]> 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:

TaskWarrior: Use default .taskrc
[etc/taskwarrior.git] / tasklib / task.py
index 5e2b71b66a48d4b87e6ad8c27ccbb932bb76ace4..23f9277b4aef67b67d3a8359db0b33413eb87f53 100644 (file)
+from __future__ import print_function
+import copy
+import datetime
 import json
+import logging
 import os
+import pytz
+import six
+import sys
 import subprocess
-import tempfile
-import uuid
-
+import tzlocal
 
+DATE_FORMAT = '%Y%m%dT%H%M%SZ'
+REPR_OUTPUT_SIZE = 10
 PENDING = 'pending'
+COMPLETED = 'completed'
+
+VERSION_2_1_0 = six.u('2.1.0')
+VERSION_2_2_0 = six.u('2.2.0')
+VERSION_2_3_0 = six.u('2.3.0')
+VERSION_2_4_0 = six.u('2.4.0')
+VERSION_2_4_1 = six.u('2.4.1')
+VERSION_2_4_2 = six.u('2.4.2')
+
+logger = logging.getLogger(__name__)
+local_zone = tzlocal.get_localzone()
 
 
 class TaskWarriorException(Exception):
     pass
 
 
-class Task(object):
+class ReadOnlyDictView(object):
+    """
+    Provides simplified read-only view upon dict object.
+    """
+
+    def __init__(self, viewed_dict):
+        self.viewed_dict = viewed_dict
+
+    def __getitem__(self, key):
+        return copy.deepcopy(self.viewed_dict.__getitem__(key))
+
+    def __contains__(self, k):
+        return self.viewed_dict.__contains__(k)
+
+    def __iter__(self):
+        for value in self.viewed_dict:
+            yield copy.deepcopy(value)
+
+    def __len__(self):
+        return len(self.viewed_dict)
+
+    def get(self, key, default=None):
+        return copy.deepcopy(self.viewed_dict.get(key, default))
+
+    def items(self):
+        return [copy.deepcopy(v) for v in self.viewed_dict.items()]
+
+    def values(self):
+        return [copy.deepcopy(v) for v in self.viewed_dict.values()]
+
+
+class SerializingObject(object):
+    """
+    Common ancestor for TaskResource & TaskFilter, since they both
+    need to serialize arguments.
+
+    Serializing method should hold the following contract:
+      - any empty value (meaning removal of the attribute)
+        is deserialized into a empty string
+      - None denotes a empty value for any attribute
+
+    Deserializing method should hold the following contract:
+      - None denotes a empty value for any attribute (however,
+        this is here as a safeguard, TaskWarrior currently does
+        not export empty-valued attributes) if the attribute
+        is not iterable (e.g. list or set), in which case
+        a empty iterable should be used.
+
+    Normalizing methods should hold the following contract:
+      - They are used to validate and normalize the user input.
+        Any attribute value that comes from the user (during Task
+        initialization, assignign values to Task attributes, or
+        filtering by user-provided values of attributes) is first
+        validated and normalized using the normalize_{key} method.
+      - If validation or normalization fails, normalizer is expected
+        to raise ValueError.
+    """
+
+    def _deserialize(self, key, value):
+        hydrate_func = getattr(self, 'deserialize_{0}'.format(key),
+                               lambda x: x if x != '' else None)
+        return hydrate_func(value)
+
+    def _serialize(self, key, value):
+        dehydrate_func = getattr(self, 'serialize_{0}'.format(key),
+                                 lambda x: x if x is not None else '')
+        return dehydrate_func(value)
+
+    def _normalize(self, key, value):
+        """
+        Use normalize_<key> methods to normalize user input. Any user
+        input will be normalized at the moment it is used as filter,
+        or entered as a value of Task attribute.
+        """
+
+        # None value should not be converted by normalizer
+        if value is None:
+            return None
+
+        normalize_func = getattr(self, 'normalize_{0}'.format(key),
+                                 lambda x: x)
+
+        return normalize_func(value)
+
+    def timestamp_serializer(self, date):
+        if not date:
+            return ''
+
+        # Any serialized timestamp should be localized, we need to
+        # convert to UTC before converting to string (DATE_FORMAT uses UTC)
+        date = date.astimezone(pytz.utc)
+
+        return date.strftime(DATE_FORMAT)
+
+    def timestamp_deserializer(self, date_str):
+        if not date_str:
+            return None
+
+        # Return timestamp localized in the local zone
+        naive_timestamp = datetime.datetime.strptime(date_str, DATE_FORMAT)
+        localized_timestamp = pytz.utc.localize(naive_timestamp)
+        return localized_timestamp.astimezone(local_zone)
+
+    def serialize_entry(self, value):
+        return self.timestamp_serializer(value)
+
+    def deserialize_entry(self, value):
+        return self.timestamp_deserializer(value)
+
+    def normalize_entry(self, value):
+        return self.datetime_normalizer(value)
+
+    def serialize_modified(self, value):
+        return self.timestamp_serializer(value)
+
+    def deserialize_modified(self, value):
+        return self.timestamp_deserializer(value)
+
+    def normalize_modified(self, value):
+        return self.datetime_normalizer(value)
