    and dare to write immediately, skipping the schedule altogether:
-                log.debug(f"immediately writing {parID = } @ cycle '0' (measurement stopped)")
-                return self.write(parID, new_value)
-
-        if not future_cycle > self.current_cycle:
-            log.warn(f"attempting to schedule past cycle, hope you know what you're doing");
-            pass  # and at least let's debug it in MQTT browser (see also doc-string above)!
-
-        topic, qos, retain = "IC_Command/Write/Scheduled", 1, False
-        log.info(f"scheduling '{parID}' ~> [{new_value}] for cycle ({future_cycle})")
-        cmd = _build_write_command(parID, new_value, future_cycle)
-        payload = {
-            "Header": _build_header(),
-            "CMDs": [ cmd, ]
-        }
-        return self.publish_with_ack(topic, json.dumps(payload), qos=qos, retain=retain)
-
-    def schedule_filename(self, path, future_cycle):
-        '''Start writing to a new .h5 file with the beginning of 'future_cycle'.'''
-        assert str(path), "filename cannot be empty!"
-        # try to make sure that IoniTOF accepts the path:
-        if self.host == '127.0.0.1':
-            os.makedirs(os.path.dirname(path), exist_ok=True)
-            try:
-                with open(path, 'x'):
-                    log.info("touched new file:", path)
-            except FileExistsError as exc:
-                log.error(f"new filename '{path}' already exists and will not be scheduled!")
-                return
-
-        return self.schedule('ACQ_SRV_SetFullStorageFile', path.replace('/', '\\'), future_cycle)
-
-    def start_measurement(self, path=None):
-        '''Start a new measurement and block until the change is confirmed.
-
-        If 'path' is not None, write to the given .h5 file.
-        '''
-        if not path:
-            self.write('ACQ_SRV_Start_Meas_Quick', True)
-        else:
-            self.write('ACQ_SRV_Start_Meas_Record', path.replace('/', '\\'))
-        timeout_s = 30
-        started_at = time.monotonic()
-        while time.monotonic() < started_at + timeout_s:
-            if self.is_running:
-                break
-
-            time.sleep(10e-3)
-        else:
-            self.disconnect()
-            raise TimeoutError(f"[{self}] error starting measurement");
-
-    def stop_measurement(self, future_cycle=None):
-        '''Stop the current measurement and block until the change is confirmed.
-
-        If 'future_cycle' is not None and in the future, schedule the stop command.'''
-        if future_cycle is None or not future_cycle > self._overallcycle[0]:
-            self.write('ACQ_SRV_Stop_Meas', True)
-        else:
-            self.schedule('ACQ_SRV_Stop_Meas', True, future_cycle)
-        # may need to wait until the scheduled event..
-        if future_cycle is not None:
-            self.block_until(future_cycle)
-        # ..for this timeout to be applicable:
-        timeout_s = 30
-        started_at = time.monotonic()
-        while time.monotonic() < started_at + timeout_s:
-            # confirm change of state:
-            if not self.is_running:
-                break
-
-            time.sleep(10e-3)
-        else:
-            self.disconnect()
-            raise TimeoutError(f"[{self}] error stopping measurement");
-
-    def block_until(self, cycle):
-        '''Blocks the current thread until at least 'cycle' has passed or acquisition stopped.
-
-        Returns the actual current cycle.
-        '''
-        while self.is_running:
-            if self._overallcycle[0] >= int(cycle):
-                break
-            time.sleep(10e-3)
-        else:
-            return 0
-
-        return self._overallcycle[0]
-
-    def iter_specdata(self, timeout_s=None, buffer_size=300):
-        '''Returns an iterator over the fullcycle-data as long as it is available.
-
-        * This will wait up to `timeout_s` (or indefinitely if `None`) for a
-          measurement to start or raise a TimeoutError (default: None).
-        * Elements will be buffered up to a maximum of `buffer_size` cycles (default: 300).
-        * Cycles recorded prior to calling `next()` on the iterator may be missed,
-          so ideally this should be set up before any measurement is running.
-        * [Important]: When the buffer runs full, a `queue.Full` exception will be raised!
-          Therefore, the caller should consume the iterator as soon as possible while the
-          measurement is running.
-        '''
-        q = queue.Queue(buffer_size)
-        topic = "DataCollection/Act/ACQ_SRV_FullCycleData"
-        qos = 2
-
-        def callback(client, self, msg):
-            try:
-                q.put_nowait(_parse_fullcycle(msg.payload, need_add_data=True))
-                log.debug(f"received fullcycle, buffer at ({q.qsize()}/{q.maxsize})")
-            except queue.Full:
-                # DO NOT FAIL INSIDE THE CALLBACK!
-                log.error(f"iter_specdata({q.maxsize}): fullcycle buffer overrun!")
-                client.unsubscribe(topic)
-
-        if not self.is_connected:
-            raise Exception("no connection to MQTT broker")
-
-        # Note: when using a simple generator function like this, the following lines
-        #  will not be excecuted until the first call to `next` on the iterator!
