]>
git.madduck.net Git - etc/taskwarrior.git/blobdiff - tasklib/task.py
madduck's git repository
Every one of the projects in this repository is available at the canonical
URL git://git.madduck.net/madduck/pub/<projectpath> — see
each project's metadata for the exact URL.
All patches and comments are welcome. Please squash your changes to logical
commits before using git-format-patch and git-send-email to
patches@ git. madduck. net .
If you'd read over the Git project's submission guidelines and adhered to them,
I'd be especially grateful.
SSH access, as well as push access can be individually
arranged .
If you use my repositories frequently, consider adding the following
snippet to ~/.gitconfig and using the third clone URL listed for each
project:
[url "git://git.madduck.net/madduck/"]
insteadOf = madduck:
from __future__ import print_function
import copy
from __future__ import print_function
import copy
import json
import logging
import os
import six
import sys
import json
import logging
import os
import six
import sys
-from backends import TaskWarrior, TaskWarriorException
-from serializing import SerializingObject
+from .serializing import SerializingObject
DATE_FORMAT = '%Y%m%dT%H%M%SZ'
REPR_OUTPUT_SIZE = 10
PENDING = 'pending'
COMPLETED = 'completed'
DATE_FORMAT = '%Y%m%dT%H%M%SZ'
REPR_OUTPUT_SIZE = 10
PENDING = 'pending'
COMPLETED = 'completed'
+DELETED = 'deleted'
+WAITING = 'waiting'
+RECURRING = 'recurring'
logger = logging.getLogger(__name__)
logger = logging.getLogger(__name__)
def __len__(self):
return len(self.viewed_dict)
def __len__(self):
return len(self.viewed_dict)
+ def __unicode__(self):
+ return six.u('ReadOnlyDictView: {0}'.format(repr(self.viewed_dict)))
+
+ __repr__ = __unicode__
+
def get(self, key, default=None):
return copy.deepcopy(self.viewed_dict.get(key, default))
def get(self, key, default=None):
return copy.deepcopy(self.viewed_dict.get(key, default))
if update_original:
self._original_data = copy.deepcopy(self._data)
if update_original:
self._original_data = copy.deepcopy(self._data)
def __getitem__(self, key):
# This is a workaround to make TaskResource non-iterable
# over simple index-based iteration
def __getitem__(self, key):
# This is a workaround to make TaskResource non-iterable
# over simple index-based iteration
# to pass that to TaskWarrior.
data_tuples = filter(lambda t: t[1] is not '', data_tuples)
data = dict(data_tuples)
# to pass that to TaskWarrior.
data_tuples = filter(lambda t: t[1] is not '', data_tuples)
data = dict(data_tuples)
- return json.dumps(data, separators=(',',':'))
+ return json.dumps(data, separators=(',', ':'))
@property
def _modified_fields(self):
@property
def _modified_fields(self):
def __init__(self, task, data=None):
self.task = task
self._load_data(data or dict())
def __init__(self, task, data=None):
self.task = task
self._load_data(data or dict())
- super(TaskAnnotation, self).__init__(task.warrior )
+ super(TaskAnnotation, self).__init__(task.backend )
def remove(self):
self.task.remove_annotation(self)
def remove(self):
self.task.remove_annotation(self)
# their data dics are the same
return self.task == other.task and self._data == other._data
# their data dics are the same
return self.task == other.task and self._data == other._data
+ def __ne__(self, other):
+ return not self.__eq__(other)
+
- def from_input(cls, input_file=sys.stdin, modify=None, warrior =None):
+ def from_input(cls, input_file=sys.stdin, modify=None, backend =None):
"""
Creates a Task object, directly from the stdin, by reading one line.
If modify=True, two lines are used, first line interpreted as the
"""
Creates a Task object, directly from the stdin, by reading one line.