+
+    def serialize_start(self, value):
+        return self.timestamp_serializer(value)
+
+    def deserialize_start(self, value):
+        return self.timestamp_deserializer(value)
+
+    def normalize_start(self, value):
+        return self.datetime_normalizer(value)
+
+    def serialize_end(self, value):
+        return self.timestamp_serializer(value)
+
+    def deserialize_end(self, value):
+        return self.timestamp_deserializer(value)
+
+    def normalize_end(self, value):
+        return self.datetime_normalizer(value)
+
+    def serialize_due(self, value):
+        return self.timestamp_serializer(value)
+
+    def deserialize_due(self, value):
+        return self.timestamp_deserializer(value)
+
+    def normalize_due(self, value):
+        return self.datetime_normalizer(value)
+
+    def serialize_scheduled(self, value):
+        return self.timestamp_serializer(value)
+
+    def deserialize_scheduled(self, value):
+        return self.timestamp_deserializer(value)
+
+    def normalize_scheduled(self, value):
+        return self.datetime_normalizer(value)
+
+    def serialize_until(self, value):
+        return self.timestamp_serializer(value)
+
+    def deserialize_until(self, value):
+        return self.timestamp_deserializer(value)
+
+    def normalize_until(self, value):
+        return self.datetime_normalizer(value)
+
+    def serialize_wait(self, value):
+        return self.timestamp_serializer(value)
+
+    def deserialize_wait(self, value):
+        return self.timestamp_deserializer(value)
+
+    def normalize_wait(self, value):
+        return self.datetime_normalizer(value)
+
+    def serialize_annotations(self, value):
+        value = value if value is not None else []
+
+        # This may seem weird, but it's correct, we want to export
+        # a list of dicts as serialized value
+        serialized_annotations = [json.loads(annotation.export_data())
+                                  for annotation in value]
+        return serialized_annotations if serialized_annotations else ''
+
+    def deserialize_annotations(self, data):
+        return [TaskAnnotation(self, d) for d in data] if data else []
+
+    def serialize_tags(self, tags):
+        return ','.join(tags) if tags else ''
+
+    def deserialize_tags(self, tags):
+        if isinstance(tags, six.string_types):
+            return tags.split(',') if tags else []
+        return tags or []
+
+    def serialize_depends(self, value):
+        # Return the list of uuids
+        value = value if value is not None else set()
+        return ','.join(task['uuid'] for task in value)
+
+    def deserialize_depends(self, raw_uuids):
+        raw_uuids = raw_uuids or ''  # Convert None to empty string
+        uuids = raw_uuids.split(',')
+        return set(self.warrior.tasks.get(uuid=uuid) for uuid in uuids if uuid)
+
+    def datetime_normalizer(self, value):
+        """
+        Normalizes date/datetime value (considered to come from user input)
+        to localized datetime value. Following conversions happen:
+
+        naive date -> localized datetime with the same date, and time=midnight
+        naive datetime -> localized datetime with the same value
+        localized datetime -> localized datetime (no conversion)
+        """
+
+        if (isinstance(value, datetime.date)
+            and not isinstance(value, datetime.datetime)):
+            # Convert to local midnight
+            value_full = datetime.datetime.combine(value, datetime.time.min)
+            localized = local_zone.localize(value_full)
+        elif isinstance(value, datetime.datetime) and value.tzinfo is None:
+            # Convert to localized datetime object
+            localized = local_zone.localize(value)
+        else:
+            # If the value is already localized, there is no need to change
+            # time zone at this point. Also None is a valid value too.
+            localized = value
+
+        return localized
+
+    def normalize_uuid(self, value):
+        # Enforce sane UUID
+        if not isinstance(value, six.string_types) or value == '':
+            raise ValueError("UUID must be a valid non-empty string, "
+                             "not: {}".format(value))
+
+        return value
+
+
+class TaskResource(SerializingObject):
+    read_only_fields = []
+
+    def _load_data(self, data):
+        self._data = dict((key, self._deserialize(key, value))
+                          for key, value in data.items())
+        # We need to use a copy for original data, so that changes
+        # are not propagated.
+        self._original_data = copy.deepcopy(self._data)
+
+    def _update_data(self, data, update_original=False):
+        """
+        Low level update of the internal _data dict. Data which are coming as
+        updates should already be serialized. If update_original is True, the
+        original_data dict is updated as well.
+        """
+        self._data.update(dict((key, self._deserialize(key, value))
+                               for key, value in data.items()))
+
+        if update_original:
+            self._original_data = copy.deepcopy(self._data)
 