-        #  this means, the callback will not yet be executed, the queue not filled 
-        #  and we might miss the first cycles...
-        self.client.message_callback_add(topic, callback)
-        self.client.subscribe(topic, qos)
-        try:
-            # Note: Prior to 3.0 on POSIX systems, and for *all versions on Windows*,
-            #  if block is true and timeout is None, [the q.get()] operation goes into an
-            #  uninterruptible wait on an underlying lock. This means that no exceptions
-            #  can occur, and in particular a SIGINT will not trigger a KeyboardInterrupt!
-            yield q.get(block=True, timeout=timeout_s)  # waiting for measurement to run...
-
-            while self.is_running or not q.empty():
-                if q.full():
-                    # re-raise what we swallowed in the callback..
-                    raise queue.Full
-
-                if not self.is_connected:
-                    # no more data will come, so better prevent a deadlock:
-                    break
-
-                try:
-                    yield q.get(block=True, timeout=1.0)  # seconds
-                except queue.Empty:
-                    continue
-
-        except queue.Empty:
-            assert timeout_s is not None, "this should never happen"
-            raise TimeoutError("no measurement running after {timeout_s} seconds")
-
-        finally:
-            #  ...also, when using more than one iterator, the first to finish will
-            #  unsubscribe and cause all others to stop maybe before the time!
-            #  all of this might not actually be an issue right now, but
-            # TODO :: fix this weird behaviour (can only be done by implementing the
-            #  iterator-protocol properly using a helper class)!
-            self.client.unsubscribe(topic)
-            self.client.message_callback_remove(topic)
-
-    iter_specdata.__doc__ += _parse_fullcycle.__doc__
-
-    def __repr__(self):
-        return f"<{self.__class__.__name__}[{self.host}]>"
-
+import os
+import time
+import json
+import queue
+from collections import deque, namedtuple
+from datetime import datetime
+from functools import wraps
+from itertools import cycle, chain, zip_longest
+from threading import Condition, RLock
+
+from . import _logging
+from . import _par_id_file
+from .._base import itype, MqttClientBase
+
+
+log = _logging.getLogger(__name__)
+
+__all__ = ['MqttClient', 'MqttClientBase']
+
+
+with open(_par_id_file) as f:
+    from pandas import read_csv, isna
+
+    _par_id_info = read_csv(f, sep='\t').drop(0).set_index('Name')
+    if isna(_par_id_info.at['MPV_1', 'Access']):
+        log.warning(f'filling in read-properties still missing in {os.path.basename(_par_id_file)}')
+        _par_id_info.at['MPV_1', 'Access'] = 'RW'
+        _par_id_info.at['MPV_2', 'Access'] = 'RW'
+        _par_id_info.at['MPV_3', 'Access'] = 'RW'
+
+
+
+## >>>>>>>>    adaptor functions    <<<<<<<< ##
+
+def _build_header():
+    ts = datetime.now()
+    header = {
+        "TimeStamp": {
+            "Str": ts.isoformat(),
+            "sec": ts.timestamp() + 2082844800,  # convert to LabVIEW time
+        },
+    }
+    return header
+
+def _build_data_element(value, unit="-"):
+    elm = {
+        "Datatype": "",
+        "Index": -1,
+        "Value": str(value),
+        "Unit": str(unit),
+    }
+    if isinstance(value, bool):
+        # Note: True is also instance of int! Therefore, we must check it first:
+        elm.update({"Datatype": "BOOL", "Value": str(value).lower()})
+    elif isinstance(value, str):
+        elm.update({"Datatype": "STRING"})
+    elif isinstance(value, int):
+        elm.update({"Datatype": "I32"})
+    elif isinstance(value, float):
+        elm.update({"Datatype": "DBL"})
+    else:
+        raise NotImplemented("unknown datatype")
+
+    return elm
+
+def _build_write_command(parID, value, future_cycle=None):
+    cmd = {
+        "ParaID": str(parID),
+        "Value": str(value),
+        "Datatype": "",
+        "CMDMode": "Set",
+        "Index": -1,
+    }
+    if future_cycle is not None:
+        cmd.update({
+            "SchedMode": "OverallCycle",
+            "Schedule": str(future_cycle),
+        })
+    if isinstance(value, bool):
+        # Note: True is also instance of int!
