from collections import OrderedDict
from inspect import isclass
from artiq.protocols import pyon
from artiq.language import units
__all__ = ["NoDefault",
"PYONValue", "BooleanValue", "EnumerationValue",
"NumberValue", "StringValue",
"HasEnvironment", "Experiment", "EnvExperiment"]
[docs]class NoDefault:
"""Represents the absence of a default value."""
pass
class DefaultMissing(Exception):
"""Raised by the ``default`` method of argument processors when no default
value is available."""
pass
class _SimpleArgProcessor:
def __init__(self, default=NoDefault):
# If default is a list, it means multiple defaults are specified, with
# decreasing priority.
if isinstance(default, list):
raise NotImplementedError
if default is not NoDefault:
self.default_value = default
def default(self):
if not hasattr(self, "default_value"):
raise DefaultMissing
return self.default_value
def process(self, x):
return x
def describe(self):
d = {"ty": self.__class__.__name__}
if hasattr(self, "default_value"):
d["default"] = self.default_value
return d
[docs]class PYONValue(_SimpleArgProcessor):
"""An argument that can be any PYON-serializable value."""
def __init__(self, default=NoDefault):
# Override the _SimpleArgProcessor init, as list defaults are valid
# PYON values
if default is not NoDefault:
self.default_value = default
def process(self, x):
return pyon.decode(x)
def describe(self):
d = {"ty": self.__class__.__name__}
if hasattr(self, "default_value"):
d["default"] = pyon.encode(self.default_value)
return d
[docs]class BooleanValue(_SimpleArgProcessor):
"""A boolean argument."""
pass
[docs]class EnumerationValue(_SimpleArgProcessor):
"""An argument that can take a string value among a predefined set of
values.
:param choices: A list of string representing the possible values of the
argument.
"""
def __init__(self, choices, default=NoDefault):
_SimpleArgProcessor.__init__(self, default)
assert default is NoDefault or default in choices
self.choices = choices
def describe(self):
d = _SimpleArgProcessor.describe(self)
d["choices"] = self.choices
return d
[docs]class NumberValue(_SimpleArgProcessor):
"""An argument that can take a numerical value.
If ndecimals = 0, scale = 1 and step is integer, then it returns
an integer value. Otherwise, it returns a floating point value.
The simplest way to represent an integer argument is
``NumberValue(step=1, ndecimals=0)``.
When ``scale`` is not specified, and the unit is a common one (i.e.
defined in ``artiq.language.units``), then the scale is obtained from
the unit using a simple string match. For example, milliseconds (``"ms"``)
units set the scale to 0.001. No unit (default) corresponds to a scale of
1.0.
For arguments with uncommon or complex units, use both the unit parameter
(a string for display) and the scale parameter (a numerical scale for
experiments).
For example, ``NumberValue(1, unit="xyz", scale=0.001)`` will display as
1 xyz in the GUI window because of the unit setting, and appear as the
numerical value 0.001 in the code because of the scale setting.
:param unit: A string representing the unit of the value.
:param scale: A numerical scaling factor by which the displayed value is
multiplied when referenced in the experiment.
:param step: The step with which the value should be modified by up/down
buttons in a UI. The default is the scale divided by 10.
:param min: The minimum value of the argument.
:param max: The maximum value of the argument.
:param ndecimals: The number of decimals a UI should use.
"""
def __init__(self, default=NoDefault, unit="", scale=None,
step=None, min=None, max=None, ndecimals=2):
if scale is None:
if unit == "":
scale = 1.0
else:
try:
scale = getattr(units, unit)
except AttributeError:
raise KeyError("Unit {} is unknown, you must specify "
"the scale manually".format(unit))
if step is None:
step = scale/10.0
if default is not NoDefault:
self.default_value = default
self.unit = unit
self.scale = scale
self.step = step
self.min = min
self.max = max
self.ndecimals = ndecimals
def _is_int(self):
return (self.ndecimals == 0
and int(self.step) == self.step
and self.scale == 1)
def default(self):
if not hasattr(self, "default_value"):
raise DefaultMissing
if self._is_int():
return int(self.default_value)
else:
return float(self.default_value)
def process(self, x):
if self._is_int():
return int(x)
else:
return float(x)
def describe(self):
d = {"ty": self.__class__.__name__}
if hasattr(self, "default_value"):
d["default"] = self.default_value
d["unit"] = self.unit
d["scale"] = self.scale
d["step"] = self.step
d["min"] = self.min
d["max"] = self.max
d["ndecimals"] = self.ndecimals
return d
[docs]class StringValue(_SimpleArgProcessor):
"""A string argument."""
pass
class TraceArgumentManager:
def __init__(self):
self.requested_args = OrderedDict()
def get(self, key, processor, group):
self.requested_args[key] = processor, group
return None
class ProcessArgumentManager:
def __init__(self, unprocessed_arguments):
self.unprocessed_arguments = unprocessed_arguments
def get(self, key, processor, group):
if key in self.unprocessed_arguments:
r = processor.process(self.unprocessed_arguments[key])
else:
r = processor.default()
return r
[docs]class HasEnvironment:
"""Provides methods to manage the environment of an experiment (arguments,
devices, datasets)."""
def __init__(self, managers_or_parent, *args, **kwargs):
if isinstance(managers_or_parent, tuple):
self.__device_mgr = managers_or_parent[0]
self.__dataset_mgr = managers_or_parent[1]
self.__argument_mgr = managers_or_parent[2]
else:
self.__device_mgr = managers_or_parent.__device_mgr
self.__dataset_mgr = managers_or_parent.__dataset_mgr
self.__argument_mgr = managers_or_parent.__argument_mgr
self.__in_build = True
self.build(*args, **kwargs)
self.__in_build = False
[docs] def build(self):
"""Should be implemented by the user to request arguments.