-    def __init__(self, warrior, data={}):
-        self.warrior = warrior
-        self._data = data
 
     def __getitem__(self, key):
+        # This is a workaround to make TaskResource non-iterable
+        # over simple index-based iteration
+        try:
+            int(key)
+            raise StopIteration
+        except ValueError:
+            pass
+
+        if key not in self._data:
+            self._data[key] = self._deserialize(key, None)
+
         return self._data.get(key)
 
-    def __setitem__(self, key, val):
-        self._data[key] = val
+    def __setitem__(self, key, value):
+        if key in self.read_only_fields:
+            raise RuntimeError('Field \'%s\' is read-only' % key)
+
+        # Normalize the user input before saving it
+        value = self._normalize(key, value)
+        self._data[key] = value
+
+    def __str__(self):
+        s = six.text_type(self.__unicode__())
+        if not six.PY3:
+            s = s.encode('utf-8')
+        return s
+
+    def __repr__(self):
+        return str(self)
+
+    def export_data(self):
+        """
+        Exports current data contained in the Task as JSON
+        """
+
+        # We need to remove spaces for TW-1504, use custom separators
+        data_tuples = ((key, self._serialize(key, value))
+                       for key, value in six.iteritems(self._data))
+
+        # Empty string denotes empty serialized value, we do not want
+        # 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=(',',':'))
+
+    @property
+    def _modified_fields(self):
+        writable_fields = set(self._data.keys()) - set(self.read_only_fields)
+        for key in writable_fields:
+            new_value = self._data.get(key)
+            old_value = self._original_data.get(key)
+
+            # Make sure not to mark data removal as modified field if the
+            # field originally had some empty value
+            if key in self._data and not new_value and not old_value:
+                continue
+
+            if new_value != old_value:
+                yield key
+
+    @property
+    def modified(self):
+        return bool(list(self._modified_fields))
+
+
+class TaskAnnotation(TaskResource):
+    read_only_fields = ['entry', 'description']
+
+    def __init__(self, task, data={}):
+        self.task = task
+        self._load_data(data)
+
+    def remove(self):
+        self.task.remove_annotation(self)
 