+        cmd.update({"Datatype": "BOOL", "Value": str(value).lower()})
+    elif isinstance(value, str):
+        cmd.update({"Datatype": "STRING"})
+    elif isinstance(value, int):
+        cmd.update({"Datatype": "I32"})
+    elif isinstance(value, float):
+        cmd.update({"Datatype": "DBL"})
+    else:
+        raise NotImplemented("unknown datatype")
+
+    return cmd
+
+
+## >>>>>>>>    parsing functions    <<<<<<<< ##
+
+class ParsingError(Exception):
+    pass
+
+
+def _parse_data_element(elm):
+    '''
+    raises: ParsingError, KeyError
+    '''
+    # make a Python object of a DataElement
+    if elm["Datatype"] == "BOOL":
+        return bool(elm["Value"])
+    elif elm["Datatype"] == "DBL":
+        return float(elm["Value"])
+    elif elm["Datatype"] == "SGL":
+        return float(elm["Value"])
+    elif elm["Datatype"] == "I32":
+        return int(elm["Value"])
+    elif elm["Datatype"] == "I16":
+        return int(elm["Value"])
+    elif elm["Datatype"] == "STRING":
+        return str(elm["Value"])
+    raise ParsingError("unknown datatype: " + str(elm["Datatype"]))
+
+def _parse_fullcycle(byte_string, need_add_data=False):
+    '''Parses 'timecycle', 'intensity', 'mass_cal' and 'add_data' from bytes.
+
+    Important: the byteorder of the parsed arrays will be big-endian! This
+     may be aligned if needed with the `.byteswap()`-method on the array,
+     but is not automatically performed to avoid any extra copy.
+
+    @params
+    - need_add_data if `False`, the 'mass_cal' and 'add_data' returned will be None
+
+    Parsing the AddData-cluster is much slower than parsing the intensity-array!
+     This may be skipped to improve performance, but is necessary for loading
+     the 'mass_cal' anyway. For orientation:
+
+    performance (on a Intel Core i5, 8th Gen Ubuntu Linux):
+      < 2 ms  when `need_add_data=False` (default)
+      6-7 ms  when needing to parse the AddData-cluster (else)
+    
+    @returns a namedtuple ('timecycle', 'intensity', 'mass_cal', 'add_data')
+    '''
+    import numpy as np
+
+    _f32 = np.dtype(np.float32).newbyteorder('>')
+    _f64 = np.dtype(np.float64).newbyteorder('>')
+    _i16 = np.dtype(np.int16).newbyteorder('>')
+    _i32 = np.dtype(np.int32).newbyteorder('>')
+    _i64 = np.dtype(np.int64).newbyteorder('>')
+    _chr = np.dtype(np.int8).newbyteorder('>')
+
+    offset = 0
+
+    def rd_single(dtype=_i32):
+        nonlocal offset
+        _arr = np.frombuffer(byte_string, dtype=dtype, count=1, offset=offset)
+        offset += _arr.nbytes
+        return _arr[0]
+    
+    def rd_arr1d(dtype=_f32, count=None):
+        nonlocal offset
+        if count is None:
+            count = rd_single()
+        arr = np.frombuffer(byte_string, dtype=dtype, count=count, offset=offset)
+        offset += arr.nbytes
+        return arr
+    
+    def rd_arr2d(dtype=_f32):
+        nonlocal offset
+        n = rd_single()
+        m = rd_single()
+        arr = np.frombuffer(byte_string, dtype=dtype, count=n*m, offset=offset)
+        offset += arr.nbytes
+        return arr.reshape((n, m))
+
+    def rd_string():
+        nonlocal offset
+        return rd_arr1d(dtype=_chr).tobytes().decode('latin-1').lstrip('\x00')
+    
+    tc_cluster      = rd_arr1d(dtype=_f64, count=4)
+    run__, cpx__    = rd_arr1d(dtype=_f64, count=2)  # (discarded)
+    # SpecData #
+    intensity       = rd_arr1d(dtype=_f32)
+    sum_inty        = rd_arr1d(dtype=_f32)  # (discarded)
+    mon_peaks       = rd_arr2d(dtype=_f32)  # (discarded)
+    
+    if not need_add_data:
+        # skip costly parsing of Trace- and Add-Data cluster:
+        return itype.fullcycle_t(itype.timecycle_t(*tc_cluster), intensity, None, None)
+
+    # TraceData #  (as yet discarded)
+    tc_cluster2     = rd_arr1d(dtype=_f64, count=6)
+    twoD_raw        = rd_arr2d(dtype=_f32)
+    sum_raw         = rd_arr1d(dtype=_f32)
+    sum_corr        = rd_arr1d(dtype=_f32)
+    sum_conz        = rd_arr1d(dtype=_f32)
+    calc_traces     = rd_arr1d(dtype=_f32)
+    n_calc_trcs     = rd_single()
+    for i in range(n_calc_trcs):
+        calc_names  = rd_arr1d(dtype=_chr)
+    peak_centrs     = rd_arr1d(dtype=_f32)
+    # AddData #
+    add_data = dict()
+    n_add_data      = rd_single()
+    for i in range(n_add_data):
+        grp_name    = rd_string()
+        descr = []
+        for i in range(rd_single()):
+            descr.append(rd_string())
+        units = []
+        for i in range(rd_single()):
+            units.append(rd_string())
+        data        = rd_arr1d(dtype=_f32)
+        view        = rd_arr1d(dtype=_chr)
+        n_lv_times  = rd_single()
+        offset += 16 * n_lv_times  # skipping LabVIEW timestamp
+        add_data[grp_name] = [itype.add_data_item_t(*tup) for tup in zip_longest(data, descr, units, view)]
+
+    # MassCal #
+    mc_masses       = rd_arr1d(dtype=_f64)
+    mc_tbins        = rd_arr1d(dtype=_f64)
+    cal_paras       = rd_arr1d(dtype=_f64)
+    segmnt_cal_pars = rd_arr2d(dtype=_f64)
+    mcal_mode       = rd_single(dtype=_i16)
+    mass_cal = itype.masscal_t(mcal_mode, mc_masses, mc_tbins, cal_paras, segmnt_cal_pars)
+
+    return itype.fullcycle_t(itype.timecycle_t(*tc_cluster), intensity, mass_cal, add_data)
+
+
+class CalcConzInfo:
+
+    def __init__(self):
+        self.tables = {
+            "primary_ions": list(),