Other initialization steps such as requesting devices may also be
performed here.
There are two situations where the requested devices are replaced by
``DummyDevice()`` and arguments are set to their defaults (or ``None``)
instead: when the repository is scanned to build the list of
available experiments and when the dataset browser ``artiq_browser``
is used to open or run the analysis stage of an experiment. Do not
rely on being able to operate on devices or arguments in ``build()``.
Datasets are read-only in this method.
Leftover positional and keyword arguments from the constructor are
forwarded to this method. This is intended for experiments that are
only meant to be executed programmatically (not from the GUI)."""
pass
[docs] def get_argument(self, key, processor, group=None):
"""Retrieves and returns the value of an argument.
This function should only be called from ``build``.
:param key: Name of the argument.
:param processor: A description of how to process the argument, such
as instances of ``BooleanValue`` and ``NumberValue``.
:param group: An optional string that defines what group the argument
belongs to, for user interface purposes.
"""
if not self.__in_build:
raise TypeError("get_argument() should only "
"be called from build()")
return self.__argument_mgr.get(key, processor, group)
[docs] def setattr_argument(self, key, processor=None, group=None):
"""Sets an argument as attribute. The names of the argument and of the
attribute are the same.
The key is added to the instance's kernel invariants."""
setattr(self, key, self.get_argument(key, processor, group))
kernel_invariants = getattr(self, "kernel_invariants", set())
self.kernel_invariants = kernel_invariants | {key}
[docs] def get_device_db(self):
"""Returns the full contents of the device database."""
return self.__device_mgr.get_device_db()
[docs] def get_device(self, key):
"""Creates and returns a device driver."""
return self.__device_mgr.get(key)
[docs] def setattr_device(self, key):
"""Sets a device driver as attribute. The names of the device driver
and of the attribute are the same.
The key is added to the instance's kernel invariants."""
setattr(self, key, self.get_device(key))
kernel_invariants = getattr(self, "kernel_invariants", set())
self.kernel_invariants = kernel_invariants | {key}
[docs] def set_dataset(self, key, value,
broadcast=False, persist=False, save=True):
"""Sets the contents and handling modes of a dataset.
Datasets must be scalars (``bool``, ``int``, ``float`` or NumPy scalar)
or NumPy arrays.
:param broadcast: the data is sent in real-time to the master, which
dispatches it.
:param persist: the master should store the data on-disk. Implies
broadcast.
:param save: the data is saved into the local storage of the current
run (archived as a HDF5 file).
"""
self.__dataset_mgr.set(key, value, broadcast, persist, save)
[docs] def mutate_dataset(self, key, index, value):
"""Mutate an existing dataset at the given index (e.g. set a value at
a given position in a NumPy array)
If the dataset was created in broadcast mode, the modification is
immediately transmitted.
If the index is a tuple of integers, it is interpreted as
``slice(*index)``.
If the index is a tuple of tuples, each sub-tuple is interpreted
as ``slice(*sub_tuple)`` (multi-dimensional slicing)."""
self.__dataset_mgr.mutate(key, index, value)
[docs] def get_dataset(self, key, default=NoDefault):
"""Returns the contents of a dataset.
The local storage is searched first, followed by the master storage
(which contains the broadcasted datasets from all experiments) if the
key was not found initially.
If the dataset does not exist, returns the default value. If no default
is provided, raises ``KeyError``.
"""
try:
return self.__dataset_mgr.get(key)
except KeyError:
if default is NoDefault:
raise
else:
return default
[docs] def setattr_dataset(self, key, default=NoDefault):
"""Sets the contents of a dataset as attribute. The names of the
dataset and of the attribute are the same."""
setattr(self, key, self.get_dataset(key, default))
[docs]class Experiment:
"""Base class for top-level experiments.
Deriving from this class enables automatic experiment discovery in
Python modules.
"""
[docs] def prepare(self):
"""Entry point for pre-computing data necessary for running the
experiment.
Doing such computations outside of ``run`` enables more efficient
scheduling of multiple experiments that need to access the shared
hardware during part of their execution.
This method must not interact with the hardware.
"""
pass
[docs] def run(self):
"""The main entry point of the experiment.
This method must be overloaded by the user to implement the main
control flow of the experiment.
This method may interact with the hardware.
The experiment may call the scheduler's ``pause`` method while in
``run``.
"""
raise NotImplementedError
[docs] def analyze(self):
"""Entry point for analyzing the results of the experiment.
This method may be overloaded by the user to implement the analysis
phase of the experiment, for example fitting curves.
Splitting this phase from ``run`` enables tweaking the analysis
algorithm on pre-existing data, and CPU-bound analyses to be run
overlapped with the next experiment in a pipelined manner.
This method must not interact with the hardware.
"""
pass
[docs]class EnvExperiment(Experiment, HasEnvironment):
"""Base class for top-level experiments that use the ``HasEnvironment``
environment manager.
Most experiment should derive from this class."""
pass
def is_experiment(o):
"""Checks if a Python object is a top-level experiment class."""
return (isclass(o)
and issubclass(o, Experiment)
and o is not Experiment
and o is not EnvExperiment)