     def __unicode__(self):
-        return self._data.get('description')
+        return self['description']
+
+    def __eq__(self, other):
+        # consider 2 annotations equal if they belong to the same task, and
+        # their data dics are the same
+        return self.task == other.task and self._data == other._data
+
+    __repr__ = __unicode__
+
+
+class Task(TaskResource):
+    read_only_fields = ['id', 'entry', 'urgency', 'uuid', 'modified']
+
+    class DoesNotExist(Exception):
+        pass
+
+    class CompletedTask(Exception):
+        """
+        Raised when the operation cannot be performed on the completed task.
+        """
+        pass
+
+    class DeletedTask(Exception):
+        """
+        Raised when the operation cannot be performed on the deleted task.
+        """
+        pass
+
+    class NotSaved(Exception):
+        """
+        Raised when the operation cannot be performed on the task, because
+        it has not been saved to TaskWarrior yet.
+        """
+        pass
+
+    @classmethod
+    def from_input(cls, input_file=sys.stdin, modify=None, warrior=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
+        original state of the Task object, and second line as its new,
+        modified value. This is consistent with the TaskWarrior's hook
+        system.
+
+        Object created by this method should not be saved, deleted
+        or refreshed, as t could create a infinite loop. For this
+        reason, TaskWarrior instance is set to None.
+
+        Input_file argument can be used to specify the input file,
+        but defaults to sys.stdin.
+        """
+
+        # Detect the hook type if not given directly
+        name = os.path.basename(sys.argv[0])
+        modify = name.startswith('on-modify') if modify is None else modify
 
-    def regenerate_uuid(self):
-        self['uuid'] = str(uuid.uuid4())
+        # Create the TaskWarrior instance if none passed
+        if warrior is None:
+            hook_parent_dir = os.path.dirname(os.path.dirname(sys.argv[0]))
+            warrior = TaskWarrior(data_location=hook_parent_dir)
+
+        # TaskWarrior instance is set to None
+        task = cls(warrior)
+
+        # Load the data from the input
+        task._load_data(json.loads(input_file.readline().strip()))
+
+        # If this is a on-modify event, we are provided with additional
+        # line of input, which provides updated data
+        if modify:
+            task._update_data(json.loads(input_file.readline().strip()))
+
+        return task
+
+    def __init__(self, warrior, **kwargs):
+        self.warrior = warrior
+
+        # Check that user is not able to set read-only value in __init__
+        for key in kwargs.keys():
+            if key in self.read_only_fields:
+                raise RuntimeError('Field \'%s\' is read-only' % key)
+
+        # We serialize the data in kwargs so that users of the library
+        # do not have to pass different data formats via __setitem__ and
+        # __init__ methods, that would be confusing
+
+        # Rather unfortunate syntax due to python2.6 comaptiblity
+        self._data = dict((key, self._normalize(key, value))
+                          for (key, value) in six.iteritems(kwargs))
+        self._original_data = copy.deepcopy(self._data)
+
+        # Provide read only access to the original data
+        self.original = ReadOnlyDictView(self._original_data)
+
+    def __unicode__(self):
+        return self['description']
+
+    def __eq__(self, other):
+        if self['uuid'] and other['uuid']:
+            # For saved Tasks, just define equality by equality of uuids
+            return self['uuid'] == other['uuid']
+        else:
+            # If the tasks are not saved, compare the actual instances
+            return id(self) == id(other)
+
+
+    def __hash__(self):
+        if self['uuid']:
+            # For saved Tasks, just define equality by equality of uuids
+            return self['uuid'].__hash__()
+        else:
+            # If the tasks are not saved, return hash of instance id
+            return id(self).__hash__()
+
+    @property
+    def completed(self):
+        return self['status'] == six.text_type('completed')
+
+    @property
+    def deleted(self):
+        return self['status'] == six.text_type('deleted')
+
+    @property
+    def waiting(self):
+        return self['status'] == six.text_type('waiting')
+
+    @property
+    def pending(self):
+        return self['status'] == six.text_type('pending')
+
+    @property
+    def saved(self):
+        return self['uuid'] is not None or self['id'] is not None
+
+    def serialize_depends(self, cur_dependencies):
+        # 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)
+
+        return super(Task, self).serialize_depends(cur_dependencies)
+
+    def format_depends(self):
+        # We need to generate added and removed dependencies list,
+        # since Taskwarrior does not accept redefining dependencies.
+
+        # This cannot be part of serialize_depends, since we need
+        # to keep a list of all depedencies in the _data dictionary,
+        # not just currently added/removed ones
+
+        old_dependencies = self._original_data.get('depends', set())
+
+        added = self['depends'] - old_dependencies
+        removed = old_dependencies - self['depends']
+
+        # Removed dependencies need to be prefixed with '-'
+        return 'depends:' + ','.join(
+                [t['uuid'] for t in added] +
+                ['-' + t['uuid'] for t in removed]
+            )
+
+    def format_description(self):
+        # Task version older than 2.4.0 ignores first word of the
+        # task description if description: prefix is used
+        if self.warrior.version < VERSION_2_4_0:
+            return self._data['description']
+        else:
+            return "description:'{0}'".format(self._data['description'] or '')
 