+            "transmission": list(),
+        }
+
+    @staticmethod
+    def load_json(json_string):
+        cc = CalcConzInfo()
+        j = json.loads(json_string)
+        delm = j["DataElement"]
+        for li in delm["Value"]["PISets"]["PiSets"]:
+            if not li["PriIonSetName"]:
+                log.info(f'loaded ({len(cc.tables["primary_ions"])}) primary-ion settings')
+                break
+
+            masses = map(float, filter(lambda x: x > 0, li["PriIonSetMasses"]))
+            values = map(float, li["PriIonSetMultiplier"])
+            cc.tables["primary_ions"].append(itype.table_setting_t(str(li["PriIonSetName"]), list(zip(masses, values))))
+
+        for li in j["DataElement"]["Value"]["TransSets"]["Transsets"]:
+            if not li["Name"]:
+                log.info(f'loaded ({len(cc.tables["transmission"])}) transmission settings')
+                break
+
+            masses = map(float, filter(lambda x: x > 0, li["Mass"]))
+            values = map(float, li["Value"])
+            # float(li["Voltage"])  # (not used)
+            cc.tables["transmission"].append(itype.table_setting_t(str(li["Name"]), list(zip(masses, values))))
+
+        return cc
+
+
+## >>>>>>>>    callback functions    <<<<<<<< ##
+
+def follow_calc_conz_info(client, self, msg):
+    if not msg.payload:
+        # empty payload will clear a retained topic
+        self._calcconzinfo = MqttClient._calcconzinfo
+        return
+
+    if not self._calcconzinfo[0] is _NOT_INIT:
+        # nothing to do..
+        return
+
+    log.debug(f"updating tm-/pi-table from {msg.topic}...")
+    self._calcconzinfo.append(CalcConzInfo.load_json(msg.payload.decode('latin-1')))
+
+follow_calc_conz_info.topics = ["PTR/Act/PTR_CalcConzInfo"]
+
+def follow_schedule(client, self, msg):
+    with follow_schedule._lock:
+        if msg.topic.endswith("SRV_ScheduleClear"):
+            self._sched_cmds.clear()
+            return
+
+        if msg.topic.endswith("SRV_Schedule"):
+            if not msg.payload:
+                log.warn("empty ACQ_SRV_Schedule payload has cleared retained topic")
+                self._sched_cmds.clear()
+                return
+
+            if msg.retain:
+                # Note: we either have received a message that has been
+                #  retained because of a new connection..
+                payload = json.loads(msg.payload.decode())
+                self._sched_cmds.clear()
+                self._sched_cmds.extend(payload["CMDs"])
+            else:
+                #  ..or the schedule as maintained by IoniTOF has changed,
+                #  which we handle ourselves below:
+                pass
+
+        if msg.topic.startswith("IC_Command"):
+            if not msg.payload:
+                log.error("empty IC_Command! has topic been cleared?")
+                return
+
+            # these are the freshly added scheduling requests:
+            payload = json.loads(msg.payload.decode())
+            self._sched_cmds.extend(payload["CMDs"])
+
+follow_schedule.topics = [
+    "DataCollection/Act/ACQ_SRV_Schedule",
+    "DataCollection/Set/ACQ_SRV_ScheduleClear",
+    "IC_Command/Write/Scheduled"
+]
+follow_schedule._lock = RLock()
+
+def follow_state(client, self, msg):
+    if not msg.payload:
+        # empty payload will clear a retained topic
+        self._server_state = MqttClient._server_state
+        return
+
+    payload = json.loads(msg.payload.decode())
+    state = payload["DataElement"]["Value"]
+    log.debug(f"[{self}] new server-state: " + str(state))
+    # replace the current state with the new element:
+    self._server_state.append(state)
+    meas_running = (state == "ACQ_Aquire")  # yes, there's a typo, plz keep it :)
+    if meas_running:
+        # signal the relevant thread(s) that we need an update:
+        self._calcconzinfo.append(_NOT_INIT)
+    if not meas_running:
+        # Note: the user-interface in `.current_sourcefile` checks for the
+        #  above _server_state and expects an initialized filename if and
+        #  only if the server is running! Therefore, we can safely
+        #  invalidate the source-file until we get a new one:
+        self._sf_filename.append(_NOT_INIT)
+
+follow_state.topics = ["DataCollection/Act/ACQ_SRV_CurrentState"]
+
+def follow_sourcefile(client, self, msg):
+    if not msg.payload:
+        # empty payload will clear a retained topic
+        self._sf_filename = MqttClient._sf_filename
+        return
+
+    payload = json.loads(msg.payload.decode())
+    path = payload["DataElement"]["Value"]
+    log.debug(f"[{self}] new source-file: " + str(path))
+    # replace the current path with the new element:
+    self._sf_filename.append(path)
+
+follow_sourcefile.topics = ["DataCollection/Act/ACQ_SRV_SetFullStorageFile"]
+
+def follow_act_set_values(client, self, msg):
+    if not msg.payload:
+        # empty payload will clear a retained topic
+        return
+
+    try:
+        server, kind, parID = msg.topic.split('/')
+        if server == "DataCollection":
+            # Note: this topic doesn't strictly follow the convention and is handled separately
+            if kind != "Set":
+                return
+
+        if server == "Sequencer":
+            # Note: this is a separate program and will be ignored (has its own AUTO_-numbers et.c.)