If modify=True, two lines are used, first line interpreted as the
modify = name.startswith('on-modify') if modify is None else modify
# Create the TaskWarrior instance if none passed
modify = name.startswith('on-modify') if modify is None else modify
# Create the TaskWarrior instance if none passed
+ if backend is None:
+ backends = importlib.import_module('tasklib.backends')
hook_parent_dir = os.path.dirname(os.path.dirname(sys.argv[0]))
hook_parent_dir = os.path.dirname(os.path.dirname(sys.argv[0]))
- warrior = TaskWarrior(data_location=hook_parent_dir)
+ backend = backends. TaskWarrior(data_location=hook_parent_dir)
# TaskWarrior instance is set to None
# TaskWarrior instance is set to None
# Load the data from the input
task._load_data(json.loads(input_file.readline().strip()))
# Load the data from the input
task._load_data(json.loads(input_file.readline().strip()))
- def __init__(self, warrior , **kwargs):
- super(Task, self).__init__(warrior )
+ def __init__(self, backend , **kwargs):
+ super(Task, self).__init__(backend )
# Check that user is not able to set read-only value in __init__
for key in kwargs.keys():
# Check that user is not able to set read-only value in __init__
for key in kwargs.keys():
# If the tasks are not saved, compare the actual instances
return id(self) == id(other)
# If the tasks are not saved, compare the actual instances
return id(self) == id(other)
+ def __ne__(self, other):
+ return not self.__eq__(other)
def __hash__(self):
if self['uuid']:
def __hash__(self):
if self['uuid']:
def pending(self):
return self['status'] == six.text_type('pending')
def pending(self):
return self['status'] == six.text_type('pending')
+ @property
+ def recurring(self):
+ return self['status'] == six.text_type('recurring')
+
@property
def active(self):
return self['start'] is not None
@property
def active(self):
return self['start'] is not None
# Check that all the tasks are saved
for task in (cur_dependencies or set()):
if not task.saved:
# Check that all the tasks are saved
for task in (cur_dependencies or set()):
if not task.saved:
- raise Task.NotSaved('Task \'%s\' needs to be saved before '
- 'it can be set as dependency.' % task)
+ raise Task.NotSaved(
+ 'Task \'%s\' needs to be saved before '
+ 'it can be set as dependency.' % task,
+ )
return super(Task, self).serialize_depends(cur_dependencies)
def delete(self):
if not self.saved:
return super(Task, self).serialize_depends(cur_dependencies)
def delete(self):
if not self.saved:
- raise Task.NotSaved("Task needs to be saved before it can be deleted")
+ raise Task.NotSaved(
+ 'Task needs to be saved before it can be deleted',
+ )
# Refresh the status, and raise exception if the task is deleted
self.refresh(only_fields=['status'])
if self.deleted:
# Refresh the status, and raise exception if the task is deleted
self.refresh(only_fields=['status'])
if self.deleted:
- raise Task.DeletedTask("Task was already deleted" )
+ raise Task.DeletedTask('Task was already deleted' )
self.backend.delete_task(self)
self.backend.delete_task(self)
def start(self):
if not self.saved:
def start(self):
if not self.saved:
- raise Task.NotSaved("Task needs to be saved before it can be started")
+ raise Task.NotSaved(
+ 'Task needs to be saved before it can be started',
+ )
# Refresh, and raise exception if task is already completed/deleted
self.refresh(only_fields=['status'])
if self.completed:
# Refresh, and raise exception if task is already completed/deleted
self.refresh(only_fields=['status'])
if self.completed:
- raise Task.CompletedTask("Cannot start a completed task" )
+ raise Task.CompletedTask('Cannot start a completed task' )
- raise Task.DeletedTask("Deleted task cannot be started" )
+ raise Task.DeletedTask('Deleted task cannot be started' )
- raise Task.ActiveTask("Task is already active" )
+ raise Task.ActiveTask('Task is already active' )
self.backend.start_task(self)
self.backend.start_task(self)
def stop(self):
if not self.saved:
def stop(self):
if not self.saved:
- raise Task.NotSaved("Task needs to be saved before it can be stopped")
+ raise Task.NotSaved(
+ 'Task needs to be saved before it can be stopped',
+ )
# Refresh, and raise exception if task is already completed/deleted
self.refresh(only_fields=['status'])
if not self.active:
# Refresh, and raise exception if task is already completed/deleted
self.refresh(only_fields=['status'])
if not self.active:
- raise Task.InactiveTask("Cannot stop an inactive task" )
+ raise Task.InactiveTask('Cannot stop an inactive task' )
self.backend.stop_task(self)
self.backend.stop_task(self)
def done(self):
if not self.saved:
def done(self):
if not self.saved:
- raise Task.NotSaved("Task needs to be saved before it can be completed")
+ raise Task.NotSaved(
+ 'Task needs to be saved before it can be completed',
+ )
# Refresh, and raise exception if task is already completed/deleted
self.refresh(only_fields=['status'])
if self.completed:
# Refresh, and raise exception if task is already completed/deleted
self.refresh(only_fields=['status'])
if self.completed:
- raise Task.CompletedTask("Cannot complete a completed task" )
+ raise Task.CompletedTask('Cannot complete a completed task' )
- raise Task.DeletedTask("Deleted task cannot be completed" )
+ raise Task.DeletedTask('Deleted task cannot be completed' )
self.backend.complete_task(self)
self.backend.complete_task(self)
def add_annotation(self, annotation):
if not self.saved:
def add_annotation(self, annotation):
if not self.saved:
- raise Task.NotSaved("Task needs to be saved to add annotation" )
+ raise Task.NotSaved('Task needs to be saved to add annotation' )
self.backend.annotate_task(self, annotation)
self.refresh(only_fields=['annotations'])
def remove_annotation(self, annotation):
if not self.saved:
self.backend.annotate_task(self, annotation)
self.refresh(only_fields=['annotations'])
def remove_annotation(self, annotation):
if not self.saved:
- raise Task.NotSaved("Task needs to be saved to remove annotation" )
+ raise Task.NotSaved('Task needs to be saved to remove annotation' )
if isinstance(annotation, TaskAnnotation):
annotation = annotation['description']
if isinstance(annotation, TaskAnnotation):
annotation = annotation['description']
def refresh(self, only_fields=None, after_save=False):
# Raise error when trying to refresh a task that has not been saved
if not self.saved:
def refresh(self, only_fields=None, after_save=False):
# Raise error when trying to refresh a task that has not been saved
if not self.saved:
- raise Task.NotSaved("Task needs to be saved to be refreshed" )
+ raise Task.NotSaved('Task needs to be saved to be refreshed' )
new_data = self.backend.refresh_task(self, after_save=after_save)
if only_fields:
to_update = dict(
new_data = self.backend.refresh_task(self, after_save=after_save)
if only_fields:
to_update = dict(
- [(k, new_data.get(k)) for k in only_fields])
+ [(k, new_data.get(k)) for k in only_fields],
+ )
self._update_data(to_update, update_original=True)
else:
self._load_data(new_data)
self._update_data(to_update, update_original=True)
else:
self._load_data(new_data)
class TaskQuerySet(object):
"""
Represents a lazy lookup for a task objects.
"""
class TaskQuerySet(object):
"""
Represents a lazy lookup for a task objects.
"""
- def __init__(self, warrior=None , filter_obj=None):
- self.warrior = warrior
+ def __init__(self, backend , filter_obj=None):
+ self.backend = backend
self._result_cache = None
self._result_cache = None
- self.filter_obj = filter_obj or TaskWarriorFilter(warrior )
+ self.filter_obj = filter_obj or self.backend.filter_class(backend )
def __deepcopy__(self, memo):
"""
Deep copy of a QuerySet doesn't populate the cache
"""
def __deepcopy__(self, memo):
"""
Deep copy of a QuerySet doesn't populate the cache
"""
+ obj = self.__class__(backend=self.backend )
for k, v in self.__dict__.items():
if k in ('_iter', '_result_cache'):
obj.__dict__[k] = None
for k, v in self.__dict__.items():
if k in ('_iter', '_result_cache'):
obj.__dict__[k] = None
def __repr__(self):
data = list(self[:REPR_OUTPUT_SIZE + 1])
if len(data) > REPR_OUTPUT_SIZE:
def __repr__(self):
data = list(self[:REPR_OUTPUT_SIZE + 1])
if len(data) > REPR_OUTPUT_SIZE:
- data[-1] = "...(remaining elements truncated)..."
+ data[-1] = '...(remaining elements truncated)...'
return repr(data)
def __len__(self):
return repr(data)
def __len__(self):
if klass is None:
klass = self.__class__
filter_obj = self.filter_obj.clone()
if klass is None:
klass = self.__class__
filter_obj = self.filter_obj.clone()
- c = klass(warrior=self.warrior , filter_obj=filter_obj)
+ c = klass(backend=self.backend , filter_obj=filter_obj)
c.__dict__.update(kwargs)
return c
c.__dict__.update(kwargs)
return c
"""
Fetch the tasks which match the current filters.
"""
"""
Fetch the tasks which match the current filters.
"""
- return self.warrior .filter_tasks(self.filter_obj)
+ return self.backend .filter_tasks(self.filter_obj)
def completed(self):
return self.filter(status=COMPLETED)
def completed(self):
return self.filter(status=COMPLETED)
+ def deleted(self):
+ return self.filter(status=DELETED)
+
+ def waiting(self):
+ return self.filter(status=WAITING)
+
+ def recurring(self):
+ return self.filter(status=RECURRING)
+
def filter(self, *args, **kwargs):
"""
Returns a new TaskQuerySet with the given filters added.
def filter(self, *args, **kwargs):
"""
Returns a new TaskQuerySet with the given filters added.
if not num:
raise Task.DoesNotExist(
'Task matching query does not exist. '
if not num:
raise Task.DoesNotExist(
'Task matching query does not exist. '
- 'Lookup parameters were {0}'.format(kwargs))
+ 'Lookup parameters were {0}'.format(kwargs),
+ )
raise ValueError(
'get() returned more than one Task -- it returned {0}! '
raise ValueError(
'get() returned more than one Task -- it returned {0}! '
- 'Lookup parameters were {1}'.format(num, kwargs))
+ 'Lookup parameters were {1}'.format(num, kwargs),
+ )