     def delete(self):
-        self.warrior.delete_task(self['uuid'])
+        if not self.saved:
+            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:
+            raise Task.DeletedTask("Task was already deleted")
+
+        self.warrior.execute_command([self['uuid'], 'delete'])
+
+        # Refresh the status again, so that we have updated info stored
+        self.refresh(only_fields=['status', 'start', 'end'])
+
+    def start(self):
+        if not self.saved:
+            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:
+            raise Task.CompletedTask("Cannot start a completed task")
+        elif self.deleted:
+            raise Task.DeletedTask("Deleted task cannot be started")
+
+        self.warrior.execute_command([self['uuid'], 'start'])
+
+        # Refresh the status again, so that we have updated info stored
+        self.refresh(only_fields=['status', 'start'])
 
     def done(self):
-        self.warrior.complete_task(self['uuid'])
+        if not self.saved:
+            raise Task.NotSaved("Task needs to be saved before it can be completed")
 
-    def save(self, delete_first=True):
-        if self['uuid'] and delete_first:
-            self.delete()
-        if not self['uuid'] or delete_first:
-            self.regenerate_uuid()
-        self.warrior.import_tasks([self._data])
+        # Refresh, and raise exception if task is already completed/deleted
+        self.refresh(only_fields=['status'])
 
-    __repr__ = __unicode__
+        if self.completed:
+            raise Task.CompletedTask("Cannot complete a completed task")
+        elif self.deleted:
+            raise Task.DeletedTask("Deleted task cannot be completed")
+
+        self.warrior.execute_command([self['uuid'], 'done'])
+
+        # Refresh the status again, so that we have updated info stored
+        self.refresh(only_fields=['status', 'start', 'end'])
+
+    def save(self):
+        if self.saved and not self.modified:
+            return
+
+        args = [self['uuid'], 'modify'] if self.saved else ['add']
+        args.extend(self._get_modified_fields_as_args())
+        output = self.warrior.execute_command(args)
+
+        # Parse out the new ID, if the task is being added for the first time
+        if not self.saved:
+            id_lines = [l for l in output if l.startswith('Created task ')]
+
+            # Complain loudly if it seems that more tasks were created
+            # Should not happen
+            if len(id_lines) != 1 or len(id_lines[0].split(' ')) != 3:
+                raise TaskWarriorException("Unexpected output when creating "
+                                           "task: %s" % '\n'.join(id_lines))
+
+            # Circumvent the ID storage, since ID is considered read-only
+            self._data['id'] = int(id_lines[0].split(' ')[2].rstrip('.'))
+
+        # Refreshing is very important here, as not only modification time
+        # is updated, but arbitrary attribute may have changed due hooks
+        # altering the data before saving
+        self.refresh()
+
+    def add_annotation(self, annotation):
+        if not self.saved:
+            raise Task.NotSaved("Task needs to be saved to add annotation")
+
+        args = [self['uuid'], 'annotate', annotation]
+        self.warrior.execute_command(args)
+        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")
+
+        if isinstance(annotation, TaskAnnotation):
+            annotation = annotation['description']
+        args = [self['uuid'], 'denotate', annotation]
+        self.warrior.execute_command(args)
+        self.refresh(only_fields=['annotations'])
+
+    def _get_modified_fields_as_args(self):
+        args = []
+
+        def add_field(field):
+            # Add the output of format_field method to args list (defaults to
+            # field:value)
+            serialized_value = self._serialize(field, self._data[field])
+
+            # Empty values should not be enclosed in quotation marks, see
+            # TW-1510
+            if serialized_value is '':
+                escaped_serialized_value = ''
+            else:
+                escaped_serialized_value = "'{0}'".format(serialized_value)
+
+            format_default = lambda: "{0}:{1}".format(field,
+                                                      escaped_serialized_value)
+
+            format_func = getattr(self, 'format_{0}'.format(field),
+                                  format_default)
+
+            args.append(format_func())
+
+        # If we're modifying saved task, simply pass on all modified fields
+        if self.saved:
+            for field in self._modified_fields:
+                add_field(field)
+        # For new tasks, pass all fields that make sense
+        else:
+            for field in self._data.keys():
+                if field in self.read_only_fields:
+                    continue
+                add_field(field)
+
+        return args
+
+    def refresh(self, only_fields=[]):
+        # 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")
+
+        # We need to use ID as backup for uuid here for the refreshes
+        # of newly saved tasks. Any other place in the code is fine
+        # with using UUID only.
+        args = [self['uuid'] or self['id'], 'export']