+            return
+
+        if parID == "PTR_CalcConzInfo":
+            # another "special" topic handled in 'follow_calc_conz_info' ...
+            return
+
+        if parID not in _par_id_info.index:
+            log.warning(f"unknown par-ID in [{msg.topic}]")
+            return
+
+        payload = json.loads(msg.payload.decode())
+        if kind == "Act":
+            self.act_values[parID] = _parse_data_element(payload["DataElement"])
+        if kind == "Set":
+            self.set_values[parID] = _parse_data_element(payload["DataElement"])
+    except json.decoder.JSONDecodeError as exc:
+        log.error(f"{exc.__class__.__name__}: {exc} :: while processing [{msg.topic}] ({msg.payload})")
+        raise
+    except KeyError as exc:
+        log.error(f"{exc.__class__.__name__}: {exc} :: while processing [{msg.topic}] ({msg.payload})")
+        pass
+    except ParsingError as exc:
+        log.error(f"while parsing [{parID}] :: {str(exc)}")
+        pass
+
+follow_act_set_values.topics = ["+/Act/+", "+/Set/+"]
+
+def follow_cycle(client, self, msg):
+    if not msg.payload:
+        # empty payload will clear a retained topic
+        return
+
+    payload = json.loads(msg.payload.decode())
+    current = int(payload["DataElement"]["Value"])
+    # replace the current timecycle with the new element:
+    self._overallcycle.append(current)
+
+follow_cycle.topics = ["DataCollection/Act/ACQ_SRV_OverallCycle"]
+
+# collect all follow-functions together:
+_subscriber_functions = [fun for name, fun in list(vars().items())
+    if callable(fun) and name.startswith('follow_')]
+
+
+_NOT_INIT = object()
+
+
+class MqttClient(MqttClientBase):
+    """a simplified client for the Ionicon MQTT API.
+
+    > mq = MqttClient()
+    > mq.write('TCP_MCP_B', 3400)
+    ValueError()
+
+    """
+
+    _sched_cmds   = deque([_NOT_INIT], maxlen=None)
+    _server_state = deque([_NOT_INIT], maxlen=1)
+    _calcconzinfo = deque([_NOT_INIT], maxlen=1)
+    _sf_filename  = deque([""],        maxlen=1)
+    _overallcycle = deque([0],         maxlen=1)
+    act_values    = dict()
+    set_values    = dict()
+
+    set_value_limit = {
+        "TCP_MCP_B": 3200.0,
+    }
+    
+    @property
+    def is_connected(self):
+        '''Returns `True` if connection to IoniTOF could be established.'''
+        return (super().is_connected
+            and self._server_state[0] is not _NOT_INIT
+            and (len(self._sched_cmds) == 0 or self._sched_cmds[0] is not _NOT_INIT))
+
+    @property
+    def is_running(self):
+        '''Returns `True` if IoniTOF is currently acquiring data.'''
+        return self.current_server_state == 'ACQ_Aquire'  # yes, there's a typo, plz keep it :)
+
+    @property
+    def current_schedule(self):
+        '''Returns a list with the upcoming write commands in ascending order.'''
+        if not self.is_connected:
+            return []
+
+        current_cycle = self._overallcycle[0]
+        filter_fun = lambda cmd: float(cmd["Schedule"]) > current_cycle
+        sorted_fun = lambda cmd: float(cmd["Schedule"])
+
+        return sorted(filter(filter_fun, self._sched_cmds), key=sorted_fun)
+
+    @property
+    def current_server_state(self):
+        '''Returns the state of the acquisition-server. One of:
+
+        - "ACQ_Idle"
+        - "ACQ_JustStarted"
+        - "ACQ_Aquire"
+        - "ACQ_Stopping"
+
+        or "<unknown>" if there's no connection to IoniTOF.
