…erences

…oth transition.

New classes are now called JSONSerializableDecoder/-Encoder .

…ible with old for transition period and marked old routines as deprecated.

All now implement get_serialization_data and deserialize for old and new
serialization routines for a transition period.

Methods did not call Serializable::get_serialization_data() and thus did not inculde #type and #identifier keys in the returned dictionaries. Tests did not cover this.
Moved identifier_name and type_identifier_name from JSONSerializableEncoder/-Decoder to Serializable.

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

Implementation was inconsistent with _temporary_storage usage in other places.

…, TablePT

DummySerializable now behaves as follows: Takes any number of keyword-arguments and stores them as class attributes. get_serialization_data() returns a dictionary with (key, value) pairs for all attributes given this way. This is compatible with previous usage (where only a "data" attribute of type Any was given and get_serialization_data() returned a dictionary with only key "data" and whatever given value).
Removed duplicated definition of DummySerializable further down the same file.
NestedDummySerializable now is also compatible with new serialization
routines and implements deserialize().

..to adapt syntax/type to behavior.

Slight change in DummySerializableTests.make_kwargs to extend testing of DummySerializable funcitonality a bit.

…format.

Pulse is read using old serialization routines from a source storage backend and destination storage backend to a new using the new routines.

Method should list all available serializables. No tests currently.

…tly.

convert_pulses_in_storage converts all pulses in a given source StorageBackend
from the old serialization format into the new and stores them in a given
destination StorageBackend.
Both conversion methods now raise an exception if they would need to
overwrite identifiers already existing in the destination backend.
Commit also includes more extensive tests for both conversion methods as
well as some style corretions for already existing tests.

Added get_default_pulse_registration() method to query current default pulse registration.
Added method PulseStorage.as_default_registry(), a context manager method that causes the PulseStorage object to replace the current default pulse registration for the duration of a "with:" block.
Added method PulseStorage.set_to_default_registry() which permanently sets the PulseStorage object as the default pulse registration.

…erializableTests.

Adapted tests.pulses.sequencing_dummies.DummyPulseTemplate to have sensible serialization semantics so it can by higher order PTs during tests.

Closes #270. However, test_flush_destroy_object (testing that PulseStorage flushes on destruction) inexplicably fails now. Probably some timing issues with garbage collection. Refer to issue #281 .

This makes method flush() obsolete, so it was removed.

JSONEncoder previously did not check whether subpulses are identical to already known pulses with the same id but just replaced them with a reference to the already known subpulse in the serialization. This could lead to unexpected behavior. JSONEncoder now raises a RuntimeError instead in this case.
PulseStorage adopts a transaction model and takes care that in case of such an exception, a pulse is not partially serialized, i.e., if a pulse serialization aborts due to a duplicated subpulse identifier, no changes to the storage is made (that is, new subpulses that were already succesfully serialized will be discarded if an error is encountered for another subpulse).

…_compatability

backward-compatible implementations of get_serialization_data and deserialize

…t to each init method.

Trigger garbage collection on each pulse initialization.

DummySerializable's constructor now also passes on the registry argument.
Fixed all failing tests.

…erializable with the same name

Fix (global) pulse registry

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

…_flushing

…_flushing

NestedDummySerializable's constructor now passed the registry argument through to superclass.

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

Fix issues 271&281

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

…dded __delitem__() to PulseStorage.

__delitem__() has no test routines yet.
Closes #273.

…on_output

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

…_context_manager

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

Issue 270

…manager

Merge issue #275

# Conflicts:
#	tests/pulses/table_pulse_template_tests.py

…test

Add test for __delitem__

Increase test coverage

…' into issues/250_serialization_conversion_script

… routines.

Also added tests (these were also missing for the old serialization routines!).

…ires a Serializer object as non-optional positional argument anymore.

Fixes for new serialization

…stry().

set_default_pulse_registry() allows to set a new user-specified registry as the global default pulse registry.
new_default_pulse_registry() creates a new, empty global default pulse registry (as a WeakValueDictionary).

…_conversion_script

…try_none

…_as_default_issue

…nes.

Ensures that the conversion using convert_pulses_in_storage() results in an equal PulseTemplate.

…efault registry.

…t_id in SerializableTests (and added assert_equal_instance building on it)

and made sure that all implementations adhere to that

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

… under a different than its own identifier.

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

…d Serializable.__init__()

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

…lt_issue

Issues/301 pulse storage as default issue

…re they serialize correctly.

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

…egistration

…try_none

…global registry after they are completely initialized.

Serializable now does not perform registration in __init__ but provides a method _register() which MUST be called by leaf-level subclasses. (This is probably also not a good solution in the long term as it effectively prevents further subclassing from those classes but it works for now).

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

Implementations of __len__ and __iter__ curently raise NotImplementedError s.

Signed-off-by: Lukas Prediger <lukas.prediger@rwth-aachen.de>

…_conversion_script

Issues/299 default registry none

…on_script

Serialization conversion script

…id_inconsistencies

…egistration

…istencies

Issues/272 pulse storage id inconsistencies

…egistration

Issues/298 early pulse registration

…pe_annotations_for_registry

…egistry occurrences accordingly.

…tions_for_registry

Issues/308 consistent type annotations for registry

…nt over several instantiations.

…t for this

Increased test coverage for serialization