+        new_data = json.loads(self.warrior.execute_command(args)[0])
+        if only_fields:
+            to_update = dict(
+                [(k, new_data.get(k)) for k in only_fields])
+            self._update_data(to_update, update_original=True)
+        else:
+            self._load_data(new_data)
+
+class TaskFilter(SerializingObject):
+    """
+    A set of parameters to filter the task list with.
+    """
+
+    def __init__(self, filter_params=[]):
+        self.filter_params = filter_params
+
+    def add_filter(self, filter_str):
+        self.filter_params.append(filter_str)
+
+    def add_filter_param(self, key, value):
+        key = key.replace('__', '.')
+
+        # Replace the value with empty string, since that is the
+        # convention in TW for empty values
+        attribute_key = key.split('.')[0]
+
+        # Since this is user input, we need to normalize before we serialize
+        value = self._normalize(attribute_key, value)
+        value = self._serialize(attribute_key, value)
+
+        # If we are filtering by uuid:, do not use uuid keyword
+        # due to TW-1452 bug
+        if key == 'uuid':
+            self.filter_params.insert(0, value)
+        else:
+            # Surround value with aphostrophes unless it's a empty string
+            value = "'%s'" % value if value else ''
+
+            # We enforce equality match by using 'is' (or 'none') modifier
+            # Without using this syntax, filter fails due to TW-1479
+            modifier = '.is' if value else '.none'
+            key = key + modifier if '.' not in key else key
+
+            self.filter_params.append("{0}:{1}".format(key, value))
+
+    def get_filter_params(self):
+        return [f for f in self.filter_params if f]
+
+    def clone(self):
+        c = self.__class__()
+        c.filter_params = list(self.filter_params)
+        return c
+
+
+class TaskQuerySet(object):
+    """
+    Represents a lazy lookup for a task objects.
+    """
+
+    def __init__(self, warrior=None, filter_obj=None):
+        self.warrior = warrior
+        self._result_cache = None
+        self.filter_obj = filter_obj or TaskFilter()
+
+    def __deepcopy__(self, memo):
+        """
+        Deep copy of a QuerySet doesn't populate the cache
+        """
+        obj = self.__class__()
+        for k, v in self.__dict__.items():
+            if k in ('_iter', '_result_cache'):
+                obj.__dict__[k] = None
+            else:
+                obj.__dict__[k] = copy.deepcopy(v, memo)
+        return obj
+
+    def __repr__(self):
+        data = list(self[:REPR_OUTPUT_SIZE + 1])
+        if len(data) > REPR_OUTPUT_SIZE:
+            data[-1] = "...(remaining elements truncated)..."
+        return repr(data)
+
+    def __len__(self):
+        if self._result_cache is None:
+            self._result_cache = list(self)
+        return len(self._result_cache)
+
+    def __iter__(self):
+        if self._result_cache is None:
+            self._result_cache = self._execute()
+        return iter(self._result_cache)
+
+    def __getitem__(self, k):
+        if self._result_cache is None:
+            self._result_cache = list(self)
+        return self._result_cache.__getitem__(k)
+
+    def __bool__(self):
+        if self._result_cache is not None:
+            return bool(self._result_cache)
+        try:
+            next(iter(self))
+        except StopIteration:
+            return False
+        return True
+
+    def __nonzero__(self):
+        return type(self).__bool__(self)
+
+    def _clone(self, klass=None, **kwargs):
+        if klass is None:
+            klass = self.__class__
+        filter_obj = self.filter_obj.clone()
+        c = klass(warrior=self.warrior, filter_obj=filter_obj)
+        c.__dict__.update(kwargs)
+        return c
+
+    def _execute(self):
+        """
+        Fetch the tasks which match the current filters.
+        """
+        return self.warrior.filter_tasks(self.filter_obj)
+
+    def all(self):
+        """
+        Returns a new TaskQuerySet that is a copy of the current one.
+        """
+        return self._clone()
+
+    def pending(self):
+        return self.filter(status=PENDING)
+
+    def completed(self):
+        return self.filter(status=COMPLETED)
+
+    def filter(self, *args, **kwargs):
+        """
+        Returns a new TaskQuerySet with the given filters added.
+        """
+        clone = self._clone()
+        for f in args:
+            clone.filter_obj.add_filter(f)
+        for key, value in kwargs.items():
+            clone.filter_obj.add_filter_param(key, value)
+        return clone
+
+    def get(self, **kwargs):
+        """
+        Performs the query and returns a single object matching the given
+        keyword arguments.
+        """
+        clone = self.filter(**kwargs)
+        num = len(clone)
+        if num == 1:
+            return clone._result_cache[0]
+        if not num:
+            raise Task.DoesNotExist(
+                'Task matching query does not exist. '
+                'Lookup parameters were {0}'.format(kwargs))
+        raise ValueError(
+            'get() returned more than one Task -- it returned {0}! '
+            'Lookup parameters were {1}'.format(num, kwargs))
 