+        '''
+        if self.is_connected:
+            return self._server_state[0]
+        return "<unknown>"
+
+    @property
+    def current_sourcefile(self):
+        '''Returns the path to the hdf5-file that is currently being written.
+        
+        Returns an empty string if no measurement is running.
+        '''
+        if not self.is_running:
+            return ""
+
+        if self._sf_filename[0] is not _NOT_INIT:
+            return self._sf_filename[0]
+
+        # Note: '_NOT_INIT' is set by us on start of acquisition, so we'd expect
+        #  to receive the source-file-topic after a (generous) timeout:
+        timeout_s = 15
+        started_at = time.monotonic()
+        while time.monotonic() < started_at + timeout_s:
+            if self._sf_filename[0] is not _NOT_INIT:
+                return self._sf_filename[0]
+    
+            time.sleep(10e-3)
+        else:
+            raise TimeoutError(f"[{self}] unable to retrieve source-file after ({timeout_s = })");
+
+    @property
+    def current_cycle(self):
+        '''Returns the current 'AbsCycle' (/'OverallCycle').'''
+        if self.is_running:
+            return self._overallcycle[0]
+        return 0
+
+    def __init__(self, host='127.0.0.1', port=1883):
+        # this sets up the mqtt connection with default callbacks:
+        super().__init__(host, port, _subscriber_functions, None, None, None)
+        log.debug(f"connection check ({self.is_connected}) :: {self._server_state = } / {self._sched_cmds = }");
+
+    def disconnect(self):
+        super().disconnect()
+        log.debug(f"[{self}] has disconnected")
+        # reset internal queues to their defaults:
+        self._sched_cmds   = MqttClient._sched_cmds
+        self._server_state = MqttClient._server_state
+        self._calcconzinfo = MqttClient._calcconzinfo
+        self._sf_filename  = MqttClient._sf_filename
+        self._overallcycle = MqttClient._overallcycle
+        self.act_values    = MqttClient.act_values
+        self.set_values    = MqttClient.set_values
+
+    def get(self, parID, kind="set"):
+        '''Return the last known value for the given `parID`.
+
+        - kind: one of 'set'/'act' (default: 'set')
+
+        A `KeyError` will be raised if the given `parID` is unknown!
+        '''
+        if not self.is_connected:
+            raise Exception(f"[{self}] no connection to instrument");
+
+        _lut = self.act_values if kind.lower() == "act" else self.set_values
+        is_read_only = ('W' not in _par_id_info.loc[parID].Access)  # may raise KeyError!
+        if _lut is self.set_values and is_read_only:
+            raise ValueError(f"'{parID}' is read-only, did you mean `kind='act'`?")
+
+        if not parID in _lut:
+            # Note: The values should need NO! time to be populated from the MQTT topics,
+            #  because all topics are published as *retained* by the PTR-server.
+            #  However, a short timeout is respected before raising a `KeyError`:
+            time.sleep(200e-3)
+            rv = _lut.get(parID)
+            if rv is not None:
+                return rv
+
+            # still not found? give some useful hints for the user not to go crazy:
+            error_hint = (
+                "act" if parID in self.act_values else
+                "set" if parID in self.set_values else
+                "")
+            raise KeyError(str(parID) + (' (did you mean `kind="%s"`?)' % error_hint) if error_hint else "")
+        return _lut[parID]
+
+    def get_table(self, table_name):
+        timeout_s = 10
+        started_at = time.monotonic()
+        try:
+            while time.monotonic() < started_at + timeout_s:
+                # confirm change of state:
+                if not self._calcconzinfo[0] is _NOT_INIT:
+                    return self._calcconzinfo[0].tables[table_name]
+    
+                time.sleep(10e-3)
+            else:
+                raise TimeoutError(f"[{self}] unable to retrieve calc-conz-info from PTR server");
+        except KeyError as exc:
+            raise KeyError(str(exc) + f", possible values: {list(CalcConzInfo.tables.keys())}")
+
+    def set(self, parID, new_value, unit='-'):
+        '''Set a 'new_value' to 'parID' in the DataCollection.'''
+        if not self.is_connected:
+            raise Exception(f"[{self}] no connection to instrument");
+
+        raise NotImplementedError("DataCollection/Set, did you mean .write(parID)?")
+
+        topic, qos, retain = "DataCollection/Set/" + str(parID), 1, True
+        log.info(f"setting '{parID}' ~> [{new_value}]")
+        payload = {
+            "Header":      _build_header(),
+            "DataElement": _build_data_element(new_value, unit),
+        }
+        return self.publish_with_ack(topic, json.dumps(payload), qos=qos, retain=retain)
+
+    def filter_schedule(self, parID):
+        '''Returns a list with the upcoming write commands for 'parID' in ascending order.'''
+        return (cmd for cmd in self.current_schedule if cmd["ParaID"] == str(parID))
+
+    def write(self, parID, new_value):
+        '''Write a 'new_value' to 'parID' directly.'''
+        if not self.is_connected:
+            raise Exception(f"[{self}] no connection to instrument");
+
+        if not 'W' in _par_id_info.loc[parID].Access:  # may raise KeyError!
+            raise ValueError(f"'{parID}' is read-only")
+
+        if parID in __class__.set_value_limit and new_value > __class__.set_value_limit[parID]:
+            raise ValueError("set value limit of {__class__.set_value_limit[parID]} on '{parID}'")
+
+        topic, qos, retain = "IC_Command/Write/Direct", 1, False
+        log.info(f"writing '{parID}' ~> [{new_value}]")
+        cmd = _build_write_command(parID, new_value)
+        payload = {
+            "Header": _build_header(),
+            "CMDs": [ cmd, ]
+        }
+        return self.publish_with_ack(topic, json.dumps(payload), qos=qos, retain=retain)
+
+    def schedule(self, parID, new_value, future_cycle):
+        '''Schedule a 'new_value' to 'parID' for the given 'future_cycle'.
+
+        If 'future_cycle' is in fact in the past, the behaviour is defined by IoniTOF
+        (most likely the command is ignored). To be sure, the '.current_cycle' should
+        be checked before and after running the '.schedule' command programmatically!
+        '''
+        if not self.is_connected:
+            raise Exception(f"[{self}] no connection to instrument");
+
+        if not 'W' in _par_id_info.loc[parID].Access:  # may raise KeyError!
+            raise ValueError(f"'{parID}' is read-only")
+
+        if parID in __class__.set_value_limit and new_value > __class__.set_value_limit[parID]:
+            raise ValueError("set value limit of {__class__.set_value_limit[parID]} on '{parID}'")
+
+        if (future_cycle == 0 and not self.is_running):
+            # Note: ioniTOF40 doesn't handle scheduling for the 0th cycle!
+            if parID == "AME_ActionNumber":
+                # a) the action-number will trigger a script for the 0th cycle, so
+                #    we *must* be scheduling it!
+                self.write("AME_ActionNumber", new_value)
+            elif parID.startswith("AME_"):
+                # b) the AME-numbers cannot (currently) be set (i.e. written), but since
+                #    they are inserted just *before* the cycle, this will work just fine:
+                future_cycle = 1
+            else:
+                # c) in all other cases, let's assume the measurement will start soon
+                #    and dare to write immediately, skipping the schedule altogether:
+                log.debug(f"immediately writing {parID = } @ cycle '0' (measurement stopped)")
+                return self.write(parID, new_value)
+
+        if not future_cycle > self.current_cycle:
+            log.warn(f"attempting to schedule past cycle, hope you know what you're doing");
+            pass  # and at least let's debug it in MQTT browser (see also doc-string above)!
+
+        topic, qos, retain = "IC_Command/Write/Scheduled", 1, False
+        log.info(f"scheduling '{parID}' ~> [{new_value}] for cycle ({future_cycle})")
+        cmd = _build_write_command(parID, new_value, future_cycle)
+        payload = {
+            "Header": _build_header(),
+            "CMDs": [ cmd, ]
+        }
+        return self.publish_with_ack(topic, json.dumps(payload), qos=qos, retain=retain)
+
+    def schedule_filename(self, path, future_cycle):
+        '''Start writing to a new .h5 file with the beginning of 'future_cycle'.'''
+        assert str(path), "filename cannot be empty!"
+        # try to make sure that IoniTOF accepts the path:
+        if self.host == '127.0.0.1':
+            os.makedirs(os.path.dirname(path), exist_ok=True)
+            try:
+                with open(path, 'x'):
+                    log.info("touched new file:", path)
+            except FileExistsError as exc:
+                log.error(f"new filename '{path}' already exists and will not be scheduled!")
+                return
+
+        return self.schedule('ACQ_SRV_SetFullStorageFile', path.replace('/', '\\'), future_cycle)
+
+    def start_measurement(self, path=None):
+        '''Start a new measurement and block until the change is confirmed.
+
+        If 'path' is not None, write to the given .h5 file.
+        '''
+        if not path:
+            self.write('ACQ_SRV_Start_Meas_Quick', True)
+        else:
+            self.write('ACQ_SRV_Start_Meas_Record', path.replace('/', '\\'))
+        timeout_s = 30
+        started_at = time.monotonic()
+        while time.monotonic() < started_at + timeout_s:
+            if self.is_running:
+                break
+
+            time.sleep(10e-3)
+        else:
+            self.disconnect()
+            raise TimeoutError(f"[{self}] error starting measurement");
+
+    def stop_measurement(self, future_cycle=None):
+        '''Stop the current measurement and block until the change is confirmed.
+
+        If 'future_cycle' is not None and in the future, schedule the stop command.'''
+        if future_cycle is None or not future_cycle > self._overallcycle[0]:
+            self.write('ACQ_SRV_Stop_Meas', True)
+        else:
+            self.schedule('ACQ_SRV_Stop_Meas', True, future_cycle)
+        # may need to wait until the scheduled event..