 
 class TaskWarrior(object):
-    DEFAULT_FILTERS = {
-        'status': 'pending',
-    }
+    def __init__(self, data_location='~/.task', create=True, taskrc_location='~/.taskrc'):
+        data_location = os.path.expanduser(data_location)
+        self.taskrc_location = os.path.expanduser(taskrc_location)
 
-    def __init__(self, data_location='~/.task', create=True):
-        if not os.path.exists(data_location):
+        # If taskrc does not exist, pass / to use defaults and avoid creating
+        # dummy .taskrc file by TaskWarrior
+        if not os.path.exists(self.taskrc_location):
+            self.taskrc_location = '/'
+
+        if create and not os.path.exists(data_location):
             os.makedirs(data_location)
         self.config = {
             'data.location': os.path.expanduser(data_location),
+            'confirmation': 'no',
+            'dependency.confirmation': 'no',  # See TW-1483 or taskrc man page
+            'recurrence.confirmation': 'no',  # Necessary for modifying R tasks
         }
+        self.tasks = TaskQuerySet(self)
+        self.version = self._get_version()
+
+    def _get_command_args(self, args, config_override={}):
+        command_args = ['task', 'rc:{0}'.format(self.taskrc_location)]
+        config = self.config.copy()
+        config.update(config_override)
+        for item in config.items():
+            command_args.append('rc.{0}={1}'.format(*item))
+        command_args.extend(map(str, args))
+        return command_args
+
+    def _get_version(self):
+        p = subprocess.Popen(
+                ['task', '--version'],
+                stdout=subprocess.PIPE,
+                stderr=subprocess.PIPE)
+        stdout, stderr = [x.decode('utf-8') for x in p.communicate()]
+        return stdout.strip('\n')
 
-    def _generate_command(self, command):
-        args = ['task', 'rc:/']
-        for item in self.config.items():
-            args.append('rc.{0}={1}'.format(*item))
-        args.append(command)
-        return ' '.join(args)
-
-    def _execute(self, command):
-        p = subprocess.Popen(self._generate_command(command), shell=True,
-                             stdout=subprocess.PIPE, stderr=subprocess.PIPE)
-        stdout, stderr = p.communicate()
-        if p.returncode:
-            raise TaskWarriorException(stderr.strip())
+    def execute_command(self, args, config_override={}, allow_failure=True):
+        command_args = self._get_command_args(
+            args, config_override=config_override)
+        logger.debug(' '.join(command_args))
+        p = subprocess.Popen(command_args, stdout=subprocess.PIPE,
+                             stderr=subprocess.PIPE)
+        stdout, stderr = [x.decode('utf-8') for x in p.communicate()]
+        if p.returncode and allow_failure:
+            if stderr.strip():
+                error_msg = stderr.strip()
+            else:
+                error_msg = stdout.strip()
+            raise TaskWarriorException(error_msg)
         return stdout.strip().split('\n')
 
-    def _format_filter_kwarg(self, kwarg):
-        key, val = kwarg[0], kwarg[1]
-        if key in ['tag', 'tags']:
-            key = 'tags.equal'
-        key = key.replace('__', '.')
-        return '{0}:{1}'.format(key, val)
-
-    def get_tasks(self, **filter_kwargs):
-        filters = self.DEFAULT_FILTERS
-        filters.update(filter_kwargs)
-        filter_commands = ' '.join(map(self._format_filter_kwarg,
-                                       filters.items()))
-        command = '{0} export'.format(filter_commands)
+    def enforce_recurrence(self):
+        # Run arbitrary report command which will trigger generation
+        # of recurrent tasks.
+
+        # Only necessary for TW up to 2.4.1, fixed in 2.4.2.
+        if self.version < VERSION_2_4_2:
+            self.execute_command(['next'], allow_failure=False)
+
+    def filter_tasks(self, filter_obj):
+        self.enforce_recurrence()
+        args = ['export', '--'] + filter_obj.get_filter_params()
         tasks = []
-        for line in self._execute(command):
+        for line in self.execute_command(args):
             if line:
-                tasks.append(Task(self, json.loads(line.strip(','))))
+                data = line.strip(',')
+                try:
+                    filtered_task = Task(self)
+                    filtered_task._load_data(json.loads(data))
+                    tasks.append(filtered_task)
+                except ValueError:
+                    raise TaskWarriorException('Invalid JSON: %s' % data)
         return tasks
 
-    def get_task(self, task_id):
-        command = '{0} export'.format(task_id)
-        return Task(self, json.loads(self._execute(command)[0]))
-
-    def add_task(self, description, project=None):
-        args = ['add', description]
-        if project is not None:
-            args.append('project:{0}'.format(project))
-        self._execute(' '.join(args))
-
-    def delete_task(self, task_id):
-        self._execute('{0} rc.confirmation:no delete'.format(task_id))
-
-    def complete_task(self, task_id):
-        self._execute('{0} done'.format(task_id))
+    def merge_with(self, path, push=False):
+        path = path.rstrip('/') + '/'
+        self.execute_command(['merge', path], config_override={
+            'merge.autopush': 'yes' if push else 'no',
+        })
 
-    def import_tasks(self, tasks):
-        fd, path = tempfile.mkstemp()
-        with open(path, 'w') as f:
-            f.write(json.dumps(tasks))
-        self._execute('import {0}'.format(path))
+    def undo(self):
+        self.execute_command(['undo'])