+        if future_cycle is not None:
+            self.block_until(future_cycle)
+        # ..for this timeout to be applicable:
+        timeout_s = 30
+        started_at = time.monotonic()
+        while time.monotonic() < started_at + timeout_s:
+            # confirm change of state:
+            if not self.is_running:
+                break
+
+            time.sleep(10e-3)
+        else:
+            self.disconnect()
+            raise TimeoutError(f"[{self}] error stopping measurement");
+
+    def block_until(self, cycle):
+        '''Blocks the current thread until at least 'cycle' has passed or acquisition stopped.
+
+        Returns the actual current cycle.
+        '''
+        while self.is_running:
+            if self._overallcycle[0] >= int(cycle):
+                break
+            time.sleep(10e-3)
+        else:
+            return 0
+
+        return self._overallcycle[0]
+
+    def iter_specdata(self, timeout_s=None, buffer_size=300):
+        '''Returns an iterator over the fullcycle-data as long as it is available.
+
+        * This will wait up to `timeout_s` (or indefinitely if `None`) for a
+          measurement to start or raise a TimeoutError (default: None).
+        * Elements will be buffered up to a maximum of `buffer_size` cycles (default: 300).
+        * Cycles recorded prior to calling `next()` on the iterator may be missed,
+          so ideally this should be set up before any measurement is running.
+        * Once the measurement stops, this iterator will raise `StopIteration`.
+        * [Important]: When the buffer runs full, a `queue.Full` exception will be raised!
+          Therefore, the caller should consume the iterator as soon as possible while the
+          measurement is running.
+        '''
+        q = queue.Queue(buffer_size)
+        topic = "DataCollection/Act/ACQ_SRV_FullCycleData"
+        qos = 2
+
+        def callback(client, self, msg):
+            try:
+                _fc = _parse_fullcycle(msg.payload, need_add_data=True)
+                # IoniTOF cycle-indexing starts at 1, while 0 marks idle state:
+                if _fc.timecycle.abs_cycle > 0:
+                    q.put_nowait(_fc)
+                log.debug(f"received fullcycle, buffer at ({q.qsize()}/{q.maxsize})")
+            except queue.Full:
+                # DO NOT FAIL INSIDE THE CALLBACK!
+                log.error(f"iter_specdata({q.maxsize}): fullcycle buffer overrun!")
+                client.unsubscribe(topic)
+            except Exception as ex:
+                # DO NOT FAIL INSIDE THE CALLBACK!
+                log.warning(f"got {ex!r} while parsing {len(msg.payload) = }")
+
+        if not self.is_connected:
+            raise Exception("no connection to MQTT broker")
+
+        # Note: when using a simple generator function like this, the following lines
+        #  will not be excecuted until the first call to `next` on the iterator!
+        #  this means, the callback will not yet be executed, the queue not filled 
+        #  and we might miss the first cycles...
+        self.client.message_callback_add(topic, callback)
+        self.client.subscribe(topic, qos)
+        try:
+            # Note: Prior to 3.0 on POSIX systems, and for *all versions on Windows*,
+            #  if block is true and timeout is None, [the q.get()] operation goes into an
+            #  uninterruptible wait on an underlying lock. This means that no exceptions
+            #  can occur, and in particular a SIGINT will not trigger a KeyboardInterrupt!
+            if timeout_s is None and not self.is_running:
+                log.warn(f"waiting indefinitely for measurement to run...")
+
+            yield q.get(block=True, timeout=timeout_s)
+
+            # make double sure that there's more to come..
+            _started_at = time.monotonic()
+            while timeout_s is None or time.monotonic() < _started_at + timeout_s:
+                if self.is_running:
+                    break
+
+                time.sleep(10e-3)
+            else:
+                raise TimeoutError(f"[{self}] received specdata, but measurement won't start");
+
+            while self.is_running or not q.empty():
+                if q.full():
+                    # re-raise what we swallowed in the callback..
+                    raise queue.Full
+
+                if not self.is_connected:
+                    # no more data will come, so better prevent a deadlock:
+                    break
+
+                try:
+                    yield q.get(block=True, timeout=1.0)  # seconds
+                except queue.Empty:
+                    continue
+
+        except queue.Empty:
+            assert timeout_s is not None, "this should never happen"
+            raise TimeoutError(f"no measurement running after {timeout_s} seconds")
+
+        finally:
+            #  ...also, when using more than one iterator, the first to finish will
+            #  unsubscribe and cause all others to stop maybe before the time!
+            #  all of this might not actually be an issue right now, but
+            # TODO :: fix this weird behaviour (can only be done by implementing the
+            #  iterator-protocol properly using a helper class)!
+            self.client.unsubscribe(topic)
+            self.client.message_callback_remove(topic)
+
+    iter_specdata.__doc__ += _parse_fullcycle.__doc__
+
+    def __repr__(self):
+        return f"<{self.__class__.__name__}[{self.host}]>"
+